Writers of Pro Football Prospectus 2008

Most Recent FO Features

RiceSid07.jpg

» Sidney Rice: What Could Have Been?

Sidney Rice has retired. Is he the most random single-season DYAR leader ever? One-year wonder? Injury prone? We offer a career retrospective for the second-best wide receiver named Rice in NFL history.

10 Aug 2006

Never Punting

Guest column by Jason Scheib

(Ed. note: There are many writers at Football Outsiders, and sometimes its hard to convince people that an article on our site represents "the author says X" rather than "Football Outsiders says X." This is extra true for guest columns. This article should not be taken to mean "Football Outsiders says never punt." But Jason's theory certainly is intriguing and unconventional, and we thought it was worth consideration and discussion. We hope you'll find it as thought-provoking as we did. Be prepared, this article is REALLY long.)

Introduction

When a football team fails to convert on third down, it usually punts. And when it punts, it is turning the ball over to the other team. So why isn't a punt considered a turnover, just like an interception or fumble?

I started with this simple idea and began exploring it as far and in as many different directions as it would take me. Over time, it has grown into a theory that redefines a turnover and uses this new definition to see what a team can do to improve their net turnovers and win more games. This theory presents two significant implications:

  • A team would win more games if they never punted, and
  • A team that never punts would not just be employing a different strategy but would approach the game in a fundamentally different way, which would further add to their success.

This is not about taking more risks and punting less often. That could cost you games depending on when you decided to punt and when you decided not to. The key is to never punt. Never punting takes away the risk because it allows the averages to work in your favor. It also opens you up to different play calling opportunities, primarily on third down. The two go together and are dependent on each other in order to make this work.

Thanks to Peter Watson for discussing these ideas with me and helping me get this theory this far.

Redefining The Turnover

Every once in a while you will hear a football announcer or commentator say that the outcome of a play on the field was, in effect, a turnover. What they are saying is that it is not a turnover -- at least by definition -- but had the same effect. For example, when a team goes for it on fourth down and fails, it is commonly called a turnover on downs -- but it is not counted as a turnover in the stats.

A turnover is traditionally defined as an interception or a lost fumble, and therefore those are the only events that are included in the turnover ratio. And because of that, it can sometimes be difficult to use the turnover ratio as an indication of how good a team is, considering that interceptions and fumbles are often the result of a bad bounce of the ball. A team may have a good record because of a few lucky bounces and not really because they are as good as their record. Interceptions and fumbles are looked at as momentum changers and turning points in games, and for good reason; they often involve good field position for the team forcing the turnover. But there is no fundamental difference between these events and a turnover on downs.

Or a punt for that matter. A punt may be a controlled turnover, done because of the perceived benefit in field position, but punting is giving the ball to the other team. And in fact it is common for people to say that a team forces the other team to punt. Not that they mean a punt is involuntary, but the effect is the same as forcing any other kind of turnover.

So let's redefine a turnover as ANY TIME a drive ends by giving the ball to the other team without scoring. This includes interceptions and lost fumbles along with punts, turnovers on downs, and missed (or blocked) field goals. It also includes a successful onside kick as a turnover against the receiving team (I will explain the reasoning for this later on). And so consider the following: If a turnover is defined as any time you give the ball to the other team without scoring, scoring is then defined as not turning the ball over. One or the other happens as the result of each drive. And so we can say that fewer turnovers equals more points, and also that fewer turnovers by the opposition equals more points against you. So then, since both teams have the same number of drives in a game (plus or minus one) the team that turns the ball over less by definition scores on more drives and so tends to win.

If this is a valid definition then we could expect to see a correlation between a team's net turnovers and their win-loss record. To measure net turnovers, let's add these other items into the turnover ratio and see what we get. Here are the results for what I call Actual Turnover Ratio (ATR) for the 2004 NFL season:

Rank Team ATR Wins Rank Team ATR Wins
1 NE 34 14 17 CAR 0 7
2 IND 31 12 18 NYG -1 6
3 PHI 30 13 19 HOU -3 7
4 NYJ 28 10 20 DET -3 6
5 PIT 23 15 21 WAS -4 6
6 DEN 19 10 22 JAC -7 9
7 SD 18 12 23 ARI -8 6
8 BUF 6 9 24 CLE -12 4
9 BAL 5 9 25 MIA -15 4
10 KC 5 7 26 NO -17 8
11 SEA 4 9 27 OAK -17 5
12 MIN 4 8 28 TEN -18 5
13 GB 3 10 29 DAL -20 6
14 CIN 3 8 30 STL -24 8
15 ATL 1 11 31 CHI -28 5
16 TB 1 5 32 SF -38 2

And here is a sample of what the full breakdown of these numbers looks like for the top and bottom teams and the league average:

OFFENSE (Giveaways) DEFENSE (Takeaways)
Team Punts Fum/Int TO on Downs Miss FG Ons vs. Punts Fum/Int TO on Downs Miss FG Ons For Off. TO Def. TO ATR Wins
NE 56 27 6 2 0 69 36 17 3 0 91 125 34 14
League Average 78.53 28.25 7.34 5.22 .41 78.53 28.25 7.34 5.22 .41 119.75 119.75 0 8
SF 96 40 5 4 0 79 21 3 4 0 145 107 -38 2

I'm not a statistician, but in looking at these results there does appear to be a good correlation between Actual Turnover Ratio and wins.

(Ed. Note: I'm not a statistician either, but I play one on TV. The correlation coefficient is 0.834, so there's a real correlation.)

For the most part, the teams with the most wins are at the top and the teams with the fewest wins are at the bottom. But that's just one season. Here is the average ATR vs. wins for the 2002-2004 NFL seasons (not counting successful onside kicks, which are rare; I wasn't able to pour through the play-by-play from 2002-2003):

ATR Wins
31 to 35 12.3
26 to 30 11.8
21 to 25 12.3
16 to 20 11.0
11 to 15 9.8
6 to 10 9.2
1 to 5 8.1
-4 to 0 7.2
-9 to -5 7.6
-14 to -10 5.7
-19 to -15 5.6
-24 to -20 5.4
-29 to -25 4.0
-34 to -30 4.5
-39 and less 3.0

So the statistics match our expectations. And it shouldn't be a surprise, considering that Actual Turnover Ratio is really just the inverse of net points, and net points are essentially what determines whether you win or lose a game. This doesn't discount the role of field position in a game; where a team begins a drive will affect how likely they are to either score or turn the ball over on that drive. But the question needs to be asked: which is more significant, the field position or the fact that the ball was turned over in the first place? Because if Actual Turnover Ratio is any indication, the turnover itself seems to be more significant. And that could have a serious implication in regards to punts. But before we look at punts, there are some other things to look at first...

Onside Kicks, Implied Drives, and Safeties

What about kickoffs? Are kickoffs turnovers? After all, when a team kicks off they have possession of the ball until they kick. So, then, you could say that when you score you maintain possession. The rules of the game require that you then kick off. The rules don't require that you give possession to the other team. If that were the case, then onside kicks would not be legal. So in theory, you could onside kick every time and if successful every time you could keep the other team's offense from ever being on the field. Of course that's not realistic, but in theory you could.

But do we count kickoffs into Actual Turnover Ratio? On a drive you either turn the ball over or score. Now if we count kickoffs into that equation we then would have to say that when you do score you turn the ball over anyway unless you kick onside and recover. So then, almost every drive would end in a turnover. Or, we could say that after you score, it is a new possession. Except that you don't have a first and ten, you have to kick. The difficulty with this is that only turnovers count into ATR, so the fact that you score on one possession and then turn it over on the next 'possession' (kickoff) still results in one turnover towards the ratio and so that is no different either way you word it. And so we go back to the definition of a turnover, where a drive ends by giving the ball to the other team without scoring. So kickoffs can't be turnovers.

Here's something to think about though: what if we looked at a successful onside kick as a turnover against the receiving team? It seems backwards of what we've been saying. How can a team that doesn't have possession turn the ball over? But when a successful onside kick happens against your team, it sure feels like you turned the ball over. And there's a reason why it feels that way.

It is implied in the nature of the game of football that when you score the next drive belongs to the other team. When your team successfully recovers their onside kick attempt, the other team loses their opportunity at a drive. But it's more than that. It is implied in the nature of the game that the next drive after your score was theirs. And so it is implied that they had a drive. Granted, lost "drive" is one in which they never played a down, but that lost drive still had a result. And since that drive did not end in a score, it by definition ended in a turnover. An implied turnover. That's why a successful onside kick feels like a turnover -- because it is one. This is why a successful onside kick can give you a chance to get back in a game at the end: because it closes the gap in the game's Actual Turnover Ratio by one.

Implied drives can be used to explain more than just successful onside kicks. For a kickoff, punt, fumble recovery, or interception returned for a touchdown, since the return team (or recovering team) did not have a drive, and it was implied that the next drive belonged to them, the implied drive ended with a score instead of a turnover. And so implied drives are counted as drives in figuring Actual Turnover Ratio because they have an outcome of either a score or a turnover.

Safeties are a little trickier to explain, because the team that gets the points didn't have the ball, and also because they do not then kick off but receive the ball from the other team. But a turnover does not take place at any point in a safety. A drive ends with either a turnover or a score. You have the ball and are sacked for a safety. So you have a drive and it ended in a score. Not a score for you, but the drive still ended with a score, still fitting the definition. As far as the definition is concerned, it doesn't matter who scored, just that the drive ended in a score. And because your drive ended with a score it is implied that the next drive belongs to the other team. By that definition it is no different than a touchdown, which would also be your drive ending with a score and so the next drive belongs to the other team. And in that sense there isn't a turnover on a safety any more than there is when you kick off after a touchdown.

To Punt Or Not To Punt

If a better Actual Turnover Ratio means more wins, the question is: What can a team do to improve their ratio? In taking a look at the components of Actual Turnover Ratio, some of what a team can do are things they already try to do, such as correcting mistakes, trying to find an accurate field goal kicker, or "forcing" the opponents to punt. But the one thing that stands out the most is that punts are not only the one component that is done voluntarily, but also the component that occurs most often. So the question is now: If a team never punts, can they improve their turnover ratio? Let's take a look at what never punting would do to ATR.

If a team were to never punt, the offensive side (giveaways) of ATR would improve significantly. While punts drop to zero, turnovers on downs would increase, but only by the amount of unsuccessful fourth down attempts. On the defensive side (takeaways), the only component that significantly changes is punts to your team. This is because the other team would be less likely to punt if they had more drives start in better field position. What that means is that the odds of them turning the ball back over to you without scoring is lower. So by not punting, only two other components of your ATR are affected by any significant degree, one on offense and one on defense. On both sides of the ratio the number of turnovers would go down, which means by definition both you and your opponents would score more. The question is which side of the ratio is affected more. If we can show that your giveaways would go down more than your takeaways, the implication would be that by never punting you would win more games.

To find the answer to this, I went through the play-by-play of every game of the 2004 NFL season and calculated the points scored after every drive result. In other words, how likely was a team to score if they received the ball via a kickoff, punt, interception or fumble, etc., and also what was the average number of points per score. These numbers are meaningful because according to our new definition, a score is the inverse of a turnover. These point values came out as follows:

Points Per Score 5.54
Ball Received Via:
Kickoff 1.59
Punt 1.65
Fumble or Interception 2.80
Turnover on Downs 1.42
Missed or Blocked Field Goal 1.33
Successful Onside Kick 2.08

These point values reflect the average field position your team starts from. Higher values mean you have a better average starting field position than lower values. The turnover on downs point value is skewed because most of those are in short yardage situations in opponent's territory. For the purpose of looking at additional turnovers on downs in place of never punting, it makes more sense to go with the 2.80 of fumbles and interceptions, since those can happen anywhere on the field, in your territory or opponent's territory and averaging somewhere around midfield. You could then look at these numbers and say, of course you would want to punt, the numbers prove you are more likely to be scored upon if you go for it and don't make it. But let's look at it more in context.

I also kept track of the number of times each of these events happened per team per game. The average number of drives per team per game was 12.12, with 11.21 having a result (score or turnover) and the other .91 not having a result because time ran out at the end of either half. Out of those 11.21 drives, 3.88 end in scores, 4.88 end in punts, 1.71 end in fumbles or interceptions, 0.38 in turnovers on downs, 0.33 in missed field goals, and 0.03 in successful onside kicks against. And so we can put together the following chart:

Number Points Per Total Points
Result of a Team's Drives:
Scores 3.88 5.54 21.4952
Punts 4.88 0 0
Fumbles/Interceptions 1.71 0 0
Turnovers on Downs 0.38 0 0
Missed Field Goals 0.33 0 0
Successful Onside Kicks Against 0.03 0 0
No Result (End of Either Half) 0.91 0 0
Opponent Receives Ball Via:
Punts 4.88 1.65 8.052
Fumbles/Interceptions 1.71 2.80 4.788
Turnovers on Downs 0.38 1.42 0.5396
Missed Field Goals 0.33 1.33 0.4389
Successful Onside Kicks Against 0.03 2.08 0.0624
Kickoffs 4.79 1.59 7.6161

The total of the points in the 'Opponent Receives Ball Via' section equals the same as the points scored or about 21.5 (since this is just a league average, points scored equals points against).

So what happens if you don't punt? Of course, just because you are successful on a fourth down attempt doesn't necessarily mean you end up scoring on that drive; you could still turn the ball over. But even if you only ended up scoring one out of those approximately five drives where you now go for it instead, this is what the chart would then look like -- if your opponents still punted. I added an additional item in the chart to count for the additional turnovers on downs that we are figuring at the higher 2.80 point value:

Number Points Per Total Points
Result of a Team's Drives:
Scores 4.88 5.54 27.0352
Punts 0 0 0
Fumbles/Interceptions 1.71 0 0
Turnovers on Downs (in Place of Punts) 3.88 0 0
Turnovers on Downs 0.38 0 0
Missed Field Goals 0.33 0 0
Successful Onside Kicks Against 0.03 0 0
No Result (End of Either Half) 0.91 0 0
Opponent Receives Ball Via:
Punts 0 1.65 0
Fumbles/Interceptions 1.71 2.80 4.788
Turnovers on Downs (in Place of Punts) 3.88 2.80 10.8644
Turnovers on Downs 0.38 1.42 0.53966
Missed Field Goals 0.33 1.33 0.43899
Successful Onside Kicks Against 0.03 2.08 0.06244
Kickoffs 5.79 1.59 9.20611

So out of 4.88 drives where your team would traditionally have punted, if now you score on just one of those drives, the other 3.88 count as turnovers on downs at 2.8 points each. And since you score one more time you kick off one more time. So now you would score 27.0352 points and your opponents would score 25.899 (the total of the point values for your opponent's drives).

We can also invert these charts to show the number of turnovers on each drive, since a turnover is the opposite of a score. The number of turnovers for each drive is, in effect then, the percent of times a team will turn the ball over when receiving the ball each different way. And since you don't turn the ball over every time you don't score because of time running out at the end of a half, we can figure in from these numbers the fact that 89% of drives have a result and so 89% of the 5.54 points per score (or 4.93) is the inverse of a turnover. And so the chart would look like this:

Number Turnovers For Each Total Turnovers
Result of
a Team's Drives:
Scores 3.88 0 0
Punts 4.88 1 4.88
Fumbles/Interceptions 1.71 1 1.71
Turnovers on Downs 0.38 1 0.38
Missed Field Goals 0.33 1 0.33
Successful Onside Kicks Against 0.03 1 0.03
No Result (End of Either Half) 0.91 0 0
Opponent Receives Ball Via:
Punts 4.88 0.625 3.05
Fumbles/Interceptions 1.71 0.441 0.75
Turnovers
on Downs
0.38 0.662 0.25
Missed Field Goals 0.33 0.676 0.22
Successful Onside Kicks Against 0.03 0.556 0.02
Kickoffs 4.79 0.635 3.04

The total turnovers are the same both for the total of the results of your drives and your opponents drives at 7.33. And as in the example before, if by not punting you ended up scoring on one more drive than before the chart would look like this:

Number Turnovers For Each Total Turnovers
Result of a Team's Drives:
Scores 4.88 0 0
Punts 0 1 0
Fumbles/Interceptions 1.71 1 1.71
Turnovers on Downs (in Place of Punts) 3.88 1 3.88
Turnovers on Downs 0.38 1 0.38
Missed Field Goals 0.33 1 0.33
Successful Onside Kicks Against 0.03 1 0.03
No Result (End of Either Half) 0.91 0 0
Opponent Receives Ball Via:
Punts 0 0.625 0
Fumbles/Interceptions 1.71 0.441 0.75
Turnovers on Downs (in Place of Punts) 3.88 0.441 1.71
Turnovers on Downs 0.38 0.662 0.25
Missed Field Goals 0.33 0.676 0.22
Successful Onside Kicks Against 0.03 0.556 0.02
Kickoffs 5.79 0.635 3.68

By scoring on one additional drive, your turnovers now total 6.33 and your opponent's turnovers total 6.63. And that's with only about a 20 percent success rate of converting the fourth down attempts and then scoring on that drive. The higher that success rate goes, the larger the point difference (and turnover difference) between you and your opponents.

Although the 20 percent success rate seems conservative, we need to figure out what that success rate would actually be. But before we do that, there is an important point to consider: If a team never punts, their offensive strategy completely changes. The difficulty in trying to figure out what a team's statistics would look like if they never punted is that a team that never punts would have very different stats than the stats we currently see. And that is because this team would be playing a somewhat different game. So let's take a look at how a team that never punts would approach the game differently...

Four Down Mindset

Everybody knows that in football you have four downs to gain 10 yards or you lose control of the ball. But everybody approaches football as if you only have three downs to gain 10 yards or -- most of the time -- you punt. There may be exceptions to that when a team gets close to the end zone, but in general that is the primary approach. Think about that. Convention says that you are better off punting. And maybe that's true if you approach the game as if you only have three downs. The difference is mindset.

That's the difference between saying you should be more aggressive and punt less often, depending on the situation, and saying that you should not punt at all. The first statement is based upon still approaching the game with a three down mindset. The second statement is based on approaching the game with a four down mindset.

So what exactly is four down mindset? It means you look at EVERY first-and-10 as if you have four tries for a first down instead of three. After all, the rules of the game say you have four tries.

How does this affect strategy? Well, primarily it affects what you do on third down. On third down, instead of having the mindset that you need to convert or the punting unit comes in, you have the mindset that all you need to do is get closer to the first down marker so as to put you in a better position to convert on fourth down. So where a third-and-long now usually means a pass, you could attempt a run instead to make it fourth-and-short. On purpose. And so the defense couldn't just key on the pass in that situation. In effect, you are giving yourself one more down, which may sound obvious -- but what that really means is that third down has now become no different than what second down used to be. On a conventional second down you may just try to put yourself in a better position to convert on the next down. Now you can still approach it that way on third down and you just gave yourself an extra down in between.

Would teams really just try to improve their odds of converting on fourth down rather than trying to still make the conversion on third down? Consider this: It's third-and-10. If you attempt a 10-yard pass and fail to convert, you are faced with a 10-yard conversion attempt on fourth down. If you attempt a run to set up a shorter fourth down attempt and gain no yardage, you are still faced with a 10-yard conversion attempt. Some of the times you will put yourself in shorter yardage to get a first down, and the worst case is that you are in the same position as the first example. So this strategy would improve your odds of converting.

Because you have given yourself an extra down, you can get more out of your offense. You could take a bad offensive team and get more production out of it. Or you could take a good offense and make it great. For instance, a running back that averages 2.5 yards per carry would now be just as effective to your offense as one that averages 3.3 yards per carry in a three down offense. But also think of the effect it would have on an opposing defense, knowing that they would have to stop you four times EVERY time you had a first and 10 instead of the normal three times. That could wear on them.

I mentioned that stats would be different than the ones we currently see. You could expect third down conversion rate to drop, since you are not always trying to convert on third down. You could also expect fourth down conversion rate to drop. The reason for this is that teams now generally only go for it on fourth down in short yardage situations and so the conversion rate is higher. And yet overall there would be more first downs. Why would you convert more often if both third down conversion rate and fourth down conversion rate drop? I contend that fourth down conversion rate is misrepresented. Should it not take into consideration what happens on every fourth down and not just the ones where you go for it? It would then count punts and field goal attempts as well. If you go for it and are successful, or if you make a field goal, you convert. If you go for it and are not successful, or if you punt or miss a field goal, you fail to convert. By that measure fourth down conversion rate in the NFL would have been just 24 percent rather than just under 50 percent. But let's go further still and look at the conversion rate of any first-and-10 series regardless of which down the conversion happened on. The NFL average for this rate for the 2004 season was 73 percent (more on this later). Not punting would increase that rate because while fourth down conversion rate as traditionally defined would go down, fourth down conversion rate as more accurately defined would go up significantly.

