Writers of Pro Football Prospectus 2008

30 Nov 2005

Turnovers, Early Deficits Lead to Losses

When I was a kid, I always used to enjoy Dick Vermeil and Brent Musberger on college football broadcasts. (Musberger would occasionally ask Vermeil if he thought he'd ever return to coaching, and Vermeil insisted he wouldn't.) But one thing I didn't like was how Vermeil always talked about how the team that runs more almost always wins the game. I started looking at box scores, and found that he was right, but I also started to notice that teams usually started running more after they had already taken the lead. What I really wanted was a statistic that showed whether teams establish the run because that's the way to win, or whether teams establish the run after they're already ahead to milk the clock.

Much later, Dick Vermeil became a head coach again and I became a writer at Football Outsiders, where we examine such ideas. Turns out, establishing the run isn't the way teams build leads.

So this article tells us that teams usually lose when they give up sacks, lose the time of possession battle or allow a 100-yard rusher, but it doesn't tell us whether those are the things bad teams do, or whether bad teams fall behind a lot and have to pass in a futile effort to catch up, which leads to those things.

Posted by: Michael David Smith on 30 Nov 2005

38 comments, Last at 01 Dec 2005, 5:40pm by Smeghead


by princeton73 (not verified) :: Wed, 11/30/2005 - 2:56pm

We know this because Insider Jeremy Green, the Cleveland Browns' director of pro personnel during 2000-2004, tracked some of the time-honored statistical categories.

how'd THAT computer program work out for you, Jeremy?

by Josh (not verified) :: Wed, 11/30/2005 - 3:01pm

I think the more interesting article here is the one on factors that don't matter as much as commonly believed (see link)

by Tally (not verified) :: Wed, 11/30/2005 - 3:26pm

Correlation does not imply causation.

Some of these so-called factors were the subject of studies by the FO staff, such as 100 yd rushing games correlating with but not causing wins.

The myths are equally pointless. 300 yd passing games not being correlative with victory probably indicates that the team was playing from behind and had to pass more.

All in all, another ignorant article disguised as scientific and erudite.

by Becephalus (not verified) :: Wed, 11/30/2005 - 3:27pm

I also hear that having fewer points when the time runs out leads to more losses. JK MDS.

by S (not verified) :: Wed, 11/30/2005 - 3:35pm

Hey there. Been gone for a while but had to comment once this got posted.

Honestly, I agree with #1, this article is probably an illustration of why Mr. Green is no longer an NFL executive. The analysis lacks the sophistication of the studies that typically apear on this site, and his claims are huge stretches based on the data he uses. Honestly, of his ten claims, the only one that a)hasn't already been demonstrated using better analysis or b)isn't riddled with confounding variables, is myth #5 (the kick returns). The only explanation I have for that one is that kick returns must be somewhat random events (like fumble recoveries). Does anyone know of any data about the above?

However, this is what you get from mainstream sources most of the time. And for fun and discussion purposes, I guess it's ok. But there's nothing new here.

by Drew (not verified) :: Wed, 11/30/2005 - 3:36pm

The "five deadly sins" and "big-time myths" articles are both interesting.

The myths article is probably more interesting, because most of the deadly sins are pretty intuitive. The reasons for them happening may not be, but the fact that they are bad things to do is fairly obvious to even a casual fan.

My only beef with the myths article is with #3. The contention seems to be that being the top seed isn't important. But the data provided suggests that the top seed makes it to the Super Bowl as often as all the other seeds combined. Maybe it's not a total lock, but it still sounds pretty darn important to me.

by Ryan Mc (not verified) :: Wed, 11/30/2005 - 3:37pm

Like other readers I also have a few issues with some of these. The falling behind in the first quarter stat might be cute, but is completely useless since I don't know any coach who actually game plans to fall behind early. In other words, what does this stat actually teach a coach?

I know turnovers is every coaches favourite stat. My problem with it is losing often leads to turnovers as much as turnovers lead to losing. For example, a lot was made of Daunte Culpeppers 5 INTs against Cincy earlier this year, but 4 of them came after his team trailed by 27, and were therefore the result of excessive risk-taking which wouldn't have occurred had his team been close.