Maybe we can use Canadian football as an example of how much the difference of a down can affect a team. There are of course some notable differences between the two games: the field is 110 yards, each team has 12 men on the field and two men in motion on defense. But by far the biggest difference is that in the CFL each team has only three downs instead of four. And that difference changes the whole game. What would you expect to happen to the game if there was a change from four downs to three? More passing since it would be much harder to convert running the ball. More punts. More turnovers on downs. All of these are characteristics of the CFL. The average passing yards per team per game for the 2004 CFL season was 284.1, much higher than the 210.5 for the NFL. Rush yards were lower at 100.9 compared to 116.6 for the NFL. Each team punts an average of seven times per game instead of five. Turnovers on downs are higher (.642 compared to .459). Fumbles and interceptions are higher. And most surprising is that there are more than twice as many missed field goals as in the NFL. Then of course, if you have a harder time converting with one less down, attempting a long field goal may seem like your best option more often. To one degree or another, we could trace all of these differences to having only three downs. And all in all, teams in the CFL average three more turnovers per game than in the NFL (by the definition here) . If the CFL, by approaching the game with one less down, causes the game to change so significantly, is it too much of a stretch to think that approaching American football with a four down mindset would have just as significant an effect to an offense?

A Look At Third Down Statistics

Do teams really approach third down differently than second down? Or do they try to convert on first and second down just as much as on third down, but with less success because there are more yards on average to gain? Part of my argument about four down mindset was that on third down a team might -- on purpose -- just try to get closer to the first down marker to increase their odds of converting on fourth down. But then I can‘t imagine a coach ever saying that they don‘t try just as hard to convert on second down, and that on second down they just try to get closer to the first down marker as to increase their odds of converting on third down. And that thought bugged me. What also bugged me was the thought that maybe I'm exaggerating the effect of four down mindset. In how many situations would it really affect your approach? Maybe it wouldn't affect third and short because you are still just as likely to run the ball then, and maybe it wouldn't affect third and very long because then you would still be inclined to pass anyway. And so maybe it would only affect play calling in the third and, say, 7-to-10 yard range, which doesn't seem that significant.

I realized I needed to look at some statistics to see if my argument was valid. And so I compared the play calling (run vs. pass) of second down and third down (sacks and quarterback scrambles are counted as passing plays). These statistics come from the play-by-play of 32 games during the last two weeks of the 2004 NFL season. While not as large of a sample size as a whole season, it does take into account every team's abilities and tendencies, and does give a good indication of the difference in play calling between the two downs. And the difference is significant:

Second Down Third Down
Yards to Go Plays Runs % Passes % Yards to Go Plays Runs % Passes %
1 61 43 70.5 18 29.5 1 100 67 67.0 33 33.0
2 73 44 60.3 29 39.7 2 88 35 39.8 53 60.2
3 69 41 59.4 28 40.6 3 68 13 19.1 55 80.9
4 77 38 49.4 39 50.6 4 67 12 17.9 55 82.1
5 118 66 55.9 52 44.1 5 69 5 7.2 64 92.8
6 113 56 49.6 57 50.4 6 63 8 12.7 55 87.3
7 125 52 41.6 73 58.4 7 67 3 4.5 64 95.5
8 117 37 31.6 80 68.4 8 58 7 12.1 51 87.9
9 106 37 34.9 69 65.1 9 53 4 7.5 49 92.5
10 311 122 39.2 189 60.8 10 80 1 1.3 79 98.8
11 39 15 38.5 24 61.5 11 28 1 3.6 27 96.4
12 33 12 36.4 21 63.6 12 21 3 14.3 18 85.7
13 20 6 30.0 14 70.0 13 19 2 10.5 17 89.5
14 14 7 50.0 7 50.0 14 16 2 12.5 14 87.5
15 26 10 38.5 16 61.5 15 28 0 0.0 28 100.0
Over 15 49 19 38.8 30 61.2 Over 15 55 9 16.4 46 83.6

On second down, teams called 605 running plays to 746 pass plays, while on third down teams called only 172 running plays to 708 pass plays. Which is 44.8 percent runs on second down and only 19.5 percent runs on third down. But even more significant is the fact that this discrepancy in play calling on third down took place for every yards-to-go situation. So this answered both of my concerns: not only does this show that teams do use a different approach on third down than on second, but it also shows that the difference in approach is just as significant in the third and short and third and very long as it is in the third and 7-to-10 range.

I also calculated the average number of yards gained for runs and passes in each situation. The average yards gained on a run was 4.22 on second down and 4.63 on third down. The average yards gained on a pass was 5.53 on second down and 5.13 on third down. Short-yardage rushing was about the same: the average yards gained on a run when there are three or fewer yards to go was 3.88 on second down and 3.85 on third down. While there may be a perception that running the ball on third and short is tougher than on second and short, the statistics show otherwise. Note also that teams pretty much ALWAYS pass on third-and-10 (98.8 percent of the time). There were only 19 conversions out of 80 when passing on third and 10. Running more often on third down would increase your odds of eventually converting on that series.

We can figure out how many more times teams would run the ball for each yards to go situation if they approached third down with the same play calling tendencies as on second down:

3rd And: Plays Runs Passes Increase in Runs % Change
1 100 71 29 4 4.0%
2 88 53 35 18 20.5%
3 68 40 28 27 39.7%
4 67 33 34 21 31.3%
5 69 39 30 34 49.3%
6 63 31 32 23 36.5%
7 67 28 39 25 37.3%
8 58 18 40 11 19.0%
9 53 18 35 14 26.4%
10 80 31 49 30 37.5%
11 28 11 17 10 35.7%
12 21 8 13 5 23.8%
13 19 6 13 4 21.1%
14 16 8 8 6 37.5%
15 28 6 22 6 21.4%
Over 15 55 21 34 12 21.8%

The total change in play calling would then be 250 more runs for an overall percent increase of 28.4 percent. And with the average yards gained per run being less than the average yards gained per pass play, that would then drop the average yards gained per third down play from 5.03 to 4.89. And since we expected the change in third down play calling to cause third down conversion rate to drop somewhat, this change in yardage gained fits that expectation. Now we can see if this information can help us see how successful a team would be by never punting and approaching the game with a four down mindset.

Connecting The Dots

Earlier I mentioned a statistic that measures the conversion rate on any first-and-10 series. This conversion rate, not surprisingly, correlates well to a team's total offensive turnovers, as indicated by the following chart (using statistics from 2002-2004):

Conversion Rate For All Downs Offensive Turnovers
.800+ 88.3
.775-.779 101.6
.750-.774 108.3
.725-.749 115.5
.700-.724 124.5
.675-.699 133.6
.650-.674 139.3
under .650 162.8

What we want to do now is try and see if we can calculate what a team's total conversion rate would become if the team never punted and approached the game with a four down mindset. Then we want to see if that decrease in offensive turnovers (giveaways) is greater than the decrease in defensive turnovers (takeaways).

First let's look at the breakdown of how this conversion rate is figured. Every first-and-10 ends in either a score, a turnover, or a non-scoring first down. The only difference between a first-and-10 (series) and a drive is that you can succeed without necessarily scoring. So the number of series equals: offensive turnovers + field goals + total first downs (a TD counts as a first down but a FG does not). Since scores and non-scoring first downs are conversions, we divide the total of those two by the total of all three for a league average conversion rate of .730. Here are the numbers for the 2004 NFL season:

Offensive Turnovers FG Made Total First Downs Total Series Total Conversion Rate
3819 703 9614 14136 .730

I have said that third down conversion rate would drop from its average of .376, but it wouldn't drop quite to the level of second down conversion rate because on third down you have on average fewer yards to go to get a first down than on second down. So we know it would be somewhere in the middle. From the second down statistics I mentioned earlier for Weeks 16 and 17 of the 2004 NFL season that second down conversion rate was .315. That should be pretty close to the average for an entire season, considering the third down conversion rate of those two weeks was .380, which is pretty close to .376. To be consistent we'll use the .380 rate. By figuring, as I said, that 5.03 yards per play on third down led to a .380 conversion rate, then with a little algebra we can figure the conversion rate with 4.89 yards per play:

.380 / 5.03 = x / 4.89
x = .369

I'm not a mathematician, so this is the best way I know of to figure this rate even if it‘s not the best way. But that said, this .369 rate is in the range of what we expected, being between the second down conversion rate and the third down conversion rate. And more than that, a change in yards per play of only .14 of a yard shouldn't have very much effect on the conversion rate, so we would expect for it to only drop a little bit. Even if this rate of .369 isn't exact, it should be pretty close to it. I think it's reasonable that for our purposes we can then use this rate as our new third down conversion rate under a four down mindset. The league total of 2525 third down conversions would then become 2476 and lead to 49 more fourth downs. To figure what the fourth down conversion rate would be, we know it would drop from its average of .482 when teams go for it but not down as far as third down conversion rate because the average yards to go would be less than on third down. So again, to use some algebra, if the average of 4.94 yards per play on second down leads to the average yards to go improving from 7.73 to 6.99 (or a .740 difference), then with 4.89 yards gained per play on third down we can figure the improvement of yards to go:

4.94 / .740 = 4.89 / x
x = .733

Again, even if this isn't the best way to figure this, the answer seems reasonable considering that a small change in average yards per play of .05 yards should only have a small effect on the change in yards to go. So then if there is a conversion rate of .380 when there are on average 6.99 yards to go to on third down, we can figure how much that rate would increase when the average yards to go on fourth down would be .733 yards less than on third down:

.380 / 6.99 = x / .733
x = .040
.380 + .040 = .420

This .420 rate goes along with what we expected, being in between normal third down and fourth down conversion rates, and somewhat closer to third down conversion rate, since normal fourth down conversion rate is skewed by a preponderance of fourth-and-short situations.

So back to figuring the change in the conversion rate on any down. we now use this .420 fourth down rate applied to the original 454 fourth down conversion attempts plus the additional 49 fourth downs we figured because of the decreased third down conversions, and also the 2513 punts that are now fourth down attempts. Here is the above chart broken down into conversions per down:

Offensive Turnovers FG Made Conversions on First and Second Down Conversions on Third Down Conversions on Fourth Down Total Series Total Conversion Rate
3819 703 6870 2525 219 14136 0.730

And by not punting and approaching the game with a four down mindset this chart changes to the following:

Offensive Turnovers FG Made Conversions on First and Second Down Conversions on Third Down Conversions on Fourth Down Total Series Total Conversion Rate
2820 703 6870 2476 1267 14136 .801

Which puts us in the range at the top of the chart earlier comparing the conversion rate on any down to offensive turnovers, where a rate of .800+ correlated to 88.3 turnovers. But really all we have to do is divide the 2820 offensive turnovers in that equation by 32 teams to get an exact number of 88.125. That number of turnovers divided by 16 games comes out to 5.51 offensive turnovers per game. So now let's put that number into the chart showing the number of turnovers per drive result and work backwards. The total of 5.51 turnovers per game causes the turnovers on downs (in place of punts) to be 3.06, since that is the only component on the offensive side of Actual Turnover Ratio other than punts affected when you don't punt.

Number Turnovers For Each Total Turnovers
Result of a Team's Drives:
Scores 5.70 0 0
Punts 0 1 0
Fumbles/Interceptions 1.71 1 1.71
Turnovers on Downs (in Place of Punts) 3.06 1 3.06
Turnovers on Downs 0.38 1 0.38
Missed Field Goals 0.33 1 0.33
Successful Onside Kicks Against 0.03 1 0.03
No Result (End of Either Half) 0.91 0 0
Opponent Receives Ball Via:
Punts 0 0.625 0
Fumbles/Interceptions 1.71 0.441 0.75
Turnovers on Downs (in Place of Punts) 3.06 0.441 1.35
Turnovers on Downs 0.38 0.662 0.25
Missed Field Goals 0.33 0.676 0.22
Successful Onside Kicks Against 0.03 0.556 0.02
Kickoffs 6.61 0.635 4.20

The total turnovers now come out to 5.51 for your team, like we said, but 6.79 turnovers for your opponent. Multiplied out over a 16-game season and rounded to the nearest turnover, that becomes 88 turnovers for you and 109 for your opponents, or an Actual Turnover Ratio of +21. And according to the chart shown toward the beginning, this correlates to approximately 11 or 12 wins. And if so, then by never punting and approaching the game with a four down mindset, an 8-8 team could improve their record by approximately 3 to 4 wins on average. And that's significant.

We can also plug these turnover numbers into the chart showing points per drive result. The outcome there is 31.6 points per game for you and 24.9 points per game for your opponent. Why does this work? Because the difference between giving the ball to your opponent by punting versus turning it over on downs is only about 1 point against you (each) on average, and the increase in your scoring more than makes up for that.

And, as I mentioned earlier, even if my calculations for what third down conversion rate and fourth down conversion rate would be is off somewhat, we know at the very least that third down conversion rate would be higher than the normal second down conversion rate of .315 and that fourth down conversion rate would be higher than the normal third down conversion rate of .380. But even if we wanted to be extra conservative and use .315 as third down conversion rate and .380 for fourth down conversion rate, even that is enough to make never punting work in your favor. If we use those two rates to figure the conversion rate on any down, we end up with a rate of .776 and an average of offensive turnovers per team of 98.91,which plugged into the above chart comes out to 99 offensive turnovers to 110 opponent's turnovers, or an ATR of about +11. And those are unrealistically low conversion rates and still a significant improvement.

Below Average Offenses And Below Average Defenses

Even if never punting and approaching the game with a four down mindset can be shown to help a team win more games on average, the problem is that most teams aren't average. Many teams are strong on offense and weak on defense, or weak on offense and strong on defense. And some teams aren't strong in either category.

To make sure this works for all kinds of teams and not just the ones with at least an average offense and average defense, I did all of the same calculations for three individual teams that I did for the league average. Those three teams were the 2004 Minnesota Vikings (strong on offense and weak on defense), the 2004 Baltimore Ravens (weak on offense and strong on defense), and the 2004 San Francisco 49ers (weak in both offense and defense and the lowest ATR in the league).

Without taking up a lot of space to show all of those teams' individualized numbers, here were each team's results:

  • The 2004 Minnesota Vikings had 86 offensive turnovers, best in the league, and 90 defensive turnovers, among the worst in the league, and finished the season with an ATR of +4 and an 8-8 record. If they were to have never punted and played with a four down mindset, their conversion rate on any down would have increased from .811 to .868 while their offensive turnovers decreased to 58 and their defensive turnovers decreased to 88, for an ATR of +30. This would be an increase in ATR of +26 and approximately a 12-4 record.
  • The 2004 Baltimore Ravens had 134 offensive turnovers, among the worst in the league, and 139 defensive turnovers, among the best in the league, and finished the season with an ATR of +5 and a 9-7 record. If they were to have never punted and played with a four down mindset, their conversion rate on any down would have increased from .683 to .783 while their offensive turnovers decreased to 92 and their defensive turnovers decreased to 137, for an ATR of +45. This would be an increase in ATR of +40 and off the chart, so probably 13 wins if not more.
  • The 2004 San Francisco 49ers had 145 offensive turnovers, among the worst in the league, and 107 defensive turnovers, also among the worst in the league, and finished the season with a league worst ATR of -38 and a 2-14 record. If they were to have never punted and played with a four down mindset, their conversion rate on any down would have increased from .673 to .736 while their offensive turnovers decreased to 117 and their defensive turnovers decreased to 98, for an ATR of -19. This would be an increase in ATR of +19 and approximately 5 or 6 wins.

If this works for the 2004 49ers, who did not have a strong offense or a strong defense to bail them out when they turned the ball over on downs, it should be fair to say it would work for any team. And not just the average or above average ones.

Exceptions

Up until now I have been talking about NEVER punting. But are there some situations where it may still be in your best interest to punt? Probably yes, although ideally you would want to keep them to a minimum. For example, it's nearing the end of the game and the outcome is on the line and you find yourself facing fourth down deep in your own territory. If you go for it and don't convert the other team is in a position to kick a field goal and win the game. Or maybe because you are backed up to your own end zone facing fourth-and-20.

While it may sound like I'm contradicting myself when I say this, I'm not at all. That's because what I'm advocating here is not an offensive strategy based on risk, but rather one that is based on being conservative. The point of this turnover theory is not so much about eliminating punting as it is about finding a way to improve your Actual Turnover Ratio. The purpose of not punting is to help you win more games, but if not punting in an emergency situation would likely cost you a game then the point is still to win the game. And so any rare times you do punt would normally be at the end of a close game or in overtime.

What about field goals? Does the four down mindset suggest that you should go for it at the expense of a field goal? While some people may argue that there may be an advantage to going for a touchdown as opposed to settling for the field goal, that isn't necessary with this theory. A field goal means the drive ended in a score, and you didn't turn the ball over. And that's all we're trying to do, have more drives end in scores and fewer drives end in turnovers.

Conclusion

The calculations we made were based on the assumption that even though you never punt, your opponents still punt as normal. But it is just as likely that teams would try not punting against you, especially if it's working for you. Of course if neither team punts, the playing field is even and you no longer have an advantage. You may, however, cause the other team to punt less and just pick their spots, and that may give you an even bigger advantage.

There are two reasons for this. If a team punts on some occasions and not others the numbers don't average out the same way they do if they never punt. As I explained earlier, you need to be able to get at least one extra score per game in order to put this in your favor. But if you just punt one or two less times per game, the odds aren't very good that one of those times will end up in a score. You need to go for it each of the five times teams normally punt to get that one extra score every game. Otherwise it's hit or miss, and taking a risk. The other reason is because if you punt some times but not others then you can't really have a four down mindset even if you try to. If you aren't committed to going for it every time, it affects your third down playcalling and there's no way around that.

The irony is that going for it on fourth down is considered a risky way of playing the game, when in fact spreading yourself out over four downs actually allows you to be much more conservative in your offensive play calling. Think about what that means. If a strategy is considered risky because it could cost you winning a game, and it can be shown that by never punting you would win more games, then that would have to be considered the conservative strategy while punting would be the risky strategy. Certainly that would be counterintuitive. Not that the numbers would necessarily average out in your favor in every game, which may be a difficult thing to get past. But they would over the course of a season.

Not only that, but if you never punt it takes the guesswork out of when to go for it and when not to. As it is now, coaches have to take so many things into consideration: the score, the time left in the game, the yards to go to get a first down, who has the momentum, etc. If you never punt, well, then you always go for it no matter what the situation is. And in the long term you're more successful by doing so.

Turnovers as better defined are the inverse of scoring. A punt is a turnover. Punts are also the most common turnover. The most common turnover is done voluntarily. That at least puts a different perspective on punting. And on top of that the arguments for punting aren't necessarily supported by the evidence. The fact that you turn the ball over when you punt appears to be more significant than the field position gained by punting.

Something as simple as a traditional definition can affect people's perception of the game and even how to be successful at it. I hope that even if my argument for never punting isn't convincing enough, that at least I have successfully challenged some conventional wisdom and maybe it can get some discussion going.

Responses To Feedback

I will attempt here as best as I can to respond to some feedback I have received before this article was published at FO:

1) It seems a bit too optimistic to suggest that if you never punted, you would score on one out of every five drives.

When I was constructing my argument, I first showed that even if you score on just one out of the five drives where your team would usually punt, that would be enough to make this work to your advantage. Then, when trying to figure out what it would really be, my calculations ended up going from 3.88 scores per game to 5.70, or an increase of 1.82. If one more score seems optimistic, that would certainly seem even more optimistic. A lot of this theory is counterintuitive, and I would like to think that this is counterintuitive as well. Whether this is realistic or not depends on how great of an effect four down mindset really has. And so far what I have found seems to indicate that the effect is greater than I had even first thought. But no doubt something that will probably be debated.

2) If a team really played this strategy, it would dramatically change the defensive strategy, which would dramatically change all the ratios which are involved in the theory.