Sacks I can buy though. It drives me crazy when I see QBs take a sack on first or second down when they have time to get out the pocket and throw the ball away. So excessive sacks is either a real indication of poor QB play, or poor line play, and I think we can agree that both or either of those are pretty likely to be highly correlated with losing.

by seven year lion (not verified) :: Wed, 11/30/2005 - 3:38pm

The accompanying article claims that being the top playoff seed isn't that important because that team "only" advances to the Superbowl 50% of the time. You're right ESPN, the fact that the other 5 teams combined make it an identical 50% shows that being the #1 seed is completely worthless. Was that a huge myth anyway? I've never heard anyone try and claim that the #1 seed always makes the Super Bowl.

Articles like this are why I always respond to people telling me "You know, people can use statistics to prove anything." By saying "No, idiots can use statistics to believe anything."

by Kevo (not verified) :: Wed, 11/30/2005 - 3:38pm

My only thought about kick returns is that Special Teams are almost exclusive, as opposed to offense and defense, which have an effect on each other (time of possesion, field position, point deficits).

When you score on a kickoff return, it takes about 10 seconds off the clock, and you give the ball back without having dented the opposing defense whatsoever.

by seven year lion (not verified) :: Wed, 11/30/2005 - 3:40pm

Whoops, Drew beat me to it.

by Kevo (not verified) :: Wed, 11/30/2005 - 3:42pm

Well, when you consider that the top seed is generally regarded as the best team, you'd figure it would be more than a 50% chance to win two games and make it to the Super Bowl.

by Drew (not verified) :: Wed, 11/30/2005 - 3:42pm

Re 10

With only a two-minute time gap, I'll give you credit for starting to write your post before I finished mine.

by Bassett (not verified) :: Wed, 11/30/2005 - 3:48pm

Turnovers and trailing scores lead to losses?

Breaking News:

Ice is cold and fire is hot.

by Drew (not verified) :: Wed, 11/30/2005 - 3:48pm

Re 11

I don't know. Basically what you have is a mini-tournament between 4 good teams. To say that the best of the group has a 50% chance of winning it sounds perfectly plausible to me.

by Kibbles (not verified) :: Wed, 11/30/2005 - 3:53pm

Wait, wait, wait... Turnovers and early defecits lead to losses? No way. Give me a second to get my head around this.

by bobman (not verified) :: Wed, 11/30/2005 - 3:57pm

All in all, for an article that is not exactly methodologically rigorous, it was pretty good and might lead some casual fans to look below the surface and eventually to become FO regulars. No? Like any discussion or debate, the more informed the parties are, the better for everybody it is (in my mind).

One thing not noted about the yards per carry stat is this scenario: you're behind by 14 in the 4th quarter. Sure your QB starts firing all over the field for 9 yards a shot against the prevent D, but every once in a while the change of pace run also nets 8-9 yards and the draw nets 15. Get five or six of those, and that really ups your average YPC from 3.8 to 4.8. And you still lose. This fits two of the non-winning patterns (300 yd passer and high YPC).

Maybe a refinement of the article might be to look at these stats on a quarter by quarter basis--the QB with 150 yards in the 1st half wins what percentage of the time more than the QB with 150 yards in the 2nd half (i.e. garbage time futile catch-up)? Eventually you end up looking at them on a play by play basis and... voila! You're at FO.

by Kevo (not verified) :: Wed, 11/30/2005 - 3:57pm

To have only a 50% chance of making the Super Bowl as the top seed, the team would have only about a 71% chance in each playoff game. I'd say, with home-field advantage considered, that most people would expect a higher percentage than that. Of course, it depends on how much better than the competition they are, but with a 7% addition for HFA, 71% seems a bit low to me.

by Michael David Smith :: Wed, 11/30/2005 - 3:58pm

The only one of the five myths and five sins that isn't completely obvious is the kick returns. I'd like to see how the numbers break down between punt returns and kickoff returns. For kickoffs, a bad team receives more kickoffs than a good team and should therefore return more for touchdowns, so that's pretty obvious. With punts it's less clear, although if you receive a lot of punts it means your defense isn't forcing a lot of turnovers.