I understand that defensive strategy would likely change, but I'm not sure what a defense could do to counteract your strategy in such a way that would dramatically diminish the benefit. In my mind, there would be more pressure on the defense, not only because of having to defend an extra down each series, but also because it should be more difficult to defend third down given that they would have to be equally ready for both a run or a pass and not just expect a pass. And if the opposition tried not punting as well, they would have to commit to it every time to get the averages to work in their favor, otherwise it could hurt them. If they committed to it every time, the playing field would be even again, and that should be the worst case scenario.

3) It would certainly be a good thing for teams to go to a never punting mindset as soon as they pass the 50-yard line.

I think this assumes that not punting is risky. I tried to make a case that not punting – if done every time – is actually conservative. If you go for it some times and not others, it could hurt you, because you need to go for it on every one of the five times teams typically punt in order to get the averages to work in your favor. Waiting until you cross the 50-yard line means you might only go for it one or at most two extra times. This would also mean playing in a three down mindset for much of the game and not getting the full effect of four down mindset. Perhaps that is why in other analyses such as David Romer's paper, the benefit of optimizing when you go for it on fourth down only leads to one more win every three seasons. Never punting and playing in a constant four down mindset is simpler and has a greater effect.

4) Considering touchdowns and field goals equivalent as non-turnovers is an oversimplification that is bound to skew the results.

There's no question the basis of this theory is simple, but I like to think that maybe it is that simple. Actual Turnover Ratio by definition is the inverse of net points. Net points does not consider touchdowns and field goals as equivalent, but ATR does. I would think that if counting touchdowns and field goals as equivalent skewed the results, there would be less correlation between ATR and wins compared to net points and wins. But there isn't. In fact, the first time I ever calculated ATR was for the 2003 season, and when I ranked the teams by ATR, the top 11 teams were 11 of the 12 playoff teams that year. The only way I can picture the results being skewed as a result of this strategy is if either you or your opponents started favoring more field goals in place of touchdowns or vice versa, and if (as mentioned in the next response) opponents may end up with more field goals, any skew there should be in your favor.

5) Taking a truly hard line as far as never punting would lead to a lot of field goals by the opposition, and this analysis doesn't really account for this. Virtually every four-and-out in your own territory would lead to at least a field goal, which in turn would lead to a kickoff, which would lead to a possession starting in your own territory.

Actually I thought I had accounted for this in my analysis, at least indirectly, in showing the increase in opponents points scored against you. And I think there is an overestimate here on how much opponents would actually score against you. First of all, you would have on average about five fourth down attempts where you now go for it instead of punt. Those fourth down attempts would be at varying points on the field, and maybe two to three (just a guess) would be in your territory. You'll make some of those, and so maybe a little under two would result in giving the opponent the ball in your territory. And considering that drives starting at midfield end up in a score about half the time (based on my calculations of 2.80 points per drive that start as a result of a fumble or interception which should theoretically average around midfield, and compared to the 5.54 point per score), let's say that on these drives opponents score anywhere from 50 percent up to even as much as 80 percent, that's in the neighborhood of one score against you. And if that one score is a field goal, that doesn't sound so significant.

These are the best answers I can think of for now. Certainly this theory will be debated, and that's good. So far, it's still just a theory. The more this is discussed and analyzed, maybe the closer we'll be to knowing if it really works like my calculations seem to indicate.

Guest columnist Jason Scheib lives in northwest Washington, where he has learned first hand how much trouble punting can cause a team. When not thinking about football strategy, he holds a job which would actually take longer than this article to explain.

Posted by: Guest on 10 Aug 2006

187 comments, Last at 04 Sep 2009, 3:34pm by Nelson

Comments

1
by dedkrikit (not verified) :: Thu, 08/10/2006 - 5:09pm

This article should be read in conjunction with RoboPunter.

2
by BB (not verified) :: Thu, 08/10/2006 - 5:16pm

Very interesting. Much like the "use your best reliever when you need him most and not just in the 9th" or the four-man rotation in baseball, this is a very interesting idea that will never get tried because if it messed up a game or two the coach would get fired, but theoretically very intriguing.

Someone try this in Madden and see how it works :)

3
by GlennW (not verified) :: Thu, 08/10/2006 - 5:17pm

A while back I read a similar probabilistic study where the conclusion was that a team should only punt on 4th-and-long (where there should be less such situations, given a more conservative offense) anywhere outside something like their own 10-yard-line. One major objection to that approach was that emotion is just too big a part of the game, and failures on these 4th-down attempts would be deflating and devastating. That didn't seem too logical to me, at least not if your team was mentally prepared to operate with this strategy. Maybe someday a radical coach will give it a shot, or at least a more conservative version. God knows that current strategic decisions are far from optimal.

I can't see a strategy of NEVER punting though, partially based on a such a nebulous psychological concept as the mandatory "four-down mindset". Teams currently play with a three-down mindset but based on their first and second down plays can still adapt to setting up for a short fourth-down try. Every down-and-distance at each point on the field has a probability of success and a cost of failure; I don't know why you wouldn't want to use that information. In the obvious extreme, if your offense is absolutely swamped on its first three plays deep inside your own territory, I don't know why you'd punish yourself in going for it just to maintain the strict mindset. The other team probably doesn't mind knowing that you're always going to go for it and defensing accordingly either.

4
by TheWedge (not verified) :: Thu, 08/10/2006 - 5:24pm

My first thought when I read this article was: "God, I hate it when someone plays like this in Madden" Why do I hate it though? Because it allows crappy players to stay in the game, which I guess is kind of the point. I just can't let any of my regular Madden buddies read this article, ever.

5
by karl (not verified) :: Thu, 08/10/2006 - 5:27pm

"I’m not a statistician, but in looking at these results there does appear to be a good correlation between Actual Turnover Ratio and wins."

Well yes. If you score more often than you punt and turn the ball over (ie: give the other team other opportunities to score, and stop the other team more often than you allow them to score, you will have more wins. The different types of scores are what keeps it from being a 1 correlation, I'd assume. Sorry for commenting before finishing the article.

6
by Kyle S (not verified) :: Thu, 08/10/2006 - 5:32pm

I disagree with the idea that interceptions and fumbles are often the result of fortuitous bounces of the ball. Successful defenses often force lots of turnovers. Interceptions are often the result of making the QB hurry his throw, and funbles are lot of times resultant from defenses that fly to the football - some do it better than others.

7
by karl (not verified) :: Thu, 08/10/2006 - 5:34pm

I take that back. I'm also not finishing the article. Enjoy.

8
by SJM (not verified) :: Thu, 08/10/2006 - 5:52pm

This will never ever happen in the NFL, but somebody should drop Mike Leach an email. He's looney enough (or smart enough?) to do it.

9
by Tarrant (not verified) :: Thu, 08/10/2006 - 5:59pm

I liked this article. Fascinating work.

The only thing I would state is that the article points out that on the average, teams have about five punts per game, thus there would be five more times during the game where a team would be going for it on 4th down.

However, that disregards the fact that if a team converts that 4th down and the drive itself continues, the team may face another 4th down later in that drive (which could not have happened had they punted). Is that taken into account?

Perhaps I missed this in the analysis, as it was a long article, if so, I apologize.

T.

10
by Mac (not verified) :: Thu, 08/10/2006 - 6:07pm

Very interesting article. Unfortunately, I doubt any team would adopt the strategy, even if every shred of logic in the article was flawless.

Madden really may be the best place to apply this strategy.

11
by Alex (not verified) :: Thu, 08/10/2006 - 6:10pm

tl;dr?

Followed by "Wall of Text" jokes.

I think you're taking a bit too simplistic view of the game. Stats are nice and all, but the idea that you must go for it on 4th and 8 on your 22 at the beginning of the game, and then your opponent scores an easy touchdown, would be pretty debilitating to your team. Assuming the average in every situation doesn't seem to realize that the average is rarely the actual.

And I apologize too, because I couldn't keep it all in my head at the same time. I'll try rereading this tonight.

12
by Kulko (not verified) :: Thu, 08/10/2006 - 6:11pm

I think your computation has one mayor flaw. You assume a relatively cheap value for turnover on downs. This is because people only go for 4th down in "safe situations". But if you go for it each time, you will give over at least one turnover on down in field goal range, which is much more expensive then turning over the ball on their 35 yard line.

13
by Dired (not verified) :: Thu, 08/10/2006 - 6:12pm

Madden or college. I mean, the NFL would never have created the run-and-shoot on its own for similar reasons, but a lower-tier (I forget what division Portland State is in) college program could certainly put it to the test somewhere.

14
by Brian (not verified) :: Thu, 08/10/2006 - 6:15pm

#4, I was thinking the exact same thing. Much like Denver's "we'll throw 10 guys on the LOS and you'll have no idea who's blitzing or who's dropping" defense, playing a 4-down team would be annoying as hell. Which is probably this theory might be crazy enoughto work

15
by masocc (not verified) :: Thu, 08/10/2006 - 6:17pm

I'll admit, I haven't read this yet...

Is this an Onion article? ;)

I can envision the underlying logic, but to be most effective, there really needs to be a happy medium somewhere.

Such as: Don't punt, unless it's 4th and > X and you're behind the X yard line. (Myself, in Madden, I use the opp's FG range and vary the 4th and X based on a number of factors (opp's defense, my defense vs. opp, kicker's accuracy, etc).

16
by B (not verified) :: Thu, 08/10/2006 - 6:17pm

The main problem with this strategy is it assumes that all turnovers are equal. Instead of adopting the traditional "always punt" approach or this "never punt" approach, I'd rather use a method that looks at down and distance and determines the optimium strategy. For example, 4th and 10 on your twenty yard line is a lot more risky than 4th and 3 on your 45 yard line. And when you also consider time remaining and the point differential, you get a something that's much to complicated to figure out in the time between plays.

17
by dannymac (not verified) :: Thu, 08/10/2006 - 6:20pm

Great article, you really can't find insight like this anywhere else but here. But, I have to take issue with one of your premises.

Points Per Score 5.54
Ball Received Via:
Fumble or Interception 2.80
Turnover on Downs 1.42

To account for the fact that most turnover on downs occur in opposing territory in short yardage situations you increase the "Turnover on Downs" value with from 1.42 to 2.80.

But I don't think that is nearly a big enough increase considering other factors. Most NFL games end on a turnover on downs from the losing team, and when the winning teams get the ball they make no effort to score at all. They either run out the clock or take a knee.

If you adjusted those situations out, the number would soar much higher.

18
by GlennW (not verified) :: Thu, 08/10/2006 - 6:27pm

Okay, the previous work I was referring to was the Romer study mentioned in this paper.

To elaborate further on the "four-down mindset", I don't see the drawback to always playing with a four-down mindset, but then re-evaluating the specific situation on fourth down. I don't see where that detracts from the strategy or the mindset. A statement like "if you punt sometimes but not others then you can’t really have a four down mindset even if you try to" is conjecture that goes beyond the simple probabilities of the matter.

19
by Sean (not verified) :: Thu, 08/10/2006 - 6:32pm

I tend to find that players who never punt in Madden almost always end up losing, as they are essentially handing 2-3 scores to you per game, which is a lot considering the limited number of possessions in a game with five minute quarters.

But that's just me.

20
by giving him the business (not verified) :: Thu, 08/10/2006 - 6:39pm

I employ this strategy when playing Madden. I also like to run the Punt Block defense on first down. I don't win very often.

21
by cjfarls (not verified) :: Thu, 08/10/2006 - 6:44pm

I'm kinda with GlennW... the idea is to give yourself the best probability of a successful scoring drive, and your opponent the worst probability of a successful scoring drive.

Blindly going for it on 4th & 15 from your 20 is much more likely to increase their probability of scoring more than it increases your probability of scoring... however, you're right in that it means that you should have played those first 3 downs as if you were going to go for it on 4th down: i.e. if immediately previously you had a 3rd & 15, you should have no problem going for a 7.5 yard play on 3rd down, with the assumption that a 4th & 7 is much more likely to be to your benefit than having to get 15 yards or punt.

If I had more time, or more math inclination, I think you could calculate that equilibrium point, based on where a 35 yard change field position for them decreases their probability of scoring equal to your probability of you making the 4th down and eventually scoring (Remember if you convert 4th & 8 from your 3, and get to the 11, your chance of an eventually successful scoring drive may still be very small).

Basically, I would think that this means the further back in your own end you are, the more likely it should be that you punt. At some point, the extra 35 yards probably doesn't change the opponents scoring chance as much (my assumption is, their probability of drive success starting from their 15 vs. their 40 is less a difference that starting from your 40 vs. your 15) as your probability of converting the 4th (no matter the distance), in which case going for it always is the right answer.

That net drive success probability is really what the article's model says should be maximized, and hence, the goal should always be to use all 4 downs most efficiently... and as pointed out, voluntarily giving the other team the ball doesn't do a lot in helping this net probability.

22
by giving him the business (not verified) :: Thu, 08/10/2006 - 6:45pm

Practical or not, I love the fact that this theory is being thrown out there. Anything that reduces the role of the kicker and/or punter is worth trying.

23
by Tracy (not verified) :: Thu, 08/10/2006 - 6:49pm

In regards to the assumption that turnovers on downs would average out to around midfield: I don't think this is accurate. Under this strategy, when the offense got to around the 35 yard line, they would be more likely to try a field goal then to go for it on fourth down. So turnovers on downs would only happen on 65% - 70% of the field - the portion of the field most advantageous to the other team, whereas fumbles and interceptions are just as likely to happen very near your own goal line. I would think, then, that a turnover on downs would yield more expected successful drives to the opposition then an interception or fumble.

24
by Crushinator (not verified) :: Thu, 08/10/2006 - 6:50pm

A fundamental problem with it is momentum. Teams have a tendency of becoming more aggressive the larger a deficit - This was the entire principle behind the Colts D this year - Hold the opponent early on, then just play pass the rest of the game.

The problem with this is that let's assume you go for it on 4th down every time. First series is after a kickoff, you have the ball at the 30. You're offense is faced with 4th and 12 and fails. The other team is then in position to get easy points. If they get a TD early, they can play Steeler Ball and just sit on the lead, knowing the opponent will have to lose their running game.

I do think not enough coaches go for it on 4th down enough, and I do agree with "the fourth down mentality" and its ability to work to some extent. But I don't think saying "go for it every time" is the answer. Another example -

It's 3rd and 9 from a teams own 35. They called a short pass play to pick up 6-7 yards and set them up for a 4th and 2. However, the QB is sacked and its now 4th and 17.

That was still a 4 down mindset there, but you realistically also can only do it in situations where you can make it. I think more coaches should approach the game where they could play third down as a set up for 4th down, but if that set up fails and you're at a very large distance, then punting would still be the decision. I think just deciding "never punt" is a little bit callous and removes a pretty important tool for weak offenses.

25
by Drew (not verified) :: Thu, 08/10/2006 - 6:54pm

Interesting article. I always hate it when an announcer uses the phrase "forcing a punt", even when it's 4th and inches.

My memory of "Friday Night Lights" is not so great. But from what I remember, the team that Permian plays against in the final never kicks in any fashion -- punting or placekicking. Of course, I saw that movie at Hollywood Bar and Filmworks, so someone might want to corroborate that.

Re 1,
I'd really like to see the Never Punt team play against RoboPunter's Always Punt team. Is there a way someone could sim that somehow?

26
by Josh (not verified) :: Thu, 08/10/2006 - 6:57pm

The extra roster spot that would be gained by having no punter would make this worthwhile. jk

27
by stan (not verified) :: Thu, 08/10/2006 - 7:08pm

When Kentucky was coached Hal Mumme he went for it on 4th down in his own territory a bunch. Mike Leach was his OC (btw #8).

His thought -- Ky had bad defense, if they fail on 4th, the other team is just as likely to drive it for a score after a punt anyway, try to keep the ball with that great pass attack. Made a lot of sense.

Not sure I buy THIS analysis, but I agree that coaches should go for it more than they do. The benefit cost of going for it very often does outway the benefit cost of the field position that comes from punting.

28
by Smeghead (not verified) :: Thu, 08/10/2006 - 7:11pm

Mike Holmgren should have read some of the anti-punting research going around before he let Tom Rouen suit up for Super Bowl XL.

29
by Smeghead (not verified) :: Thu, 08/10/2006 - 7:13pm

When pitting the 4-down mentality against the prospective loss of momentum, which strategy's swagger is bigger?

30
by Boston Dan (not verified) :: Thu, 08/10/2006 - 7:39pm

Actual Turnovers. This is brilliant, excellent work Jason. Are you going to track this data during the upcoming season. I wouldn't mind taking a crack at it.

31
by Reinhard (not verified) :: Thu, 08/10/2006 - 7:43pm

A safety should be considered a turnover, because, if I am on defense and safety you, then you MUST punt to me right afterwards. You had the ball and now I have it, sounds like a "Scheib turnover" to me.

It also seems to me that a factor that you are not considering is variance. On any given kind of play, how much variance is there in the number of yards? For a Colts-Edge run, he gets a certain number of yards every time, pretty much. On a Michael Vick "WCO" passing play, he might scramble and throw deep on a broken play, or get sacked. I think when you are talking about conservative vs aggressive playcalling, then you deal with this variance. Maybe I didn't articulate that so good... help me out someone!

32
by Karl Cuba (not verified) :: Thu, 08/10/2006 - 8:12pm

I've been interested for a while in a similar concept to this, the 'pick - punt'. What I mean by this is when, on 3rd down, a Qb throws a deep (40+ yards) pass with a low chance of success and a higher than normal chance of getting picked off. This seems to me to be a reasonable play as long as the WR knows it's coming and remembers to make the tackle, though it seems to run against the conventional wisdom of football. What's the difference between an interception 40 yards downfield and a punt? With the 'pick - punt' you at least get the possibility of making a big offensive play. I'm pretty sure the Rams used this philosophy in the Vermeil years (they also went for a lot of onside kicks and 4th down attempts). I'm not sure that I've explained this all that well. N.B. This obviously wouldn't work on 4th down, as all the DB would have to do is bat the ball down.

33
by VarlosZ (not verified) :: Thu, 08/10/2006 - 8:15pm

Never punting is a bit much to suggest -- it's almost certainly best to punt on 4th and 15 from your own 15 in the 1st Quarter -- but I think the idea is fundamentally strong.

And don't we know that it's a good idea intuitively? When our team has to hold off an opponent in 4-down mode, don't we expect them to march down the field, no matter how well our defense has been playing to that point? Don't we pray for for a turnover or a sack, because otherwise we won't be able to stop them? In marginal punting situations, are we ever not relieved to see the other team's punter come onto the field?

34
by Corey (not verified) :: Thu, 08/10/2006 - 8:23pm

Re: #8 and #27

If I'm not mistaken, Texas Tech (coached by Mike Leach) does not punt, except in the "emergency end-of-game" situations described in the article. Or maybe it's just that they've experimented with this strategy. Regardless, it appears to have worked for them. Mike Leach is truly my idol, and I hope he coaches in the NFL some day...

All football fans should read this article.

35
by Nathan (not verified) :: Thu, 08/10/2006 - 8:26pm

I think that after playing enough of this way in Madden/NFL2k, without actual indepth stats, I can tell you how it feels.

The swings are greater. The emotional swings on offense have to be taken into account. You feel like you're doing worse than if you punted the ball away and grinded it out. It warrents mentioning.

Second, I think it is important to not use the theory on 4th and 10. I used to use a 4th and 5 or better outlook. It still happened plenty of times, but it allowed me to control the likelyhood of it occuring.

I'd also never try it inside of my own 40. Your milage may very.

36
by Larry (not verified) :: Thu, 08/10/2006 - 8:31pm

First things first - I like it. Great out of the box analysis. Love this stuff.

My small critiques:

One thing missing, aside from a binomial statistics analysis of the true increase in scoring rate from a 4 down strategy, is the reduced number of possessions per game. If you go for it all the time, then you'll make some of them increasing the average length (in both time and yardage) of a drive, which means fewer total drives for both you and your opponent. This ought to reduce the advantage somewhat, without thinking too hard about it.

Also, you may have overestimated the positive effect. If you need one extra score to make this work - with a 20% score conversion rate, the chances of getting AT LEAST one extra score are 1-0.8^5 = 1/3, more or less. That's because sometimes you get 2 or more, so that's balanced by more games with none. The perils of working with averages. So, I think it'd play out a little differently.

Still, very cool. Nice job.

37
by Nathan (not verified) :: Thu, 08/10/2006 - 8:35pm

That isn't entirely true, I think I used a sliding scale to determine going for it.

So at my own 40, 1 yard, 50 2 yards, so on. It wasn't exact, and mostly by feel.

And my mindset was different depending on circumstances.

Generally I found you could tell if you were going to make it on 4th depending on the previous 3 downs. I know that doesn't make much sense, and was probably just mental, but I was more than often right guessing if I would make it.

I'd usually run extra safe routes on 4th. If it was common, i bet you could tell. I would almost never run a deep pass, I think that warrents inclusion.

Curls, Runs, Outs, Slants.

Nothing fancy or complicated. No Zags, or plays that took a long time to develop.

38
by GlennW (not verified) :: Thu, 08/10/2006 - 8:48pm

According to the Big-12 football website, Texas Tech punted 50 times last year. They'll try all kinds of crazy things, mix it up, but they've not gone so radical as to abandon punting, or even eschew it in most of the "obvious" situations.

39
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 9:07pm

Wow, a bunch of comments already. I'll try to answer some of them the best I can...

Re: #3 "Every down-and-distance at each point on the field has a probability of success and a cost of failure; I don’t know why you wouldn’t want to use that information."

Because if this theory works when you keep it this simple, why make it more complicated? In fact I think you get the maximum benefit by keeping it this simple. I don't think you can get enough benefit out of this if you punt some times and not others for reasons I mentioned in the article. And I think the reason people want to take these probabilities into account is because the perception is that punting is risky. For instance Roemer's paper was about optimizing risk. Not punting at all is about taking away the risk, and once you start looking at every 4th down as a separate event that concept gets missed.

40
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 9:14pm

Re: #6 - all I meant there is that the correlation between the conventional turnover ratio (fumbles and interceptions only) and wins isn't the best, which of course isn't anything new. I just mentioned that as one reason why, although not the main reason. The main reason being that the conventional turnover ratio is incomplete and doesn't include all kinds of turnovers.

41
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 9:23pm

Re: #9 - I don't think that really matters because what we're looking at here is not so much the number of 4th down situations as it is the number of drives that used to end with a punt and now saying will instead end as either a score or a different kind of turnover (other than a punt). It's the number of drives and final result of those drives that matter.

42
by Paul (not verified) :: Thu, 08/10/2006 - 9:29pm

When my wife and I play Madden, it's usually Steelers vs Colts. I play the Steelers and she uses the Colts. I frustrate her to no end by just running every down or short screens--basic ball control. With this style of play, every 4th down that I get is 3 yards or less, since I typically average 4-5 yards per carry. Punting be damned! Go for the first! And my success rate on fourth down is over 75%. I have had 7 minutes drives(quarter and a half). She goes vertical, every down. drives last no more than a minute--TD or punt. Since my style is usually more successful, as soon as she adopts the forget the punt style, the game quickly becomes a rout as the defense doesn't have to guess.
That's the key--if you have fairly high success rate (in FO terms) a punt-free style can definitely work, even in the NFL. It works best in the small-ball style offense , frequent short yardage situations where the pass works as well as the run and defense is kept guessing. It also could help when the defense is either overwhelmed (playing Indy) or highly superior (Bears playing SF). In the overwhelmed case, the idea is to match scores with a team that can score at will. In the highly superior case, the idea is a low scoring game where one score could determine the game. This one is more dependent on field position, so using it deep in your own territory is riskier. But from your own 40 might be worth it.

43
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 9:30pm

Re: #12 - I didn't go with the lower value, I went with the 2.80 value which is the same value as for fumbles and interceptions. The idea being that those events also happen at any point on the field and so should average around midfield.

As for when you are in field goal range, I talked about that in the "Exceptions" section where I said that you should kick the field goal when facing 4th down in field goal range because the whole goal here is to have your drive end in a score and a field goal accomplishes that.

44
by moloch g (not verified) :: Thu, 08/10/2006 - 9:56pm

id be worried about using the last two weeks of a season to analyze because lots of team change their strategies by then because of standings and the desire not to get guys injured. so maybe teams are running when in week 5 they would be passing or because their starting tailback is sitting the week out to rest for the playoffs or is faking an injury to ensure that he doesnt get injured in a nonimportant game to ensure he is still healthy to reach next seasons bonuses. perhaps u should adjust your stats to look at what teams did in weeks one and two when they had all their freshest ideas instead of the last two weeks

45
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 10:14pm

Re: #17 - I don't think as many games end with a turnover on downs as you think. I went through the play by play of every game and don't recall that being the case most of the time.

I'm not sure I understand why taking those out of the equation would cause the point value to go up. Even those end-of-game turnovers on downs would take place at any point on the field and any down and distance which is what the 2.80 captures.

46
by Daniel Edwards (not verified) :: Thu, 08/10/2006 - 10:20pm

A possible variation on this (for instance in "obvious" punting situations like 4th and 15) is quick-kicking. At Florida, every so often Spurrier would get irritated with his offense and call a quick-kick on third down; it would be similar here but on fourth down. You'd be even less predictable.

If having to defend four downs all the time bothered the defense, I would imagine having to play a safety way deep every time the offense lined up in the shotgun on fourth down would irritate the hell out of them. (So of course I imagine Texas Tech doing exactly this.)

47
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 10:38pm

Re: #23 - yeah, I thought of that later on, but figuring out how much to increase the point value is a different story. It may change noticeably, it may not change much at all. And in fact I didn't measure where drives starting as a result of a fumble or interception began, midfield seems logical and also fits that the 2.80 is around half of 5.54. But even a noticeable change would probably still cause the overall benefit of this strategy to be in your favor, just not to the same degree. That's something worth looking into though.

48
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 10:48pm

Re: #30 - I certainly could but not sure if enough people want to see it tracked. I'm glad you brought this up because what may be missed in all of this is that whether teams punt or not, Actual Turnover Ratio is a valid and meaningful statistic and is interesting in its own right. At least I think so.

49
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 10:59pm

Re: #31 - Funny you should mention that because I once had a long back and forth discussion with someone about that. Only it started out by me claiming a safety was one turnover and him claiming it was two turnovers, one for the other team getting points and the other for you punting the ball back to them. Of course the other team getting points is not a turnover for you, that's not the definition of a turnover. But in all of that discussion I became convinced that a safety is actually not a turnover at all. Just because the method of giving the ball to the other team after a safety is usually a punt (I say usually because according to NFL rules teams have a choice of how to kick) doesn't in and of itself make it a turnover, it's only a turnover if it was the end of your drive. But the end of your drive was the safety itself, which was a score. The kick comes after the score like in any other scenario. There are just different rules as to how safeties are scored and kicked. It could just as easily be a negative 2 points for you (which would make sense as you are 'scoring' backwards into the wrong endzone) but instead they add it to your opponent's score.

50
by Countertorque (not verified) :: Thu, 08/10/2006 - 11:04pm

RE: 49

But doesn't it improve your correlation between ATR and wins if you count safeties as turnovers? Granted, I don't think it changes much, cause there aren't many safeties.

51
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 11:09pm

Re: #36 - I did notice this but didn't mention it because it seemed insignificant. The best offense in the league will have around 2 less drives per game than the worst offense in the league, and a team that never punts would go part way up that scale for a difference of about 1 less drive per game. Which of course is then 1 less drive per game for your opponents (and 1 less drive that could potentially have ended in an opponent scoring). Along with that, all the numbers may shrink a little bit but all to the same percentage and so in the end should give you pretty much the same results.

52
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 11:25pm

Re: #42 - thanks for mentioning that even on Madden it's not just about not punting, it's also about playing with a four down mindset and running on 3rd and 7, etc. When I was first putting these thoughts together I had my brother in law's old nintendo and played some Madden 99 (outdated by today's standards I know). Although I couldn't prove it, it definitely felt like I beat the computer more when I didn't punt and played in a four down mindset. And that gut feeling is one reason I kept trying to figure out how to get to the end of this theory when I would get stuck.

53
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 11:31pm

Re: #44 - I wanted to do the whole season but just didn't have time. For some reason I started at week 17 and was going to work backwards to week 1. The numbers were normalizing by the time I stopped anyway. I thought about doing the first two weeks instead but with week 2 being a bye week I wouldn't have been able to get every team's tendencies represented in a balanced way.

54
by Jason Scheib (not verified) :: Thu, 08/10/2006 - 11:37pm

Re: #50 - I'm sure it would, but that's beside the point. If something is not a turnover it can't be counted as one. And just because it would improve ATR doesn't mean it's a turnover. Same logic as a TD (which is also a drive ending in a score) being counted as a turnover against the defense. It would also improve ATR but wouldn't be valid because it's not a turnover.

55
by Scott de B. (not verified) :: Thu, 08/10/2006 - 11:46pm

First of all, you would have on average about five fourth down attempts where you now go for it instead of punt. Those fourth down attempts would be at varying points on the field, and maybe two to three (just a guess) would be in your territory.

I have a problem with this assumption. Punts don't happen at random locations on the field, but almost always between your goal line and the opponent's 30 (because otherwise you go for the field goal). So by area alone you'd expect 5/7 of punts to occur in your own territory. This would however be balanced out to some degree by the fact that a drive is much, much more likely to start on your 20 than your 1 yard line.

What this article really needs is am analysis of where punts are taken from.

56
by Jon Fuge everybody (not verified) :: Thu, 08/10/2006 - 11:51pm

I like the idea but I wonder if the primary calculation of success rate is correct. You say you'd have to score approximately 1 out of every 5 times you go for it on fourth down. The problem I picked up on was that after you turn it over without a punt, not only are you increasing the other team's chance of scoring on that drive, you're decreasing your own chance of scoring on the next drive if you stop them. Adding that in might change it to 1 out of 4.

57
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 12:11am

Re: #55 - an analysis of where punts take place may be helpful, but all I was trying to do in that response was try to show that this strategy isn't going to end up in opponents just kicking a bunch of extra field goals against you. The actual evidence itself is not the assumption I went with as to where punts took place, but in the actual numbers of turnovers (and points) for and against you by going with this strategy. The response I gave there was more an illustration to try and explain why the numbers make more sense then someone else had thought. That's all.

58
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 12:51am

Re: #56 - 1 out of 5 is just the minimum (technically the minimum would be slightly less than that but that wouldn't sound so practical to say you could still be ok with scoring less than 1 extra time per game even though it's true). The actual calculated amount at the end came to 1.82 out of those 5 drives ending in a score, so if it did drop some from the effect you mentioned, it still has plenty of room to drop. Say it even drops to 1 in 5, you're still ok. But I'm not sure how to measure this. Part of the problem in figuring any of this is that the stats when you punt and the stats when you never punt are like comparing apples to oranges.

59
by jonnyblazin (not verified) :: Fri, 08/11/2006 - 12:51am

I'll bet some offensive coordinators would love it, others not so much. When Paul Hackett was coaching the Jets when they were good, their offense was ridiculously efficient at gaining 3-5 yards on every play (primarily C-Mart and Pennington waggle). With a sack-prone offensive scheme (or QB who holds the ball too long), this strategy would be a disaster.

60
by Kibbles (not verified) :: Fri, 08/11/2006 - 1:57am

Re #14: #4, I was thinking the exact same thing. Much like Denver’s “we’ll throw 10 guys on the LOS and you’ll have no idea who’s blitzing or who’s dropping� defense, playing a 4-down team would be annoying as hell. Which is probably this theory might be crazy enoughto work
I actually gained a healthy respect for Larry Coyer last year. I think he's the only defensive coordinator with enough balls to call a 10-man blitz in an actual live game situation- let alone calling the exact same 10-man blitz AGAIN on the very next play!

Re #32: This obviously wouldn’t work on 4th down, as all the DB would have to do is bat the ball down.

You'd be surprised. There was a play in the Miami/Denver game last year where Plummer threw it waaaay downfield on 4th down and whichever defensive back it was covering the play (I think it was the nickle guy) picked it off for his first career interception. I always thought it was ludicrous that that play was even counted as an interception in Plummer's totals, since it was a far more positive play for Denver than a simple incompletion would have been, giving them an extra 30-40 yards of field position.

Anyway, my thoughts on the "never punting" offense. Anecdotally, I don't think it would improve your chances of winning any. I mean, in the 2003 playoffs, the Chiefs and Colts both tried adopting the "never punt" offense, but only half of those teams won- and a 50% winning percentage is about what you'd expect the average "sometimes punt" team to post, anyway.

(By the way, in case it's not immediately obvious, I'm being droll).

61
by Derek (not verified) :: Fri, 08/11/2006 - 2:49am

I really enjoyed the article. I'd like to see someone with a stronger mathematical background take another look at it, however, with more sample data on which to base the conclusions. Not that I can remember much from my Stat class, but I think it would make the case stronger.

And to me it is somewhat intuitive that a 4-down mindset is better. In the CFL, if you're kicking on almost every 3rd down, then there's an expectation of yourself to get 5 yards/play, which seems really difficult. 3.33 yards/play with a 3-down mindset seems reasonable, but with the 4th down in the NFL, expecting a paltry 2.5 yards/play out of yourself sounds far, far easier. It makes sense that scoring could actually increase.

On the other hand, you don't want to go for 4th and 10 on your own 10. I'd imagine that the probability of scoring on that drive is miniscule to that of the other team scoring, so your expected points of going for it on 4th and 10 would be less than the expected points from punting.

My educated guess would be that punting is a lot like sacrifice bunting. In many situations, why give up the out/possession? But if you're merely looking to prevent the other team from scoring a touchdown, maybe the expected value of an attempt at pinning the opponent deep within its own territory is greater than that of trying to score one more FG to ice the game.

But then there's the question of attrition. How does pinning the opponent deeper in their territory affect the expected starting field position on your next drive and thus the probability of scoring on your next drive?

This certainly merits a more in-depth look.

62
by Sergio (not verified) :: Fri, 08/11/2006 - 3:24am

Whew. Took me a while to finish this puppy...

Great article. It's an interesting perspective, and I'd love to analyze the numbers for a "limited" edition of this train of thought (such as 'only punt when the success rate looks favorable'), though I do understand that's one step next to impossible.

Still, I'm willing to try it whenever I get the chance. I'm pretty sure it would, indeed, lead to more wins *over the course of an entire season*, though I don't know about a 3-5 game improvement for a 8-8 team...

63
by brent2mnen (not verified) :: Fri, 08/11/2006 - 4:34am

There are just too many variables that this article fails to take into account. Regardless of whether you have a “four-down mindset,� giving the ball up on your own 25 will result in more points per possession for your opponent than giving them the ball on their 40. Add to that the fact that you are still relatively unlikely to score, even if you were to convert your 4th down at that point, and there is your recipe for disaster. I would expect that drives that start at their opponents 25 yard line might result in nearly 5 points per possession, while drives that start at their own 40 are probably closer to 2.5. Even if you were to make a first down, the average number of points on a drive that started from your own 30 yard line would probably be slightly lower – say 2.3. If I had real numbers, and some sort of knowledge of statistics, I could probably tell you what the net result of such an action was, on average.

The fact of the matter is, the longer the drive, the less likely the score, and the fewer points per drive. This is essentially the point that Easterbrook makes with his “maroon-zone�. That the benefits of punting inside the 40 yard line, and increasing the length of the field somewhat, are less than the increased likelihood of scoring if you were to “go for it� on 4th down. This is magnified by the fact that the average net punt from inside your opponents 40 is less than the average net punt from inside your own 40 (I suspect that touchbacks account for much of this). A team could always switch to a “four-down mindset� at this point and gain the same benefits as Scheib suggests, with much less risk.

A truly interesting study would be to know what the average number of points per drive was, based on the starting point of the drive. With enough data, you could compare starting field position after kicks, punts, and turnovers – which could indicate the true statistical significance of “momentum-changing plays�, such as turnovers (or long returns). That, however, is a topic for another article, isn’t it?

64
by scarface (not verified) :: Fri, 08/11/2006 - 4:35am

Re. 35,42,52 (Madden):
I and all my friends play Madden 2k5 with kind of a 4-down mindset (but we don't have anything at stake, though :) ). On 3rd and 8, I often call a Run out of a 3-WR formation, get to 4th and 2, and of course, go for it. If I fail and am stuck at 4th and long, then I punt. Note that this works unreasonably well against the computer's AI, because the it will dumbly and predictably bring out a pass-defending Nickel defense. So Jason, your results against the computer don't count :) But yeah, I totally agree that 4-down thinking makes the offense more unpredictable, and that's always a good thing.

65
by scarface (not verified) :: Fri, 08/11/2006 - 4:36am

Re.39 (from Jason):
Jason, you're basically saying that coaches are not smart/willing enough to completely use all the info and probabilities at their disposal to calculate the optimal strategy at every point in the game - so you want to give them a simplified thumb rule like "Never Punt". Fair enough.

But ideally, I think you can have all the probabilities/statistics for EVERY possible yardage-and-down fed into a simple computer program, and have a lowly assistant who can look it up in a matter of seconds, during the game, advising the coach. And I suspect, when you're 4th and 10 on your own 30, the assistant would strongly advise the head coach to punt. Note that I still think you should use the 4 down mindset, when planning your plays - just adding that you shouldn't throw away information in an age when you have vast processing power at your disposal :)

66
by EnglishBob (not verified) :: Fri, 08/11/2006 - 5:08am

Easterbrook will love you. I agree with the concept as regards possession from about the halfway line but still think in practice it would prove suicidal if used in own half. How many games have we seen become field possession battles where a TO (I don't mean Owens) decides the result?

Also I think the defence can and would change radically to alter outcomes. If I know you are using all four downs I am more likely to run blitz to try and create tackles for losses. I can generally afford to gamble more as I know your 4 set mentality makes attempts at longer plays less likely. Going for fewer longer plays will also reduce the average no. of points you score (I am assuming), I can play lots of 8 in the box. I think there was an article in FO 2005 which demonstrated the importance of big plays.

You would however get an off-setting gain from 4 down mentality in that your time of possession should improve (even if opponents' field position improves). Which is more important? I can see how a four down approach thus helps in Madden (but its easy to run in this, its not real football!) with its 5 min quarter.

It also looks like the average points following TOD implies this does normally happen with a team chasing a game and the opposition thus are more interested in eating up the clock than scoring.

Finally (phew long post, sorry), I have seen a form of 4 down approach used in a game called statis pro football. This might be more useful as a guide as it is statistically driven and over 15 min quarters. Still just a simulation but arguably more use than Madden.
Although it is frustrating for the defence to see the offence converting on 4th, eventually you'll stop them and have better field position than would probably have been the case. You can adjust defensively to it as well.

There is some limited read across from rugby, where teams can adopt this approach. As a teenager my team was not allowed to kick by the coach (so we could develop handling skills etc). It means if you are a better team than the opposition you tend to really rack up points and can build big early leads. Against a good team however, you risk giving away good field position that they can convert into points. When weak teams or teams trailing in a game adopt this approach they typically end up with a bigger margin loss than would have been the case (fine in that situation as losing by 10 or 20 doesn't matter, it's still a loss, and maybe you raised your chances of winning by 5%).
I kind of feel that would be the actual impact of running no punts in the NFL. You'd beat the 49ers by 30 pts instead of 10 pts but the Ravens/ Bears would be grinning from ear to ear as their defense manages to stop you often enough to give their weak offences great field position- leading to more points and you losing.

67
by Jake S. (not verified) :: Fri, 08/11/2006 - 8:49am

1. Assuming it's better to never punt, it's true that punting sometimes is bad strategy, but again, assuming it's true, it's not poorer strategy than punting most times, as the author says it is.

2. It is just me, or does he totally forget about field position? It seems like his whole argument is the turnover ratio. Where does he take into account the benefit of field position?

68
by senser81 (not verified) :: Fri, 08/11/2006 - 9:00am

I guess if I were a defensive coordinator and I knew that my opponent never punted, I would try that much harder on 2nd & 3rd down to get a sack or loss of yardage play, because a 2nd-and-17 or a 3rd-and-long situation has the potential to be even more beneficial than a turnover, considering that a DB might get tackled 30 yards downfield after an INT. I wouldn't be all that concerned with tricky coverages or the running game...I'd just wait for a sack or an offensive penalty (preferably a holding penalty) to throw the 4-down offense out of whack and get good field position immediately.