by bobman (not verified) :: Wed, 11/30/2005 - 4:07pm

Kevo, I don't know about that. Assume two home games for the #1 seed. The first game they have an 80% chance of winning against another division winner. That might be high, really, even at home against another good team, even for the best record in the conf. Then the next game is 63% against what is likely to be the second best team in the conf. That's a pretty good odds spread, 63% vs 37%, between the two top teams in the conf. And 0.8 x 0.63 = 0.50 chance of winning both and making the SB. Not sure it's right, but just throwing numbers up there, they appear reasonable. (of course all the other comments above about the top seed having a 50% chance and the other seeds all splitting their 50% chance still hold.)

by Will Allen (not verified) :: Wed, 11/30/2005 - 4:10pm

Gosh, prior to this piece I was entertaining the counter-intuitive notion that falling behind early was the best way to pursue victory.

by seven year lion (not verified) :: Wed, 11/30/2005 - 4:19pm

One thing is that Kick returns are pretty rare. Most teams aren't getting more than a handful a year anyway so the game that they happen in probably pretty random.

Re 18: I'm not sure creating turnovers would affect punt rate all that much. If you recover on their side of the field, you create a short drive for their offense, and leave the extra time for an additional drive later in the game which might end up in a punt. If you recover after a long drive, you're probably only preventing a touchdown, a field goal try, an attempt at a 4th down conversion, or some other sort of non-punt outcome.

This would break down if a team is forcing a TO on every other drive or something like that, but even the best teams generally only get about 2.5 takeaways a game.

If there's an old piece of FO research I'm unaware of that shows that turnover rate vastly affect punt rate I retract all my statements.

by Andrew (not verified) :: Wed, 11/30/2005 - 4:33pm

seven year lion #8:

"The accompanying article claims that being the top playoff seed isn’t that important because that team “only� advances to the Superbowl 50% of the time. You’re right ESPN, the fact that the other 5 teams combined make it an identical 50% shows that being the #1 seed is completely worthless. Was that a huge myth anyway? I’ve never heard anyone try and claim that the #1 seed always makes the Super Bowl.

Since 1978 (excluding 1982 and 1987) and the 16 game schedule, in 25 Super Bowls:

26 #1 Seeds

16 #2 Seeds

2 #3 Seeds (1979 Rams, 2003 Panthers) have made it to the Super Bowl.

5 #4 Seeds (1980 Raiders, 1992 Bills, 1997 Broncos, 1999 Titans, 2000 Ravens)

1 #5 Seed (1985 Patriots)

No #6 Seeds.

So roughly:

2:1 for Super Bowl as a #1
3:1 for Super Bowl as a #2
25:1 for Super Bowl as a #3
10:1 for Super Bowl as a #4
50:1 for Super Bowl as a #5

More to the point, the winning percentage for having home field advantage in the Conference Championship is only around 50%, or less than what would normally be expected from a home game.

Getting just a first round bye gives a 42% chance of advancing to the Super Bowl.

by hmmm (not verified) :: Wed, 11/30/2005 - 4:47pm

Is it really true that the Bucs have never returned a kick for a score?

Pretty amazing, considering they have probably received more kicks than most teams -since I would think they have given up a lot of touchdowns in their history.

by Ryguy (not verified) :: Wed, 11/30/2005 - 5:14pm

re: this article..

this article is like how we teach our kids in 4th grade about the parts of the Civil War we think they'll understand and then teach them more in middle and high school.

I feel like I should be refunded the time I spent from reading this article.

by princeton73 (not verified) :: Wed, 11/30/2005 - 5:53pm

the one about the 300 yard passing games is particularly silly--to axe us to be SURPRISED, that is

come on, this has been beaten to death for as long as I can remember.