69
by Jake S. (not verified) :: Fri, 08/11/2006 - 9:03am

Oh, I see where he includes field position. In the ATR. Interesting.

70
by Michael LaRocca (not verified) :: Fri, 08/11/2006 - 9:07am

I can't believe I never questioned the wisdom of punting. I mean, a long long way from the NFL, I QBed a backyard team that always went for it on fourth down. So did every opponent we ever faced. It was very rare for any of us to fail. This is high school kids playing outside the school system, sandlot football, and miraculously not killing each other. So in the NFL, yes, let's ask the question. Here's hoping Mike Martz hears about this. He's an innovator, and I'd love to see what happens if a team quits punting. Alternately, Al Davis is the guy who investigated using sumo wrestlers for players (no endurance), so maybe he can tell his coach "no more punting." After over 6 years in China, I live in Thailand and will finally watch the NFL again! Or maybe on the college level, the University of Hawaii could add this to their run and shoot. Mouse Davis or June Jones, I've forgotten which is coaching, is another innovator.

71
by Michael LaRocca (not verified) :: Fri, 08/11/2006 - 9:08am

Did I say Al Davis's people should stop punting? I just remembered they have Aaron Brooks out there now. Holy crud, punting on first down might be the way to go.

72
by senser81 (not verified) :: Fri, 08/11/2006 - 9:13am

Al Davis also was the guy who spent a first round pick on a punter (Ray Guy).

73
by Shannon (not verified) :: Fri, 08/11/2006 - 9:53am

RE: 49 - yes, a safety is a score followed by a kick.

But the kick is *to* the team that scores - opposite of a kickoff.

That's why a safety is definitely a turnover.

74
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 10:22am

Re: #73 - but that would mean a drive ending in BOTH a score and a turnover, and by definition a drive always ends by one or the other. you would have to establish that there are multiple changes of possession (and multiple drives) on one play for there to be a turnover on a safety.

75
by SG (not verified) :: Fri, 08/11/2006 - 10:24am

Here's one reason you haven't considered that might make this a bad idea: fear. Put a QB on his own 20-yard line with 4th and 6, early in the 4th quarter of a close game, and he knows that not converting RIGHT HERE very well might give the other team the lead. That's a lot of extra pressure. Some QBs will rise to the challenge, but a lot won't.

76
by Pat (not verified) :: Fri, 08/11/2006 - 10:50am

Well, I won't comment on the football stuff, but this statement, in response to question 3: If you go for it some times and not others, it could hurt you, because you need to go for it on every one of the five times teams typically punt in order to get the averages to work in your favor. is wrong - it's the Gambler's fallacy in a slightly different light.

If you score once every five times not punting, that doesn't mean you need to go for it five times to score. In fact, if you score 20% of the time, going for it five times, you're not guaranteed a score. 30% of the time, you won't score at all (0.8^5). In truth, those five times aren't equal: if you're going for it on fourth down deep in your own territory, you're less than likely to score than if you did it deep in your opponent's territory, because you've still got a long way to go even if you make it. Not going for it deep in your own territory doesn't make your chances of scoring on a more favorable fourth-down attempt better.

In fact, probably the best thing for a team to do against a team that doesn't punt is punt. The most dangerous portion of the "never punting" strategy is deep in your own territory, and the easiest way to pin someone there is to punt. Against a team that never punts, it probably makes more sense to punt when you're near the 50 than it does to kick a long field goal, because the chance of them driving down the entire field is small, and the resulting field position when they do fail is likely going to be better than your current field position.

Think about it this way: if you're facing a team that doesn't punt, and you're at your own 50, and can pin the team down at the 10 (pretty reasonable), that team's drive needs to net forty yards before it becomes a bad decision for you to have punted. While average drive lengths will obviously increase for teams that don't punt, the average drive in the NFL netted 28 yards - an increase in 12 yards/drive seems a little much to me.

And, in addition, against teams like San Francisco, that becomes an obvious choice - the 49ers only had 18 yards/drive. Expecting that to go to 40 yards when they stop punting is crazy. Which means that the team that punts against them from the 50 will gain yardage by punting.

77
by senser81 (not verified) :: Fri, 08/11/2006 - 11:15am

re: #76

I agree that the best thing for the opponent of a 'no-punt' team is to punt. I was thinking of something similar in regards to kickoffs. If you were playing a 'no-punt' team, and you won the coin toss, wouldn't you always elect to kickoff to start the game? Its possible that you would elect to kickoff for both halves. If you went to OT, its possible that would pull a Mohrningweg and elect to kickoff. You would rarely need an onside kick late in the game, because you always have the opportunity to take possession inside the 'no-punt' team's 20.

Also, on a side note, the CFL having so many missed FGs has nothing to do with having only 3 downs. It has more to do with the rogue point.

78
by Eric (not verified) :: Fri, 08/11/2006 - 11:21am

I'd love to see a coach in the MAC or Sun Belt try this...

79
by Sergio (not verified) :: Fri, 08/11/2006 - 11:25am

Re:66

I don't think the 4 down mentality would significantly change the playcalling on 1st-2nd down, which are the downs when you mainly call "long plays"... in fact, it adds an extra down on which to call any kind of plays, long, short, whatever.

Besides, any team can burn a big blitz. Hot reads, maybe max protect, hell, a simple go route to the most aggressive corner's man would suffice.

My main beef with this philosophy would be that, indeed, giving the other team the ball with a pretty good chance to score (7 or 3, whatever) would demoralize your own guys. As a coach, I wouldn't mind the points so much as the "we don't stand a chance in this one" attitude...

80
by DGL (not verified) :: Fri, 08/11/2006 - 11:46am

I see this article like some of the "radical Moneyball" strategies that have come out of the sabrmetric camp in baseball (e.g., never sacrifice bunt). It's an extreme position that leads to some fascinating insights; it will likely never be fully adopted in the game; but it could lead to an evolution of strategy.

I think the next step -- tossed out for someone with more time to compile the statistics to run with -- is to come up with an "expected points" table like BBPro's Expected Runs matrix. Probably a simple table with expected points by starting field position. If you know that, you can work out optimal strategies (punt or don't punt) based on expected points for you and your opponent given where you'd have the ball after converting a first down or where they'd have the ball after TOD or punting.

Anyone have the drive data (starting field position and result of drive) in a spreadsheet?

81
by jebmak (not verified) :: Fri, 08/11/2006 - 11:53am

I think that if the definition of a turnover excludes safeties, then the definition should be examined.
I don't really understand the arguement for it not being a turnover. Your offense just lost its chance to score, and now the opposing offense has a chance to score. How is that not a turnover? Is it somehow different than the other turnovers in a way that isn't worse for your team (by giving the other team points as well as the ball)?

82
by cjfarls (not verified) :: Fri, 08/11/2006 - 12:13pm

Derek in Post #61 may have pointed out a fatal flaw... the problem is that drives are not independent... field position of your next drive is not independent of your field position on the current drive, so therefore having a short drive hurts you much more than simply the "average scoring probability" of that field position. Not only do you increase your opponents score % on the next drive, but even if you stop them, you've decreased your scoring percentage on your following drive. This is a reinforcing feedback loop that continues until a team scores or a turnover happens, and the "no punt" team is on the losing end of the loop.

Pat in #76 shows how this works, by using the "avg. drive length" and showing how punting against the "no punt" team may be the best strategy.

Now, as many have pointed out, this effect may(?) be less as you get into opposing territory or Easterbrook's "Maroon Zone".

83
by Pat (not verified) :: Fri, 08/11/2006 - 12:15pm

As a followup to #76: Consider the team on their own 50, on 4th down. They've got a 40% chance of getting better field position on their own drive - or, they could punt. If they punt, their opponent needs to drive 50 yards to do so. The average drive last year was 28 yards - the average drive that ended in a punt last year was likely much shorter, so let's just say 20 yards for fun.

Just assume the team automatically drives 20 yards before needing to convert a 4th down Crappy assumption, but it's not even needed. So the ball's now on the 30 yard line, and the opposing team now has a 60% chance of ending up in better field position. And even if they do convert, the team then drives to the 50 yard line. And now they've got a 60% chance of ending up in better field position.

In total, going for it on 4th down from the 50 gives the punting team a 40% chance of having better field position when they get the ball back. Punting, however, gives them an 84% chance of having better field position when they get the ball back. Basically, so long as you can hold the opponent to less yardage than your punt with better frequency than your 4th down conversion percentage, you're better off punting. This will always be true in cases of roughly equivalent strength, balanced teams.

It might not be true in cases of imbalanced teams - say, Cincinnati versus Indianapolis. I could definitely see a "no punt" strategy in that game.

84
by cjfarls (not verified) :: Fri, 08/11/2006 - 12:29pm

As I think about this more, I think I see where/why many coaches screw up by punting in the "maroon zone".

They seem assume that 35-40 yards of field position is of equal value no matter where it occurs on the field, because the avg net punt (~35 yards) is greater than the average drive (28 yards)... Consequently, punting would make sense, as on avg it is a 7-yard gain for your team.

However, as you get in the Maroon Zone, I would guess the avg. net punt decreases because of touchbacks, etc. If the average play gives you 3 yards, then any punt from an area where the net avg. is less than 31 yards is a bad deal... meaning a "no punt" strategy (inclduing 4-down mindset) from this area is better.

Now, of course coaches tend to have confidence in their guy's ability to pin the other team, so they'll be more likely to mentally accept the fallacy that their punter won't screw up and will get them their net 31 yards.

85
by CA (not verified) :: Fri, 08/11/2006 - 12:50pm

Really interesting article and discussion. There was a lot to read, and in response I have a lot to say, so I apologize in advance for the length of this post.

I do agree with some of the criticisms of the extremes of the proposed strategy. Jason, you are assuming that the expected point value to the opponent from a turnover on downs is always 2.8. Obviously, as others have pointed out, the actual expected point value is highly dependent on field position. A turnover on downs at your own 10 yard line is much worse than a turnover on downs at the 50.

You may be right that never punting is better than punting as often as teams actually do, but the optimal amount of punting is probably somewhere in between. For instance, perhaps you should almost always go for it on 4th and short, but usually punt on 4th and long deep in your own territory. As you get closer to midfield, perhaps you should almost never punt almost regardless of the distance. Your "4 down mindset" concept would have to be modified slightly deep inside your own territory. Of course, I think this is simply what a lot of us who frequent this site already believe; that is, basically, that teams should go for it on 4th down more often.

I do see the similarities to the proto-sabermetric revolution in baseball analysis. You are saying that the goal of football when you are on offense is not to make Actual Turnovers (which of course include punts), much as Bill James and others said that the goal of baseball when you are on offense is not to make outs. Thus, ATR becomes in some ways the equivalent of OBP. Of course, in baseball, you can lose only by exhausting your available number of outs, which are pre-set (barring, say, cancellation due to bad weather, forfits, disqualifications, and probably some other things that aren't coming to mind). In football, the number of potential actual turnovers is not officially limited, and you can lose by running out of time.

I believe that "momentum" is about as meaningful in sports as "clutch" (which is to say not at all), so I certainly disagree with the comments asserting that teams that turn the ball over on downs early will be so dispirited that they'll just give up. But perhaps that is an argument against the strategy at the Pee Wee level.

Unless I'm missing something, this strategy should lead to significantly increased time of possession for the team that employs it. If the conventional wisdom that defenses wear out more quickly than offenses is right, then the no punting strategy may be even more effective than one otherwise might think.

I second the sentiment that I would like to see a more thorough mathematical treatment of this topic by someone with a more robust understanding of statistics, econometrics, etc. If it reached similar conclusions, that would make the case that much more powerful.

Most importantly, Jason, I want to commend you on your effort and frankly your intellectual courage in writing and publishing this article. Whether or not your conclusions or even logic are right, this is a very good contribution to the football analysis landscape. Please don't let any criticism that you receive sway you from doing this kind of work.

86
by EnglishBob (not verified) :: Fri, 08/11/2006 - 1:08pm

74-81
How is a safety any different to an opponent scoring and then successfully executing an on-side kick? If you classify an onside kick as a turnover (correctly in my view) than surely the safety is no different and must also be seen as a turnover? I know its only semantic and won't affect the arguement but....

Also on the average of 2.8 pts per Turn over on downs. Presumably we can calculate this figure specifically for each team and also calculate how often they could be expected to score by making 4th down attempts. This would (I feel) result in negative net points for many teams and not always be positive.

87
by Tony (not verified) :: Fri, 08/11/2006 - 1:36pm

I have employed the never punting strategy (or as I like to call it, "punting is for women and bad teams") in Madden with great success in the past. One of the greatest benefits of it in Madden is that it wears down the opposing defense. All you need is one or two long drives and you can poke holes in even the toughest defenses.

88
by Abarine (not verified) :: Fri, 08/11/2006 - 1:36pm

#34-GREAT article, great read. I wish more football writers wrote like that.

89
by Grouchy Bills Fan (not verified) :: Fri, 08/11/2006 - 1:39pm

Based on this article, I'd say the best thing to try is exactly the "punt less often" approach that the author wants to discredit.

The people saying "this would be insane on 4th and 12 from your own 30" are missing the point - you'd almost never be in that situation. On your own half of the field, you'd be conservative and use high-percentage plays on 2nd and 3rd down. In the rare cases where that doesn't work and you're in 4th-and-long anyways, go ahead and punt I say.

When Jason says "you need to go for it on every one of the five times teams typically punt in order to get the averages to work in your favor" he is making no sense. If you have a 20% chance of success on something, that doesn't mean it's pointless to try it fewer than 5 times.

90
by Pat F. (not verified) :: Fri, 08/11/2006 - 2:02pm

This is a really interesting idea.

I think a more detailed examination using a robust computer model/simulation (perhaps modding an available sim game) would go a long way toward examining the effects of field position distribution, opponent strategies, etc.

To all those who are worried that losing possession in your own territory gives the opposition a huge advantage, remember that implicit in the author's use of averages is the assumption that an equal number of TODs will occur in opponents' territory, which are as good for you as TODs in your territory are good for your opponent. This assumption (and its subsequent ramifications) should probably be examined more closely, as the field position distribution problem referenced in #23 could really skew things by having more like 60-70% of the ToDs occur in your side of the field.

Also, keep in mind that the study is macroscopic in nature. In any given game it could blow up in your face (which of course is why no one's tried it, at the NFL level anyways), but the conjecture is that if you did this in all 16 games you'd overall come out ahead.

This article is a great first step, and many thanks to the author for doing all the research and putting this out there. The next step, I think, is to more accurately define the margins for which this strategy works (perhaps using the Point Expectation tables referenced in #80). That is, how good(or bad) does your offense and defense, and how good do your opponents', have to be for this to pay off over the conventional punting strategy. I know Jason offered evidence that you could be 2004-49ers-bad and still have it pay off in multiple wins, but I think opponent quality, the interrelatedness of possessions, and other factors may affect it, particularly on a game-by-game scale. It may be that the minimum margin for it to pay off is so low that no NFL team is actually that bad, but I'm envisioning something like the pinch-hitting guidelines that BP came up with a couple years ago: if you are this much superior to your opponent you should never punt, if you are this much inferior you should never, and between that it depends on the context of the given 4th down.

91
by Jerad M (not verified) :: Fri, 08/11/2006 - 2:29pm

My main issue here is in effect you are assuming a statistic that correlates well with wins with wins in "3 down football" would correlate just as well with wins in "4 down football." ATR obviously correlates well now but as the rest of your article argues not punting would totally change the game, is there any reason to believe that ATR would actually correlate as much to wins in a league where some teams choose to punt and some teams choose not to? Also I have some problem statisically with lumping all turnovers into the same bag for the reason of simplicities sake. I think this site has discussed that not all turnovers are equal and obviously a fumble would be much more damaging then a punt due to the relatative shit in field position from the turnover. But in the end I think ATR is the problem here. I'll grant I can't think of a better statistic to use to measure the impact of not punting, but I think not punting would be such a mammouth change in the game it would change that statistic a huge ammount we can't speculate on until we see it in action.

92
by Lou (not verified) :: Fri, 08/11/2006 - 2:45pm

Very interesting article. I think a most of the holes have been mentioned already, but it would be great to see one of The Outsiders examine this theory with further statistical analysis (hopefully before PFP2007.) Articles like this, and the ensuing discussions, are what make FO great.

93
by Nathan (not verified) :: Fri, 08/11/2006 - 2:45pm

I really haven't heard a compelling argument that doing it all the time, versus the best times would be any better.

I agree. Never Punting from your own 40 on is a better strategy. I'd also never Kick a field goal outside of 35 yards. (Whereever a high likelyhood goes to a poor one)

But it depends on position. I don't think it changes the four down strategy at all, it just adds more strategy, which is always helpful.

Adding more strategy makes those who are better at strategy, better at the game. Why make some standard never punt rule?

Play 3rd down like you aren't going to punt. If it's 3rd and 9 still, just punt.

But intend on not doing it. That isn't hard.

I was amazed no one added to my, "The swings of highs and lows is greater"

I think it's an important aspect.

94
by Nathan (not verified) :: Fri, 08/11/2006 - 2:50pm

The people saying “this would be insane on 4th and 12 from your own 30″ are missing the point - you’d almost never be in that situation. On your own half of the field, you’d be conservative and use high-percentage plays on 2nd and 3rd down.

What you're missing is, you can still try conservative plays on 1st, 2nd and 3rd down and be at 4th and 9.

And a 4 down mindset does not mean nothing but conservative plays. The biggest advantage in football is doing the unexpected.

You would definately wing some long throws and mix it up. Play 100 games of Madden like this, and you'll see what some of us mean.

I've played years and years of Madden and NFL2k with this strategy.

Saying it doesn't matter where you do it is crazy.

If you do it successfully once a game in the most optimal situation, and get 7 points where you would have had none, you are much more likely to win the game.

Why lower your chance of winning by trying it at least optimal situations, losing 7 points where you would have lost none, and having a less likely change to win the game.

I don't even need math matical proofs, play football in this style, you'll quickly start deviating to a middle ground inbetween never punting, and always punting.

Play online, play with friends, we've all found this basic senario where it makes more sense.

I'm glad someone wrote a ton on a topic us gamers have delt with a great deal, but try it out.

95
by Mr.X (not verified) :: Fri, 08/11/2006 - 3:44pm

ummm....what?

96
by David Brude (not verified) :: Fri, 08/11/2006 - 3:45pm

Re: 94

The go for often strategy works very well against a man opponent in Tecmo Super Bowl. Especially when you consider the average play (8 yards per rush, 12 per pass in this game) practically nets you a first down. I imainge this works well for most video games

The other thing that comes into play is the few possessions you get in the game and the high likihood any drive will end in a score.

If you can consistently get 4th and shorts, you can sustain drives. Of course going for it on like 4th and 25
at your own 10 and fail its a very high liklihood the other team will score. This strategy in Tecmo super bowl can deinitely blow up with an inferior player and cause them to get blown out more quickly but overall it probably improves their cahnces to win most of the time.

I think the happy medium is more apt for the NFL. Teams should definitely punt less often but all the time would be a bit extreme. The other problem is that if the strategy fails early and the other teams scores you become that much more one-dimensional due to time problems.

97
by Larry (not verified) :: Fri, 08/11/2006 - 4:00pm

The more I think about this, the more I think the negatives are understated in the analysis. Using 2.8 points per drive started by no-punt team TOD is pretty obviously understated. Turnovers occur on all parts of the field, while the no-punt team TODs would be exclusively near-midfield or better field position for the opponent, leading to higher scoring rates on the ensuing posession, not to mention the down the road field position effects even if there isn't a score.

Also, there is no way you can consider the ATR/win relation for a no-punt team. The nature of the average ATR for a no-punt team is completely different from the one that generated the initial relation. A +85 no-punt ATR may very well be the same as a +40 conventional ATR in terms of wins. By definition a no-punt team increases their ATR, since the change is turnovers (punts) either become tutnovers (downs or real on the ensuing drive) or scores. The question is whether those worse (and they are worse) new turnovers are offset by the added scores and longer drives or not. Only more detailed analysis of the new 4-down drives answers that. The point analysis above is a start to that, but probably not detailed enough.