Generally called "garbage time yards"

every Sunday night, ESPN gives its list of 100 yard rushers (mostly "W"s) and 300 yard passers (a lot of "L"s)

by Mike (not verified) :: Wed, 11/30/2005 - 5:56pm

#5: I imagine kick returns for scores don't help teams win as much as you would think for one main reason.

Aside from once a game, every kick return is the result of the other team scoring.

Think about when Dante Hall returned that kick against the Colts in the 2003 playoffs. That play was the result of great blocking, a great return man, but above all else, 6 scores by the Colts.

It's a random occurance that, unlike fumbles, is aided by being bad (namely giving up points).

by Mike (not verified) :: Wed, 11/30/2005 - 6:01pm

#22: The thing about the 42% chance of advancing to the Super Bowl after a first round bye is due to the nature of that situation, the maximum percentage you could acheive is only 50%. After all, if 2 teams get first round byes then one of them will always lose. You can go even further and say "The chances of winning a Super Bowl after getting a first round bye are less than 25%" and you would be correct.

But the fact remains that #1 seeds making it to the Super Bowl 50% of the time is a great advantage. A 50% chance of beating two teams that aren't just playoff caliber but have won at least one, sometimes two games to get there? I'll take that.

by DavidH (not verified) :: Wed, 11/30/2005 - 6:18pm

Getting just a first round bye gives a 42% chance of advancing to the Super Bowl.


84% of Super Bowl teams had first round byes.

by jebmak (not verified) :: Wed, 11/30/2005 - 6:37pm

I learned another similar thing from a Colt's announcer who said,"If you don’t hit Edgerrin James behind the line he always gets plus yards."

by James Gibson (not verified) :: Wed, 11/30/2005 - 7:09pm

What drives me nuts is that people have been forever pointing out that 100 yard rushers come about because a team takes the lead and then sits on it. The Hidden Game pointed this out in '88, Bud Goode's been point out forever that yards/pass is the "killer stat," I wrote an article in 1999 (currently lost from my archives, although the page is linked in my name) noting that passing is more important than rushing. Ron Jaworski even likes to say "points come out of the passing game." Yet announcers still talk about establishing the run. In fact, one myth and one sin go together here - allowing the 100 yard rusher and the yards/rush not being a good indicator. Does that have anything do with the team ahead running, thus accumulating more total yards, and the team behind playing knowing what's coming and therfore driving down the average?

And I didn't realize that a 300 yard passer winning was a myth. It seems like Chris Berman and Tom Jackson on Prime Time are always talking about how 300 yard passes usually lose.

by Falco (not verified) :: Wed, 11/30/2005 - 7:33pm

If you thought "Leading after the First Quarter" was interesting, wait til you see the sequel, "leading at halftime."

People have commented on the correlation vs causation problem, and it is true of virtually all of these myths and truisms.

My pet peeve correlation vs causation (non-football) is:

**Most auto accidents occur within 25 miles of a person's home.

Thus suggesting accidents occur due to some attribute of driver's to be more careless closer to home. In actuality, this stat is meaningless. Of course most accidents within 25 miles of home! Unless you are having someone drop you at your car, parked 25 miles from home, or are an over the road truck driver, or are on the trip to Wally World, most drivers are driving within 25 miles of home a high percentage of time.

On a related note to the "establishing the run" myth, a few weeks ago, I started looking at only the first half stats this year. I looked at 6 first half categories: Rush Att, Rush Yds, Rush Yds/Att, Pass Att, Net Pass Yds, Net Pass Yds/Att (documented sacks counted as pass attempts). Obviously, teams that have both the most attempts and most yards per attempt in a category would have the most raw yards also, but there are many cases where team one has more raw attempts, while the other team had more yds per attempt. I was interested in seeing where the biggest correlation to eventual wins/losses might occur.

Unfortunately, I only got through week 5 game stats, and haven't had time to go further, (nor do I have the specific numbers here) so this is a work in progress. The first thing should be intuitive--being the best in any one of these stats, without knowing any other variables, showed a winning percentage in excess of 50%.