Very interesting food for thought, though.

98
by PackerNation (not verified) :: Fri, 08/11/2006 - 4:42pm

I've been a successful HS coach. For years, I hated it when the opposing coach refused to punt it when I stopped him on 3rd down and instead, lined up to run another play.

I hated it so much I started doing it myself. Great results, great attitude builder on the offense. Get the job done or give them the ball at that spot. Even the defense seemed to play with more determination.

I really think Jason is on to something here.

If you look at the Packers, for example, who are expected to be weak offensively, AND have a weak punting game, (plus the most turnover prone quarterback in the league), it actually makes sense to just try to run the ball for two or three yards a pop, mixing in an occasional play action to the TE.

I'd also think that simply going for it on 4th down regularly would probably impact your time of possession numbers as well......which correlates positively with winning.

99
by Joe (not verified) :: Fri, 08/11/2006 - 5:22pm

Yeah, I agree with 97...the loss of 35 yards of field position by not punting is totally discounted in this analysis. Romer goes through a much more thorough (albeit much more complicated) analysis of fourth-down strategy. He DOES come to a similar conclusion, which is that teams don't go for it nearly enough on 4th down, but the "never punt" strategy is both intuitively and numerically wrong.
I respect what you're trying to do here, but your analysis is a perfect example of people's claims that "you can make numbers say anything". If you are careful and thorough, statistical analysis will give you useful results. Otherwise, though, you come to erroneous conclusions.

100
by DJAnyReason (not verified) :: Fri, 08/11/2006 - 5:23pm

I am upset that this is reply #99 in this thread and only the third mention of ROBOPunter

101
by Joe Pisarcik Magnate (not verified) :: Fri, 08/11/2006 - 5:42pm

#34-GREAT article, great read. I wish more football writers wrote like that.

Mike Leach should NOT have written that article.

/joemorgan

102
by Pat (not verified) :: Fri, 08/11/2006 - 6:44pm

I really haven’t heard a compelling argument that doing it all the time, versus the best times would be any better.

I agree. There's just absolutely no reason not to punt on 4th and 10 at your own 15, at any time in the game, unless you have the world's worst defense, and a great offense versus a crappy defense (in other words, something like 2004 Green Bay playing the Vikings).

I think the people justifying the approach from Madden results are neglecting that defenses are still some of the poorest portions of video game football.

4th and long, inside your opponent's 20? Blitz. Yah, there's a chance you could beat the blitz - call a run right at it and hope none of the blitzers can tackle the guy. Or hope the QB can get it off to the hot read fast enough.

But it's not like the other defense will care. You might get enough yards for a first down. So what? Now you're still deep in your own territory. Whee. Maybe 5-10 yards.

How often are you going to convert in those situations? 30% of the time? 40% of the time? Especially if you're inside the 20, you've probably got a better chance of gaining field position by punting the ball.

From the 20 to about the 50, I dunno. I could see arguments either way, especially depending on how good the special teams unit is. Past the 50, I'd probably agree, although it'd still be situation dependent. More than 4th and 10... why bother? What's the point?

I hated it so much I started doing it myself. Great results, great attitude builder on the offense. Get the job done or give them the ball at that spot. Even the defense seemed to play with more determination.

I think in high school you're also dealing with lesser athletes, and stamina is probably a much bigger issue. Trying to wear out a defense is probably pretty valuable. At the pro level, though, while some players are going to get frustrated, the potential for a big stop and the ability to put the offense in great field position is going to dominate.

103
by Bobman (channeling Joe Theisman) (not verified) :: Fri, 08/11/2006 - 6:56pm

I talked with ROBO PUNTER last night; in fact we had diner together then went dancing and caught a movie. And ROBO PUNTER said he LOVES this idea, even if it means he's out of a job.

I had to agree. After all, he sprang for the popcorn.

104
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 7:36pm

Re: #91 and #97 - say what you want about the never punting part and about my calculations and logic around that, but I feel pretty strongly that ATR itself is a solid statistic. ATR will always correlate to wins the same way whether you punt or not. The reason is because ATR is BY DEFINITION the inverse of net points, and net points is what determines whether you win or lose a game. For the correlation to wins to change by not punting would mean by definition that the correlation of net points to wins would have to change and there's no basis to say that. You still would end up winning or losing games based on the score. If there would be any issues with this theory, it would have to be in the calculations of what ATR would change to by not punting, but not in the correlation of ATR to wins. That's just too foundational.

105
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 8:24pm

Re: #81 and #86 - I don't think the definition is an issue. It's all just how you look at it. The drive doesn't end by you giving the ball to the other team. The drive ends at the point the score happens. There is a score at the end of the drive, and it's likely they just decided to give the other team points rather than take away your points but that would make more sense. And it's also likely that the different kind of 'kickoff' and different opponent's field position associated with it is to compensate for only giving the other team 2 points. Safeties have a quirky scoring system that I think is the issue rather than the definition of a turnover. Did you lose a chance to score? In one sense maybe, but really you did score - for the other team, or negative points, however you want to look at it - and then kicked the ball to them just like any other score.

You had a drive, it ended in a score, there was a kick to the other team. I don't see an issue.

106
by Pat (not verified) :: Fri, 08/11/2006 - 8:35pm

The reason is because ATR is BY DEFINITION the inverse of net points

It's actually closest to the inverse of net scores, which is why there isn't a perfect correlation between the two.

It's still not quite the inverse of net scores, because of two things: 1) teams trade possessions, and 2) teams actually aren't required to have an equal number of possessions. One way is overtime, but also if the coin toss winner chooses to kick, the coin toss loser can choose to receive in the second half. That team can therefore still win even though the two teams could have equal actual turnovers.

107
by Pat (not verified) :: Fri, 08/11/2006 - 8:47pm

1) teams trade possessions

Bleah. This was supposed to have more. Anyway. It was meant to say that teams trade possessions, except at the end of a half... unless you're really careful about defining a turnover, and claiming that 1) ending a half with possession is a turnover, and 2) ending a half with a score is a turnover for the other team.

Which actually makes a bit of sense - it's definitely good football strategy to try to end up with one more possession than the other team by leaving them as little time as possible in the half.

You had a drive, it ended in a score, there was a kick to the other team.

Well, if that's true, then safeties would be another reason why ATR wouldn't correlate with wins perfectly. If a game had only a safety, there'd be no turnovers, and you'd lose.

108
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 9:00pm

Re: #5 - I understand the number of drives in a game between the two teams isn't always equal, I mentioned plus or minus one but from what you said I suppose on a rare occasion it could be plus or minus two.

What I mean when I say ATR is the inverse of net points is that a score is the opposite result of a drive compared to a turnover and every team will have an average of what their number of points per score is. And so every team will have a net points value that corresponds to their ATR more or less. Is it a perfect correlation between the two? No, but it is a real correlation. What I was trying to say in the response
to the other comment is that the two are related (even if not perfectly) and you can't have one correlate to wins in that much more drastically of a different way than the other one. To use the numbers he used, if +85 ATR when not punting correlates to the number of wins that +40 would when punting, that should throw your net points way off as well, even if not to the exact same degree. But how can net points correlate that much differently to wins? If your net points gets better, you win more. If your net points gets worse, you lose more. You would then have to say you could improve your net points by a significant degree (even if not by around twice as in those numbers it would still be a lot) and still not win more games. That doesn't make any sense to me.

109
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 9:01pm

Sorry, #107 was in response to #105

110
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 9:08pm

Re: #106 - Of course it's not perfect, you can have games like the 2-0 game you mentioned. You could also have a game with the same number of drives and one team has say 4 turnovers and 6 FG's and the other team has 7 turnovers and 3 TD's and wins with a lower ATR. Over the course of an entire season those kind of results should average out. The fact that ATR correlates to wins shows that it does. But given the example I gave, there's no argument for counting a safety as a turnover just because you win without having a better ATR.

111
by Pat (not verified) :: Fri, 08/11/2006 - 9:19pm

But given the example I gave, there’s no argument for counting a safety as a turnover just because you win without having a better ATR.

I don't get it - why? The only difference between a safety and an interception/fumble recovery for a touchback is that 1) a team gets 2 points, and 2) there's a kick instead of an automatic placement at the 20. Except one's a turnover, and one isn't.

The reason I said net scores rather than net points is that net scores is much more accurate than net points. Net points have the field goal/touchdown difference (which can be very big: GB/PHI 2005). Net scores have only really really minor tweaks.

112
by Jason Scheib (not verified) :: Fri, 08/11/2006 - 9:51pm

Re: #110 - it goes back to what I said early in the article about implied drives. An fumble/interception return for a touchdown had a drive change. Your drive ended at the point of the turnover. Your opponents drive ended in a score without their offense playing a down but is still a drive. On a safety your drive ends in a score and THEN you kick. It's the opposite. For it to be the same thing, you have to establish that your drive ended in a turnover, then the other team has a drive ending in a score (which also is different because unlike a fumble or interception return they never touch the ball), then you have another drive ending when you kick but that would be yet another turnover if that were the case. It's also different because while in a return the turnover and score happen on the same play they are not the same event in the play which they would be if a safety was a turnover, and if they were the same event there's no place for a drive change inbetween like there is on a return.

113
by Anonymous (not verified) :: Fri, 08/11/2006 - 10:54pm

Re: #54, the real issue is if including safeties in ATR would improve or reduce the correlation with wins. I.e. does counting a safety as a turnover make the statistic more or less useful?

Personally, I think that counting scores by the other team as avoiding a turnover is wacky. A safety ends your drive without your team scoring; it should be a turnover. A safety is essentially a forced punt that also gives the receiving team points.

By the same logic, when Barry Foster let some Steelers opponent recover a kickoff in the endzone, that wouldn't be a turnover. Your implicit drive ended with a score by the other team. I would be able to live with this wackiness if it actually improved the model. However, I find it very unlikely that it does.

Just because the model is counter-intuitive does not mean that making it more counter-intuitive makes it better. It is good *despite* being counter-intuitive. Calling an onside kick a turnover against the receiving team makes perfect sense to me; calling a safety a successful scoring drive by the kicking team is just wrong.

It would make more sense to call it two turnovers, the first of which ended in a score by the other team (returning possession to your team) and then you punt it away as a second turnover. Think about it. Calling a safety two turnovers makes ATR an exact inverse of net scores by utilizing the same implicit drive mechanism as an onside kick. Defense forces turnover and scores (just like if they recover a ball the offense dropped in the endzone); returns the ball to the offense, whose implicit drive ends with a mandatory punt (second turnover). This is far more rational than calling a safety a successful scoring drive that ends in a kick.

I'm with you with everything but two things, the safety and choosing punts by situation. You argue that the averages work best if you always punt; however, you can split into two sets of situations (should punt or not) and make new averages.

Going for it on 4th and 26 from your own 4 does not make sense in net point terms, only in a world where a safety is better than a turnover on downs. Taking that play out will mess with your statistical prediction, but it will improve your overall odds of scoring on 4th down.

In re 4th and long, what if you always throw a hail mary? If it's intercepted, it's just like a punt. If it's received, it's a first down.

The only risks are that it might fall incomplete or that you take a sack. Any decent offense should be able to avoid a sack in that case (think quick pass into a depleted secondary if they blitz). Similarly, you can put a lot of people in the same area. Remember that an interception doesn't scare you, just a drop. You can throw into traffic safely. The risk of a run back on a hail mary interception is extremely low; too many people around.

I'm not sure that distance is the primary problem with going for it on fourth down. I think that field position is much more important.

114
by Pat (not verified) :: Fri, 08/11/2006 - 11:08pm

Yeah, I have to say, that really makes no sense. I can't see how a safety is any different than a fumble/interception touchback. The drive ended, your opponents get possession. The fact that a score happened is immaterial - your score didn't increase. Your opponents did.

I think the other reason it doesn't make sense is you spend most of the article saying boosting ATR is a good thing. Safeties help boost it (one less drive ending in a turnover), but definitely are not good.

115
by Pat (not verified) :: Fri, 08/11/2006 - 11:09pm

In re 4th and long, what if you always throw a hail mary?

If it's 4th and long, it won't be intercepted (except by rookie DBs). They'll bat it down.

116
by Englishbob (not verified) :: Sat, 08/12/2006 - 7:34am

112 said exactly everything I wanted to get at re safeties. I can only assume Jason is focusing purely on the statistical outcome and ignoring the actual practical application- kind of from the school of "lies, damned lies and statistics". Stats as we know can be very easily misinterpreted, which is why FO developed DVOA/ DPAR in the first place.

On long bombs as quasi punts- any safety with half a brain is just going to bat the ball down. I remember watching the Denver v Miami game and thinking "idiot" when Plummer's bomb wasn't just knocked down.

Great to have quotes back too!

117
by Jason Scheib (not verified) :: Sat, 08/12/2006 - 10:44am

Re: #113 "I think the other reason it doesn’t make sense is you spend most of the article saying boosting ATR is a good thing. Safeties help boost it (one less drive ending in a turnover), but definitely are not good."

I will say that's the best argument I've heard yet. Let me think about it and get back to you.

That said, I'm amazed at how much discussion is happening over safeties considering they are rare and therefore such a minor part of this whole discussion.

Re: #115 - I'm just trying to be true to the definition, that's all. I'm not trying to ignore anything. I know safeties have a negative effect to a team. I'm just being as careful as I can to define them properly.

118
by Lou (not verified) :: Sat, 08/12/2006 - 10:45am

I can't believe know one else thought of 112's safety idea. Well done sir!

119
by Nathan (not verified) :: Sat, 08/12/2006 - 11:20am

Admittedly, some of us thought the Safety issue was irrelevant because it's a rare instance, and semantics on rare instances can be delt with after all the real issues are.

:)

120
by Will Allen (not verified) :: Sat, 08/12/2006 - 2:47pm

I've been busy for the last two days (damned pesky clients!) so I haven't had time to really dig into this piece, but just from having skimmed it, kudos are due to the author, for even if there are flaws to his analysis, it is damned thought-provoking. I think Pat's reference to the gambler's fallacy is quite apropos, however.

No doubt there have been at least some teams that would gave experienced a net benefit from never punting, due to the lop-sided nature of talent distribution on their roster. On the other hand, imagine what the average margin of victory would have been for the '85 Bears, the '00 Ravens, or the great Steeler teams from the mid'70s, if nobody would have punted against them? The Ravens in particular, given how anemic their offense was, would have loved it if teams refused to punt; it would have greatly increased their chance of scoring, while increasing their opponents' chance of scoring very little. There is no such thing as an average football team.

Having said that, I have little doubt that teams punt too much, especially teams with bad defenses and good offenses.

121
by Ghandi (not verified) :: Sat, 08/12/2006 - 7:51pm

The flaw here is the assumption that it is necessary to never punt to get benefits rather than punting less. It assumes that a team needs to never punt to get significant results, but punting in situations in which the statistics are favorable would increase a team's score differential over the course of a season much more than being an ignorant fool and never using the punt. Choosing to never use the punt (until late in the game) is like choosing to always use the punt--there are situations in which punting is advantageous and not advantageous.

122
by Brian Sheppard (not verified) :: Sat, 08/12/2006 - 10:17pm

Am I the only one who thinks this analysis is just plain bad? I'm not talking about football, either. The math is wrong from top to bottom.

You cannot recommend a strategy for every situation by saying that it works in the average case, so it must always be a good idea.

The dynamic programming models are based on vastly superior math, and produce much more convincing recommendations. Something along the lines of "go for it on 4 yards or less unless you see the shadow of your own goalpost." And I am guessing that human/psychological/emotional considerations make that recommendation too aggressive on average. That's just me guessing, but it is the way pro football is actually played.

123
by Grabowski (not verified) :: Sun, 08/13/2006 - 12:00am

Hate to add fuel to the flames over the safety conundrum, but I just have to put in my two cents: The way I see it, the problem with the author's reasoning on safeties lies with his root assertion that "a drive must end with a score [regardless of which team scores] or a turnover." If ATR is supposed to benchmark success-or-failure, wouldn't it be better to reason from a paradigm of "with the exception of time expiring at the end of either half, if the team possessing the ball or receiving a kickoff does not end their drive with a score for THEIR TEAM, then a turnover should be assessed?"To support this argument, I cite the example of the "intentional safety" where a team backed up against its own goal line takes a safety on purpose, sacrificing two points to gain 20 yards of field position rather than punt from their own end line. ATR as computed by the author would assess a turnover for the end-line punt, but not for the intentional safety, even though both cases involve purposfully turning the ball over to the opponent. Why should the former be considered a turnover but not the latter?I would further argue that it needlessly complicates things to insist that a drive ends once a safety occurs. Why not just say that a drive continues after the safety, but with a forced free kick? The drive should actually end when the ball is free kicked away, and it is at THAT POINT that the ATR turnover should be assessed.In an extraordinarily rare circumstance, the kicking team could recover the free kick (and yes, just like with a kickoff, you CAN attempt an onside free kick and recover the ball after it's gone ten yards, but it's so risky that I've never seen one attempted in my lifetime) in which case the turnover and drive count should be handled just like in the case of a punt fumbled by the receiving team and recovered by the kicking team.A safety should be considered a special case where a score does NOT end a drive (and to further keep things simple, I agree with the comment that a safety should be looked at as "Team A scores -2 points" rather than "Team B scores 2 points"; that way you don't have to get into whether or not Team B had an "implied drive" to score the safety).Now if you REALLY want to get esoteric, answer this: how should the turnover and drive count be handled in the case of an NCAA defensive conversion or a CFL rouge?

124
by Jason Scheib (not verified) :: Sun, 08/13/2006 - 1:49am

I know there are flaws in my analysis. I expected my math to not be completely accurate. Like I said, I'm not a mathematician, and by saying that I never claimed that my math was correct. I had an idea and I wanted to see if it works, so I pursued the answer to that question as best as I knew how. The reason for putting my analysis out there was not to claim I solved everything, but to get the ball rolling by seeing if I could get anyone with a better math background to check my analysis and see if it really does work or not. It doesn't bother me if it doesn't work, I just wanted to know for certain one way or the other.

Thanks to those who have given their encouragement for my just getting this put out there for people to think about. And that said, I have two challenges to put out there:

1) I've seen a lot of arguments here about why people think this would not work. And they may very well be right. I challenge anyone to figure out what all the correct logical arguments are and use whatever appropriate mathematical model to calculate how many more games or less games a team would win if they never punted. If it's less, I'm ok with that, at least we would know and it would quantified. I'm probably more curious than anyone as to the exact answer to this question.

2) My second challenge is to not throw out the baby with the bathwater. If never punting doesn't work, don't by association write off Actual Turnover Ratio. ATR is a valid statistic that has meaning. And the bottom line is it still correlates to wins. Thanks to BostonDan for understanding this. Dan, I see on your site that you are going to track ATR for the upcoming season for people to see. I think that's great. I'd love to see the idea of ATR live on. Before you do that though, we should probably resolve the debate around safeties. I'm going to close with a thought on that:

Many thanks to Grabowski in post #123. My goal was to be true to the definition of a turnover in defining safeties, and in all the other arguments against my stance on safeties, you are the first person to offer a suggestion of an adjusted definition of turnovers that not only causes a safety to fit nicely in that definition but that also does not reduce the integrity of the definition as it applies to other events. I'm still not 100% sure that a safety should be a turnover, but I'm also not so sure anymore that it's not. At this point, I'll conceed that it is. And since they're so rare it shouldn't affect ATR much at all.

Thanks to everyone for your comments. I appreciate all the feedback, really I do. Positive or otherwise.

125
by Mike H (not verified) :: Sun, 08/13/2006 - 4:51pm

What about field goals? With a 4-downs mindset, should teams bother kicking for 3 points, at least in the opening stages of the game? Why not go for it all?

126
by empty13 (not verified) :: Sun, 08/13/2006 - 8:42pm

I dont play Madden but I used to play other games wbw and playing all 4 downs was often a viable strategy in those. It helps to have something (anything) going for you on offense, so you can use it or play off it tendency-wise. I wouldnt have used the play 4 downs strategy at any time with the 2005 49ers, for instance.

If you dont score, you have to turn the ball over, so every time you dont acore, you commit a turnover. Just because a punt may rescue you from horrible field position, or put the opponent in horrible field position, doesnt mean that it wasnt a turnover. One has much more scoring potential (4 points worth per Bill James) when one has the ball, at a given point on the field. Besides that, numerous teams have return specialists that kicking/punting teams try like all hell to kick away from or deny half the field to.

And yes 87/98, I would employ such a (4 down) strategy against a nominally "good" defense, but vulnerable against the run - with a small front four, ie, the Atlanta Falcons. Wear out the defense and watch a poor passing offense struggle to catch up. So, the opponent's offensive approach may be a consideration as well.

And the sooner you destroy the other team, the sooner you can rest your studs.

Also, the way some teams manage the clock in the two minute drill equates to additional turnovers. Seattle, anyone?

127
by Grabowski (not verified) :: Sun, 08/13/2006 - 9:11pm

Re #124: It took some stones on your part to put out a detailed analysis backing an idea that seems prima facie... well, to be frank... insane. For that you are to be commended. Mouse Davis must have been called looney-tunes quite a few times as well (see this excellent article for more details) but he changed the game forever. I have grave doubts that any team will or should ever wholly adopt the four-down mindset, just like no one would ever run the pure spread any more, but if you're on the right track, maybe offensive coordinators should seriously think about retooling their offense to go for it more and punt less on fourth down.

128
by Jake S. (not verified) :: Mon, 08/14/2006 - 2:45am

I'm trying this in Madden. Of course, online play requires you to punt most of the time. I should send this their way.

129
by brent2mnen (not verified) :: Mon, 08/14/2006 - 3:23pm

Safeties are definitely turnovers. Whether they count as 1, 1.5 or 2 is debatable, but they are definitely turnovers. What would the result be if every drive for one team resulted in a turnover? Their ATR would be 0, but I suspect that they would not win, even though they technically never turned the ball over.

130
by Mike Shanahan (not verified) :: Mon, 08/14/2006 - 4:47pm

Several years ago the Patriots took a safety late in the game on fourth down against the Broncos. Shortly thereafter, the Patriots won that game. Bill Belichick must have known that ATR correlates well with winning. He correctly deduced that when facing 4th and long from his own goal line, the best way to increase his ATR and by extension increase his team's chance of winning was to take a safety. If safeties were counted as actual turnovers, then that strategy would not have been appealing to the Patriots. Therefore, safeties should be counted as turnovers, if only to confuse that know-nothing hack Belichick. I'm the mastermind, and he's ruined my street cred.

131
by Andrew (A.B.) (not verified) :: Mon, 08/14/2006 - 5:14pm

Re: #122

You're not the only one. Too many flaws in the analysis completely obliterate the argument.

If FO were a peer-reviewed journal, this article would be sent back to Mr. Scheib for a whole lot of reworking.

132
by Englishbob (not verified) :: Mon, 08/14/2006 - 5:36pm

Their may be issues with Math and defining safeties as not turnovers, but give the man his due. He has gone out and written a thought provoking article that whilst flawed is worthy of reading and for that I thank him.
Anyone with really strong criticisms can always re-work the numbers themselves.

133
by Larry (not verified) :: Mon, 08/14/2006 - 7:41pm

Re: 104

The point is that the ATR/win correlation is based on a particular mix of FG/TDs, which will change if you start giving the ball up on your own 40 a lot, as opposed to punting it to the opponents' 20, plus even when your opponent doesn't score, you'll lose the field position battle, adding to the change in FG/TD mix. It's true that extra scores your opponent gets will be reflected in your ATR, though, so you probably wouldn't have something as extreme as what I postulated in #97.

I think what you need to do is consider the 4-down mindset together with the Romer analysis. Namely, on 3rd down, you don't need the first down, you just need to get enough yards to make it worthwhile to go on 4th down as opposed to punting. If you don't get it, you punt. This is different that looking at a chart on 4th down and deciding to go or not as the impact of knowing when you'd go for it on 4th down would cascade through every play calling decision you'd make. But, you wouldn't cut your punter.

Tracking ATR seems a lot like drive stats to me, and reminds me of Jim Armstrong's Drive Success Rate chapter in PFP 2005.

134
by Andy (not verified) :: Mon, 08/14/2006 - 11:03pm

Was Paul Hackett reading this article when OC at the Jets, it was just that everyone else didn't realise that he was waiting to go for it on a 4th down that never happened. His use to draw plays to get to 4th and short situations seems to fit in with these ideas.

135
by Mr Shush (not verified) :: Tue, 08/15/2006 - 12:32pm

Can I just second (or third or fourth) the calls for this study to be reworked in greater detail by someone with a mathematical/statistical background (Aaron? Pat?), paying particular attention to issues like the likely distribution of field position for turnovers on downs? Jason's maths may be off and/or damaginly simplistic, but until someone does the maths right, we won't know for certain that his conclusion is wrong (or at any rate that teams shouldn't adopt an "almost always punt" strategy).

In any case, Jason, kudos, but I'm afraid Todd Sauerbraun knows where you live. Call 1-800 G-R-A-M-M-A-T-I-C-A now if you want to live.

136
by Grabowski (not verified) :: Tue, 08/15/2006 - 10:16pm

Re #130: I was initially tempted to buy your argument, Coach, but then I thought about it and changed my mind. Correlation with winning is not the same thing as equating with winning. Strategically turning the ball over is still, nonetheless, turning the ball over. By electing to take the safety, you have still chosen to (99.99% of the time) punt away, albeit from (presumably) further up the field.Remember, the author himself argued that even with a full-on four-down mindset, there are still exceptional situations where a punt is merited. Said punt, however, is still an ATR turnover.

137
by Ben (not verified) :: Wed, 08/16/2006 - 2:14am

Re: 128

Online play in Madden requires you to punt most of the time? Really? How does this work?

138
by Brad (not verified) :: Wed, 08/16/2006 - 10:25am

The detailed analysis using a dynamic programing model has been done here taking into account field position, score, and time

http://www.footballcommentary.com/goforittables.htm

The model calculates the probability of success you should need to go for it. During the fist half in a tie game you should only go for it from your own 5 if you have greater than a 70% chance of success. You should go for it from your own 20 only if you have greater than a 60% chance of success.

This seems to indicate that you should seldom go for it inside the 20 at anything greater than 3 yds or 1-2 yds inside the ten. That doesn't mean that you can't imploy a 4 down strategy. Just if that strategy doesn't work you should not go for it on 4th and 5 from your own ten.

139
by Stung (not verified) :: Wed, 08/16/2006 - 11:22am

My girlfriend's Madden offense never ever ever punted. It would drive me crazy, but she'd look at me like I was a moron and say "I complete 75% of my passes, I have a great fullback, I make 80% of all my fourth downs. Now shut up while I score again."

In unrelated news she scored something like 150 points a game.

140
by Pat (not verified) :: Wed, 08/16/2006 - 4:05pm

The detailed analysis using a dynamic programing model has been done here taking into account field position, score, and time

One thing it doesn't take into account, though, is the quality of the opposing offense, and the quality of the same team's defense. It takes into account the quality of your offense/opposing defense by saying "if you think you have a better than X% chance of getting it", though.

I do think a four-down mindset (and punting very rarely) could reap a lot of benefits when playing unbalanced teams - especially if you're unbalanced yourself.

I think that the Tennessee/Indianapolis in 2004 game illustrated this pretty well, although Tennessee's offense just wasn't good enough to overcome it.

141
by Kevin (not verified) :: Wed, 08/16/2006 - 4:12pm

OK, I see one big problem with the mathematics. You assume the average "turnover-on-downs" will be at the 50 (and hence use the 2.8 pts per turnover). Well, 2 reasons this is wrong (one big, one small):

1) Big reason: If you are kicking field goals on 4th down when you are in field goal range, then you are taking out the turnover-on-downs possibilities in that range. That means you have a limited field, say your own 1 to the opposing 30. This means your average attempt would be more like your own 40.

2) Small Reason: Real turnovers usually are given away down field (especially interceptions), while TO-on-downs occur at line of "failed"-scrimmage. This may affect it a bit too.

142
by bit (not verified) :: Thu, 08/17/2006 - 1:46pm

sorry guys this dog don't hunt. Not punting sacrifices field position and even in theory would only work out if the other team also didn't punt. It's like saying going all in preflop with J-2 offsuit is the way to win poker hands. Yeah ok if you are playing against rhesis monkeys but sooner or later it's gonna knock you down so hard you can't get up.

143
by Brian (not verified) :: Thu, 08/17/2006 - 3:00pm

Interesting article, and not only because it killed a full hour of work to read!

I do think that the argument suffers somewhat from the mathematical simplifications. The optimal strategy would likely involve making punt/no punt decisions based on field position and score, as numerous others have pointed out. The disappointing part about that approach is that play calling is still made largely with a 3 down mentality. The big allure of Jason's analysis to me is calling plays with the knowledge that you have 4 downs and must only average 2.5 yds/play.

On the other hand, the thought of Andy Reid's play chart growing even larger to encompass the new punt/go-for-it table is worth the compromise!

144
by Dan M (not verified) :: Thu, 08/17/2006 - 3:33pm

There are a lot of comments about field position, suggesting that the author didn't take into account the fact that a turn-over on downs in one's own territory will result in more points than a punt.

The author clearly factored for this into account by saying a turn-over on downs should result in the same points-per-turnover as an interception/fumble. That's where he accidently said that this would be the points scored starting from the opponents 50 yd. line. He meant that an interception/fumble would result in a scattered set of turn-overs that could end up averaging around the 50 yd. line, but clearly account for a much higher score -per-turnover than a punt would.

He's already factored in the field-position argument.

This analysis is actually quite well-done because field-position doesn't enter in as a factor, just scoring rate per turnover.

The only field position factor not accounted for is the average yards gained by an opponent on a interception/fumble might make the 2.80 assumption a little low. If on average most turnovers on fumbles and interceptions end up past the line of scrimmage of the team being intercepted then there would be a higher scoring rate on turnovers off of downs than off of interceptions/fumbles. But I don't think this would be enough of a factor to discredit the analysis.

145
by rb (not verified) :: Thu, 08/17/2006 - 6:24pm

Nice article; the premise that NFL teams should punt less often I regard as self-evidently true.

However, it is trivial to show that 'never punting' is a silly idea, even if (and this is a *big* if) it's superior to punting on every 4th down.

Because if this theory works when you keep it this simple, why make it more complicated?

Because: a coach doesn't want to simply up his/her winning totals just a bit by employing highly unorthodox strategy; rather s(he) wants to *maximize* the probability of winning by whatever method is available.

More specifically: to the degree possible the coach should maximize the strategy of success in every situation, and it is easy to imagine a situation where punting is optimal (as a matter of fact rather than opinion).

In baseball, it is generally accepted by the numerate that "never stealing" is a superior strategy to situational base stealing as it is commonly employed. However, if I have a guy on 1st who is nearly impossibly to gun down and the pitcher is a knuckleballer and the catcher has a weak arm and bad knees, "never stealing" suddenly becomes silly.

Much the same: in football, do NOT go for it on 4th and 25 from your own 3 yard line, or even your own 43. You are not going to convert 4th and 25 (that is: a turnover as defined in this article is essentially inevitable) so the only thing to be done is mitigate the damage embedded in that turnover.

So: on 4th and 25, if you can't kick a FG, punt. This is inarguable, and the same would apply to much less extreme scenarios.

146
by Mark (not verified) :: Fri, 08/18/2006 - 12:18am

Re: #122, No you are not alone. This "research" piece should never have made it onto this site. The author does some very light, imprecise work (while filling with an inordinate number of words), and proceeds to reach a totally unsubstantiated conclusion. People have done some real and meaningful work on this topic. Why undermine it all with publishing a "no, if you just keep it real simple and do this every time instead of trying to optimize your strategy" argument? I suspect the footballotusiders editors just skimmed the article without realizing how poorly executed the study and its conclusions really are.

147
by Jason Scheib (not verified) :: Fri, 08/18/2006 - 1:49am

I know there are several issues here, but here is something in regards to the 2.80 point value that is obviously too low. I just now looked at the last chart in the article again to see what the maximum point value could be for turnover on downs in place of punts and have this still work (in other words, the point at which the total turnovers by the opposition comes down to the same number as for you, or 5.51). Since the difference in turnovers per game there was 1.28, subtract that from the 1.35 in that spot on the chart (to get .07) divided by 3.06 times per game to get a turnover value of .023. That corresponds to a point value of 5.41, or a 97.7% chance of scoring from the average field position corresponding to that point value. Teams in the red zone don't hit that kind of a scoring rate, and we know that wherever the average starting field position after these turnovers on downs would be, it wouldn't be back in the red zone, either. So anywhere in the 3's and 4's, which is more likely, still gives you a positive ATR on average.

It would take some work to figure out what that point value would really be, and like I said there are other issues with this analysis, but if this is one of the most glaring flaws, this at least shows that if there are fatal flaws in this analysis, this isn't one of them.

148
by DGL (not verified) :: Fri, 08/18/2006 - 12:13pm

Average points per drive of every drive started in the red zone last season and postseason was 4.74.

Average points per drive of every drive started in FO's "front zone" (opponents' 21-40) was 3.01.

Average points per drive of every drive started at the opponents' 10 or closer was 5.51, but there's a very small sample size (only 51 drives).

149
by David M (not verified) :: Fri, 08/18/2006 - 2:18pm

Nice read. If the writer would drop the logical fallacy that you can never punt this could be really useful. My take away is: "if you are a good running team, play more conservatively on third downs between your 35 and the opponent's 35".

150
by Jason Scheib (not verified) :: Fri, 08/18/2006 - 6:36pm

Re: #149 and others - I didn't say to never punt no matter what. There is a section called 'Exceptions' where I said "Or maybe because you are backed up to your own end zone facing fourth-and-20." That was just one example. In other words, this leaves it open for punting in certain situations. The numbers were just to see what the outcome would be if you never punted, but that doesn't mean you couldn't do even better by punting in an exception situation. So the "logical fallacy that you should never punt" is not in my article, but rather said to be in my article by those who didn't read the whole thing.

151
by Jason Scheib (not verified) :: Fri, 08/18/2006 - 6:45pm

I also feel the need to comment on the alleged "Gambler's fallacy" mentioned in comment #76 and echoed elsewhere. All I said is that you need 1 extra score on average per game. Of course you won't get that extra score every game. I didn't insinuate that you would. Sometimes you would get more than you need and other times you wouldn't get any. Over the course of a season though, that 1 extra score per game would be 16 extra scores, corresponding (roughly) to 16 less turnovers, and affecting ATR by about +16, which of course a better ATR correlates to more wins. Just because it correlates doesn't mean you will always get the win total that correlates to your ATR, just on average you would.

If I'm still missing something on this whole "Gambler's fallacy" let me know. But using the link given and reading the definition, I don't think there's an issue here.

152
by Jason Scheib (not verified) :: Fri, 08/18/2006 - 8:18pm

one more thing to add to #151 before someone responds to it. That's *IF* you scored one extra time per game on average. That's not saying you would. And the one extra time was just the minimum needed, the actual came to 1.82 extra scores per game. Again, on average. Not saying you would hit 1.82 every game or even for the season, it's just the expectation.

153
by Pat (not verified) :: Sat, 08/19/2006 - 2:54pm

If I’m still missing something on this whole “Gambler’s fallacy� let me know. But using the link given and reading the definition, I don’t think there’s an issue here.

Where the gambler's fallacy comes in is where you say "If you go for it some times and not others, it could hurt you, because you need to go for it on every one of the five times teams typically punt in order to get the averages to work in your favor."

This is the "wrong" part. It's basically a different form of the gambler's fallacy, though it's a bit difficult to see. What the original poster was saying was "Maybe you should only have a four-down mentality on the other side of the 50 yard line" - thus implying that going for it on fourth down on your own side of the field is more dangerous than in your opponent's territory. Your response was that you always need to go for it in order for the averages to work out, and that's wrong - if there are better times to go for it, and worse times to go for it, there's no reason to go for it on the worse times to make the better times work out.

The response to the last question similarly has the same problem: it suggests that there's no problem with going for it on 4th down in your own territory, because it would only happen a few times a game. That somewhat misses the point - there's no reason to do it at all.

Let the likelihood of you making the 4th down be X, the likelihood of scoring a touchdown on the drive after making the 4th down be Y, and the likelihood of scoring a field goal after the 4th down be Z. Also let the likelihood of your opponent scoring a touchdown be A, and the likelihood of your opponent scoring a field goal be B. Then it's just a simple question: is X*(7Y+3Z) even if you make it, whereas the chance of your opponent scoring is high.

You can think of it like this: what you're really interested in is increasing your ATR - that is, your net turnovers. You don't care if you turn the ball over 50 times in a game. If your opponent turns it over 60 times in a game, you're still very likely to win. So suppose your opponent has a chance p1 of turning the ball over after getting the ball at your current spot. Suppose your opponent has a chance p2 of turning the ball over after getting the ball after a punt. Clearly p1 is going to be smaller than p2. Deep in your own territory, p1 is going to be very, very small, and p2 is going to be much larger. The question then becomes, if my chance of making the 4th down is X, is X*p1 less than (1-X)*p2?

154
by Pat (not verified) :: Sat, 08/19/2006 - 3:01pm

Your response was that you always need to go for it

Well, nearly always. You gave moderately extreme examples where you wouldn't. My point is that those examples are much too few. You punt whenever it's so unlikely that you make the 1st down and still score on that drive that any benefit you would get from making it is wiped out by the benefit your opponent gets in field position. The comment here:

Waiting until you cross the 50-yard line means you might only go for it one or at most two extra times.

indicates that as well. It doesn't matter how few times you go for it. If it improves your odds, it improves your odds, and it's the right thing to do. If it reduces your odds - which it probably would on the other side of the 50 (I'm just using the 50 here for convenience - really, see the 4th Down Go For It tables previously for a decent attempt at figuring out where it turns over) there's no point to go for it.

The next statement regarding the four-down mindset seems a bit confusing as well. One of the benefits with the four-down mindset is that you can still choose to punt if your third-down attempt fails to put you in better position. You can call, for instance, a 7-yard hitch on 3rd and 10, because you think you can get a 4th and 3. But if that pass is incomplete, and you're in 4th and 10, you can still punt. Nothing's forcing you to go for it on 4th and 10.

Honestly, I sometimes wonder if Mike Martz thinks like this. I've seen him call pass patterns about a yard or two short of the first down all the time.

155
by Jason Scheib (not verified) :: Sat, 08/19/2006 - 4:03pm

Thanks Pat, that's fair and I understand that. I think where I caused some confusion (and even for myself) is in talking about two different things here. The goal of this article was not so much finding the optimal strategy as much as just trying to see what would happen if a team never punted. If a team could win more games by never punting, that's all I wanted to find. If there was an inbetween strategy that was even more successful, that would (should) be a different topic altogether. Except that because it seemed to work I thought of it as the optimal strategy and said some things to that regard. So what I need to take back is anything that indicates never punting is the optimal strategy. Even then, that doesn't take anything away from the conclusion that a team would win more games by not punting. If true, it just means it's better than the conventional strategy, not that it's the best strategy. And by looking at it that way allows for more exceptions so that it's not more extreme. And yes, I understand that a team could play with a 4 down mindset and still punt if the 4th down position was not favorable. I was just concerned that if a team punted too often, they would fall into a 3 down mindset. But maybe that's not the case.

156
by Jason Scheib (not verified) :: Sat, 08/19/2006 - 4:16pm

On another note, I went through and found the average field position when teams punted (in the 2004 season to be consistent with everything else) to be at the 36.4 yard line. You can't use that exactly to be the point at which the opposition starts their drive if you fail to convert, because if you go for it you are likely to gain a little bit of yardage even if you don't convert, maybe a yard or two. And going by the figures given in comment #148, drives starting in the opponent's 21-40 had a point value of 3.01. Those drives would average in the middle of that range at the opponent's 30. And we're farther than that, closer to the 40, so the point value would be smaller yet. Maybe the 2.80 is pretty close to being accurate after all. That would also mean that fumbles and interceptions don't average around midfield but slightly on your end of the field.

157
by cjfarls (not verified) :: Sat, 08/19/2006 - 7:07pm

As I read this, I'm starting to wonder about what effects using average (i.e. mean) starting position, vs the median and those effects on the analysis. The net points benefit of successfully converting the 4th down vs. them gettig the ball at the 36 just seems way low to me... it sems like most of the time a teams starts that close, they at least kick a field goal... how much does the few times they get stuffed impact that number, and if you're really trying to minimize you ATR, doesn't giving a 80% chance (made up number) of them at least kicking a FG hurt your net ATR more than punting away at that point?