Of the 6 categories, though, there was some separation through 5 weeks of data. First Half Net Pass Yds per Attempt was #1 in terms of win/loss record of team's winning that category. Raw Passing Attempts was sixth. The rushing stats were more closely bunched together in terms of W/L.

by Smeghead (not verified) :: Wed, 11/30/2005 - 8:02pm

Well said, #26. And notice the sleight-of-hand -- here's his claim:

Since the kicking units are involved in their share of plays, special teams must have an impact. When a team returns a punt or kickoff for a touchdown, you would imagine it would tilt the scales dramatically in the typically close games served out by the NFL.

Uh, no.

Then he gives data exclusively from kickoff returns. Returns resulting from punts -- that is, from instances where the defense stopped the offense -- are conspicuous by their absence.


To take his (unnecessarily limited) sample pool of 2005 plays only, there have been five punt returns for TDs. The returning teams in those games are 4-1.


by bravehoptoad (not verified) :: Wed, 11/30/2005 - 8:33pm

Come on, give the guy some credit. I was surprised to find such a high correlation between leading after the first quarter and winning the game. How many of you would have guessed 75%? Be honest.

by Tarrant (not verified) :: Wed, 11/30/2005 - 10:31pm

My data shows that the statistical correlation between leading at the end of the 4th quarter and winning the game is 100%.

Actually, it doesn't surprise me at all. How many drives are there in the average first quarter...three? Four? Often more, depending on the number of passes, but let's say it's an even number. If you're leading, it means you got more from your drives than they did from their same number of drives. That seems to me a good indicator of a stronger team. Not always, but if you had to guess a percentage, you'd probably guess maybe 60-65%, so 75% doesn't seem overly unreasonable.


by Pete (not verified) :: Thu, 12/01/2005 - 12:06pm

I guess I've been reading you guys too much, but I absolutely HATED this article. It's almost a case study in how to abuse statistics.

1. Suppose you inverted the question and asked "What percentage of winning teams led after the first quarter?" Would 75% seem surprising? Not to me. The winning team should lead after 1st 50% in close games, higher than that in blowouts. Similarly, what percentage of teams that win by more than 10 never trailed in the game? I'd guess 80%.

2. "Only 50% of #1 seeds make the Super Bowl." In a coin-flipping contest, the #1 seeds would only make the Super Bowl 25% of the time. So he's saying that #1 seeds are "only" twice as likely as the other seeds to make the final game. What a shocker!

3. My main objection, though, is his constant switching of sample spaces, citing one game from this year and going "Q.E.D." And in a couple of cases he doesn't even do it right! His last sentence is probably an attempt to be glib, but the "and ALMOST lost" infuriated me.

He did touch on something I've always been curious about. Maybe Aaron or one of the Outsiders stats guys could venture a theory. I've noticed, in college ball especially, that the worst teams in a conference are almost always the least penalized. Is it because they're not aggressive enough? Do the refs feel sorry for them?

by BillWallace (not verified) :: Thu, 12/01/2005 - 2:37pm

This article was awful. The only one that wasn't immediately pointless was the special teams, and even that one doesn't hold up to scrutiny.

by Smeghead (not verified) :: Thu, 12/01/2005 - 3:27pm

Re 35

I like that last question. Speculatively, how about because they spend a vastly disproportionate amount of time playing defense against running plays? I'd suspect most defensive penalties would either be offsides, defensive holding or p.i., all of which seem more likely in passing situations.

But there's an FO essay in there somewhere, isn't there, about overall relative frequency of different types of penalties, adjustments to yards/play or similar metrics based on a team's likelihood of incurring penalty yards, the frequency with which penalties are declined and the reasons, likelihood of particular officiating crews calling more or fewer penalties, whether some teams can "draw" opponent penalties more often or whether it's pretty much luck ...

by Smeghead (not verified) :: Thu, 12/01/2005 - 5:40pm

By the way, per the special teams, FO's top five special teams through week 12:

1. HOU (11.4%)
2. BUF (7.9%)
3. NYG (7.2%)
4. TEN (6.1%)
5. MIA (5.1%)

Giants are the only team among them with more than 4 wins.

Maybe it's a product of too much cap space spent on a part of the game less determinative of game outcomes.