I'm a policy guy, so could someone with a better math/statistics background comment on this difference and why using mean vs. median here is correct?

I guess I'm still with Pat here (and I think Jason's subsequent posts) that say the ideal is to only punt when the net point gain of moving the field position outweighs your chance of converting and eventually scoring. Consequently, we should probably be seeing a lot more teams going for it on 4th & 3 (all over the field) than the current mindset (always punt for 4th & 2plus).

And I haven't really seen any response or comments back on the effect of subsequent possessions not being independent variables, and the "no punt" team's continually worsening field position problem, even when the "no punt" does manage to stop the other team (average drive length being 27 yards, vs. net punt being ~35 yards).

158
by DGL (not verified) :: Sun, 08/20/2006 - 12:50pm

Taking the "average punting position" from comment 156, the 36 yard line, we can do some math. (My numbers are from the 2005 season rather than the 2004 season, but I don't expect a significant difference).

If you go for it from your own 36 yard line and make it, you're effectively starting a new drive around your own 36 yard line (exactly where depends on the yards to gain).

If you go for it from your own 36 yard line and fail, you're allowing your opponent to start a new drive around your 36 yard line (exactly where depends on what you gain or lose on fourth down).

If you punt, you're allowing your opponent to start a new drive around his own 27 yard line (average net on punts in 2005 was 37 yards).

Starting a new drive around your own 36 yard line is gives you 1.70 expected points. (Call it 1.7 +.0.2/-0.0 depending on exact yards gained.)

If your opponent starts a new drive around your 36 yard line, he gets 2.63 expected points (call it 2.63 +0.5/-0.25 depending on exact yardage on fourth down).

If you punt and your opponent starts a drive at around his own 27 yard line, he gets about 1.44 expected points.

You should punt if punting gives a higher expected points (opponent's points negative, your points positive) than going for it. Using the baseline values, the inequality is:

Punt if -1.44 > 1.70p + (-2.63)*(1-p).

p = probability of gaining the first down; 1-p = probability of failing to gain the first down. Sorry; I can't draw the probability tree in ascii art.

Doing the math, in this case you should punt if your probabilty of picking up the first down is less than about 27%. Using the most pessimistic probabilities for going for it, the threshold probability works out to 35%.

I don't know what yards to go equates to a success probability of 27%-35%. According to the 2005 OL and DL stats, no OL had a Power rating (

159
by INT (not verified) :: Sun, 08/20/2006 - 7:01pm

Forgive me if someone made this point and I missed it, but it seems that the premise of this article and the Never Punting thing is that all teams in NFL would never punt. Like Pat said, if a team never punts, a converse strategy would be to punt, as you could pin them back with better success.

160
by Alex A. (not verified) :: Sun, 08/20/2006 - 8:18pm

Hey Jason,

I really enjoyed reading your article. The math is a bit rough around the edges, but I agree, at least in principle, with most of what you say. However, as many people have commented, saying that a safety is not a turnover just doesn't make sense. I know you're trying to stay true to the definition, but wasn't the whole point of the article to redefine turnovers so that the definition made more sense?

Anyway, I have an idea about how to define a turnover so that a safety is included, and I think my definition will show that a safety is in fact 2 turnovers.

Here goes:

Call the teams Team A and Team B. Require, by definition, the following three things:

1) Drives must alternate between the two teams, that is, whenever Team A's drive ends, Team B's drive begins.

2) Team A's drive can only end with a turnover, or with a score in Team B's end zone (Field goals count because they go through Team B's end zone on the way to the goalposts). Likewise, Team B's drive can only end with a turnover, or with a score in Team A's end zone.

3) Team A can only score when it has a drive that ends in Team B's end zone (Again, field goals count because they go through Team B's end zone). Likewise, Team B can only score when it has a drive that ends in Team A's end zone.

Now for an example:

Team A's quarterback is sacked in his own end zone for a safety. This is the first turnover, because the drive ends, and there is no score in Team B's end zone. However, Team B has just scored in Team A's end zone, so Team B must have had a drive, even though no player from Team B touched the ball. Team B's drive was instantaneous, but it was still a drive. So, after the play is over, Team A has had one turnover, and Team B has had one scoring drive, so it is now Team A's drive. But since Team A is required to kick the ball off to Team B, Team A's drive won't end in a score in Team B's end zone. This is the second turnover.

I hope this helps. Thanks for a good article.

161
by DGL (not verified) :: Sun, 08/20/2006 - 11:06pm

Whoops.

No OL had a Power rate of less than 44% IIRC and no DL had a Power rate of GT 46%. So clearly going for it with less than 2 yards to go is indicated.

You can do the same math for any point on the field, and the threshold probability is in the 20%-40% range for every point from the opponents' 35 (field goal range) to around your own 20. Inside your own 20 it goes up to where it's around 50%-55% inside your own 10. So basically, as long as your outside your own 20, you'll do better over the long term going for it with 2 yards to go or less.

It would be interesting to see how quickly the success probabilities drop with yardage to go. With a table of success percentages for various yards to go, one could come up with a decision chart for fourth down based on yards to go and field position.

162
by Steve (not verified) :: Mon, 08/21/2006 - 2:54am

The Hidden NFL had a stat (if I recall correctly, it has been years) that basically determined, based on situation (ball spot, down, dist, score etc,), a given team's likelihood of winning. They applied it to any point in any game and from there they extrapolated the "value" of any given play result.

It seems to me that a perhaps more proper approach to this idea would be to create some sort of chart (call it an odds chart, crafted similarly to the "go for 2" charts NFL coaches currently use) that would determine - given situation - when going for it on 4th would directly increase the probability of winning (based on possible play outcomes) rather than diminishing it. Said chart could also give a degree of increase. I think doing something like this would INCREASE attempts to go for it on 4th down, but wouldn't require radical approaches that seem counter-intuative because they have grevious flaws that bear out in simplified math, but not if you really pick it apart.

The 1.82 figure is great, but there are only 16 games - an incredibly volatile sample size. Even if the figures bore out the distribution might fall poorly (you get huge upswings in blowouts and small downswings in close games); and in the NFL jobs are over before small sample sizes get a chance to grow for normal distribution. Reduced possessions, field position distasters, team talent variations (for example, would Green Bay have the same chances of increased success as the Colts? No. And so the base theory doesn't hold on anything but a blanket level) - all of these combine to make this theory cool, but crazy if someone took it as-is.

I've felt for a long time that re-evaluating the NFL's 4th down approach is long overdue. Kudos to you Jason, for taking that first whack. I think there is something there to be truly mined. Bravo for thinking way outside the box.

163
by Steve (not verified) :: Mon, 08/21/2006 - 2:55am

Oops DGL basically said what I said better - with cool math added. Sorry for the re-post of the same idea in normal-person speak. :)

164
by Jason S (not verified) :: Mon, 08/21/2006 - 2:58am

Love the article, this is why I read FO. For different takes on the game we love. I think the perfect set up for this type of idea would be pre-season. Wins and losses don't matter and a coach could really get a handle on adjusting the "no-punt" idea within game situations. And you know the fans would love it, well mostly, anyway.

165
by Pat (not verified) :: Mon, 08/21/2006 - 12:05pm

(Lowest power rate is 41%, Arizona.)

No OL had a Power rate of less than 44% IIRC and no DL had a Power rate of GT 46%. So clearly going for it with less than 2 yards to go is indicated.

Not necessarily. First, those numbers aren't adjusted for opponent, so the outliers could actually be worse. But still, the mean power success rate was 64%. Arizona's was 41%, or roughly 2/3 of the mean. Carolina's defensive was 46%, or again, roughly 2/3 of the mean (very roughly). Carolina versus Arizona might actually be a ~25% success rate for Arizona's power rushing.

It seems to me that a perhaps more proper approach to this idea would be to create some sort of chart (call it an odds chart, crafted similarly to the “go for 2″ charts NFL coaches currently use) that would determine - given situation - when going for it on 4th would directly increase the probability of winning (based on possible play outcomes) rather than diminishing it.

The big problem with something like this is that the values that the chart would be based on would change a lot depending on the offense and defense you're facing. A great offense/bad defense team facing another great offense/bad defense team should probably rarely, if ever, punt: Indianapolis, facing Cincinnati, was averaging 45 yards per drive, and that's truncated because of the length of the field. At that point, punting is basically pointless.

That being said, though, they did go for it on 4th down twice in that game - and it was 4th and short on both. They failed, both times. I've got a feeling that the conversion rate on 4th down would be a lot lower than Jason thinks it would be, especially against teams with a decent defense.

166
by Bruce (not verified) :: Mon, 08/21/2006 - 3:52pm

I like the thinking here. I am not sure of the mathmatics either but I will say that a happy medium could probably be reached. Why not just eliminate the punt past the fifty? It would be interesting to see what that would produce in the long haul. I just think that statistical happenstance would tend to play really heavy on the mindset of a team that failed to convert on their first three or four attempts on fourth down from within their own 35... just my opinion.

167
by Alex A. (not verified) :: Mon, 08/21/2006 - 10:09pm

"I just think that statistical happenstance would tend to play really heavy on the mindset of a team that failed to convert on their first three or four attempts on fourth down from within their own 35… just my opinion."

Funny you should say that, because that seems to be one of the most common arguments against going for it on 4th down, that the momentum swing that would accompany a failed 4th down would be a factor. However, an analysis by David Romer, in his paper on 4th down strategy, found no evidence of any effect from momentum swings after failed 4th downs.

168
by Jason Scheib (not verified) :: Mon, 08/21/2006 - 10:10pm

re: #160 - I suppose we've already seen that depending on what definition you use you could get 3 different answers for the number of turnovers in a safety. But it's been shown to me here that the definition should make the most sense of what is actually happening. I already conceeded in #124 that a safety is 1 turnover, and that seems to be by far the consensus among the readers here. It makes sense considering what is happening on the field, your drive ends without a score (for you) and you give the ball to the other team. I understand what you're trying to say but it sounds unnecessarily complicated to fit a definition, similar to what I was doing earlier in being unnecessarily complicated in trying to fit a definition. Thanks though.

169
by Jason Scheib (not verified) :: Mon, 08/21/2006 - 10:20pm

Re: #165 - Maybe the 4th down conversion rate would be a lot lower than I figured, I just tried the best I could to find it. I'm all for finding out what it really would be if someone figured it out better. That said, I even threw the current 4th down conversion rate in and still got a positive ATR (the last paragraph in the 'Connecting the Dots' section), the idea there being that even if the rate I used was too high, it should at least be higher than the current 4th down rate. I suppose that should still be the case, I'm not so sure it makes sense that it could be less.

170
by Jason Scheib (not verified) :: Mon, 08/21/2006 - 10:49pm

Re: #164 - something I had thought of but never mentioned is that if a team never punted, there's no way to know by watching the game or even looking at the game stats if a team did better than they would have had they punted as normal. The only way to know for sure is to calculate it based on theory like the one here or some adjusted version. Say for instance that before the season starts a team is expected to win 10 games. If they never punt and only win 8 games, what does that mean? Does it mean the strategy failed? Not at all. It's just as likely that the expectation was wrong and that they may have only won 6 games had they punted. It's completely subjective then, and you wouldn't necessarily know which of the two is correct. So while it makes sense for teams to give it a try in preseason because the games don't count in the standings, it wouldn't really tell you much. In other words, you can't objectively 'prove' or 'disprove' whether the strategy works by trying it out because you are comparing real outcomes with outcomes that never happened in the real world. You can only find out by using whatever correct logic and math to calculate it, which means the theory has to be correct first.

171
by Jeff Feagles (not verified) :: Tue, 08/22/2006 - 6:10am

this article sucked

172
by Alex A. (not verified) :: Tue, 08/22/2006 - 9:08pm

Re: #168 - All right, I'm willing to let it go, since you did at least agree that a safety is a turnover, and since it really won't affect much because safeties are so rare. And I will give you this: the safety itself is only one turnover.

But I still think that having to give the ball back to the other team immediately after they scored should count as another turnover. So, unless you recover an onside kick after giving up a safety, that kickoff should probably be considered another turnover.

Re: #112 - "It’s also different because while in a return the turnover and score happen on the same play they are not the same event in the play which they would be if a safety was a turnover, and if they were the same event there’s no place for a drive change inbetween like there is on a return."

I respectfully disagree. What about this situation: the defense recovers a fumble while it is still in the opponent's end zone. It's a touchdown, and a turnover (even by the old definition). The team that recovers the fumble had a drive, they scored on that drive, and the drive happened at the exact same time as the turnover, and the drive is the same event as the turnover. So it is possible for a drive and a turnover to happen at exactly the same time, and be the same event. I know it sounds really wierd, but then again, it sounds a little wierd at first to say that the receiving team on an onside kick had a drive when they never even touched the ball. By the way, I don't get why you are so resistant to the idea that a team has a drive when they get a safety, even when they don't touch the ball. After all, you showed in the article that a team can have a drive without touching the ball when the other team recovers an onside kick.

Now consider this situation, and tell me how it is fundamentally different from a safety: the defense intercepts the ball while it is still in the other team's end zone. This is definitely a turnover, by anyone's definition. Also, the defensive team scored a touchdown, at the exact same time as the turnover. They get some points, and then they have to kickoff. They attempt an onside kick, and recover it. That onside kick , as you state in the article, counts as a turnover against the receiving team (in this case, the one that threw the interception). That makes it two turnovers for the team that threw the interception. After the onside kick, the resulting situation is basically the same as that after a safety, except that the number of points scored, and probably the field position, are somewhat different.

"It is implied in the nature of the game of football that when you score the next drive belongs to the other team."

That is why the kickoff after a safety should be a turnover: because the other team scored, so the next drive belongs to you. And unless you successfully recover an onside kick, your drive ends without a score.

Sorry, I know I said I'd let it go, it's just that this has been bothering me. Anyway, I'm glad that we can all agree that a safety is a turnover.

173
by Sean Landeta (not verified) :: Thu, 08/24/2006 - 5:11am

Re #171: I agree. Do not listen to this man, he is clearly deranged.

174
by Jeff Feagles (not verified) :: Thu, 08/24/2006 - 11:58am

Fuck you Landetta, FUck you and your mom

175
by David (not verified) :: Thu, 08/24/2006 - 11:59am

This article was a joke right.

Is their someone who actually believes never punting would work because it worked on madden?

Holy fuck this site has become a joke

176
by DGL (not verified) :: Thu, 08/24/2006 - 7:58pm

I think what David meant in post #175 is...

Never punting is clearly rated too high because someone said it works in Madden. Playing football the way it's always been played since the days of Knute Rockne is way better than this. Holy fuck this site has become a joke

177
by Sean Landeta (not verified) :: Fri, 08/25/2006 - 12:27am

Hey, I was agreeing with you, Feagle-matter! I meant that the guy who said "never punt" was crazy!

178
by PackMan (not verified) :: Mon, 08/28/2006 - 12:19pm

I think the mindset is the biggest factor, and that your risk could be reduced if you use this 4 down "mindset" once you get outside the 30 yd line, if you start on the 1, you don't want to go for it from the 1, but from the 35, it is much more reasonable, and I generally play this way on madden. Also, if you are winning, going for it on 4th down may be counter productive, depending of course, on how much you are leading by, at what point in the game you are.

179
by Gray (not verified) :: Tue, 08/29/2006 - 6:03pm

It would be incredibly simple to simulate a few thousand games between teams that punt and teams that don't. I would be stunned (and suspect that the simulations were flawed) if the "never punt" teams came out ahead. A 50 yard punt is a turnover, but puts the opponent in much worse position than a fumble at the line of scrimmage.

180
by Gray (not verified) :: Tue, 08/29/2006 - 6:05pm

Just to clarify, by "simulate" I mean in Excel or something similar.

181
by SWMike (not verified) :: Sun, 09/24/2006 - 5:51pm

I didn't have time to read all the responses, so someone may have mentioned this earlier, but this "Never Punt" strategy is often used by successful youth football teams. Of course, the main reason they use this strategy is that youth punters have MUCH weaker legs than NFL punters and long snaps on the youth level are a great adventure. Add to that poor punt coverage and it is easy to see why many good youth coaches rarely punt.
Nevertheless, Never Punting is a very successful strategy in youth football for much the same reasons outlined in the article. It is much easier to 10 yards with 4 attempts than only 3 attempts. And defensing third down is much more difficult with the extra down.
I personally love the strategy although I would not take it to the extreme that the author does. As a youth coach I always for it on 4th-and-4-yard or less. I think this strategy would work all the way up to the NFL level, but it seems to me not too smart to always go for it when your chances of successful decline. The statistic that matters is the percentage change of converting a 4th Down. Inside 4 yards the odds favor the offense, while outside it the odds favor the defense.
One advantage of often (as opposed to always) going for it on 4th Down is that the defense is in a serious quandary as to whether to play its regular defense or its punt return team. Dropping a punt returner back effectively changes a 11-man defense into a 10-man defense. This makes converting the 4th Down even easier.
For this reason, I have my youth football teams run what I call a "Punt Option" on 4th-and-4 and 4th-and-5 out of our regular shotgun formation. The QB/Punter takes a peak before the snap to see whether the defense has lined up with a punt returner deep. If not, he punts. If so, he audibles to a regular running play. With the extreme specialization of the QB and Punter on the NFL level, this would probably not work up at that level, but it does show how not automatically punting on 4th down dramatically increases the strategic difficulties for the defense.
Of course, one could still use Punt Option with a traditional "Always Punt" strategy, but as the article notes his strategy increases the chances that 4th Down will be short yardage instead of long yardage.
Overall, I think the Never Punt strategy is too extreme, but it is much better than the Almost Always Punt strategy routinely used in NFL, college and High School football. My vote, however, is call plays on 1st, 2nd, and 3rd Down as if you have four downs to convert, but then punt if it is 4th-and-4 or more. While this strategy is not so contentious as the Never Punt, it would be a radical revolution compared to the present practice
Thanks for the great article.

182
by SWMike (not verified) :: Sun, 09/24/2006 - 6:34pm

Addendum to previous comment.

I should have started by saying, thanks to Jason for writing the article. The math needs to be tightened up, but this is an issue that seriously needs to be addressed. Even the best NFL coaches may be making the wrong choice on a regular basis.

183
by RH (not verified) :: Tue, 09/26/2006 - 2:00pm

Well, TMQ finally decided to say something about this in today's article. No mention of it being done here, first, more rigorously, of course. :)

184
by mpowell (not verified) :: Wed, 09/27/2006 - 1:36pm

This is an interesting article, but its not very practical. We are not really interested in finding a good strategy so much as the optimal strategy. And always puntings is obviously wrong.

Early in the article you claim, "Never punting takes away the risk because it allows the averages to work in your favor." This is just wrong. We don't have to just go with the averages. We can do BETTER than average. There is obviously a point at which punting is a good idea. The article concedes this for late game situations. But what about 4th and 30? Obviously, its a much worse situation than 4th and 10. And describing this as risk is wrong- we know when we are in 4th and 30 as opposed to 4th and 10. We can construct a punting strategy that makes optimal decisions about when to punt b/c its clear that you should be punting sometimes.

185
by Carl Weckenmann (not verified) :: Thu, 11/30/2006 - 1:04pm

The idea that because a team punts five times a game now, that going for it on every fourth down would result in only five fourth down tries is flawed. Every drive can have one and only one punt. But a drive could result in numerous fourth downs if conversions are attempted repeatedly. That is the more fourth downs a team converts successfully, the more likely they are to face another one in the same drive. My guess is an average team would need to convert about 2 on an average 70 yard drive. If the chances of conversion are 40%, the compounded chances are 16% for the drive as a whole. 35-40 yards of field position through a punt may be a better tradeoff that increasing your own odds of scoring by this much.

186
by JoshuaPerry (not verified) :: Thu, 09/06/2007 - 11:51am

Do we have to hear this every year?

187
by Nelson (not verified) :: Fri, 09/04/2009 - 3:34pm

A high school team tried this approach, among other unconventional tactics (like onside kicks 75% of the time): http://highschool.rivals.com/content.asp?CID=892888

Interesting read.