Writers of Pro Football Prospectus 2008

23 Jan 2016

NFL Coaches Are Getting Away With Crimes Against Middle-School Math

Benjamin Morris at 538.com points out that in all the madness of Andy Reid's clock mutilation and Aaron Rodgers' Hails of Mary, few noticed that both the Chiefs and Packers kicked extra points when basic math and league-wide averages said they should have gone for two. This is actually a few days old, but hey, Morris used my Tweet, so it's only fair to mention his piece here.

Posted by: Vincent Verhei on 23 Jan 2016

50 comments, Last at 28 Jan 2016, 2:41pm by Pat

Comments

1
by Bright Blue Shorts :: Sat, 01/23/2016 - 8:27am

Will just say I don't think the idea of using league averages to evaluate team's individual decisions is solid.

You need to look firstly at:
- how effective the team has been with its own play that season
- how effective the opponent has been at defending it that season
- how it's gone during the actual game

You wouldn't advise the 76 Bucs offense to go for it on 4th&1 against the 2000 Ravens just because their chances are better based on league averages of all offenses and defenses. You just wouldn't.

2
by White Rose Duelist :: Sat, 01/23/2016 - 8:52am

I wouldn't advise the '76 Bucs to go for it against anybody, since they are all at least 60 years old.

The article does go into the math in more detail, and while there is by necessity some guesswork going on, almost any estimate of probabilities results in Green Bay going for two being the better option.

18
by The Ninjalectual :: Sat, 01/23/2016 - 6:23pm

"I wouldn't advise the '76 Bucs to go for it against anybody, since they are all at least 60 years old."

Perhaps, but their opponents are all 60 as well, so it evens out.

17
by Duff Soviet Union :: Sat, 01/23/2016 - 5:17pm

The problem with this is that the sample sizes are tiny. That's why you use league wide averages.

The thing is that everyone thinks that the averages don't apply to them and that their own chances are actually much worse for whatever reason. "Well yeah, we've converted 90% of our short yardage plays the last three years, but we're 0 for 1 today, so the next one will definitely fail too".

The difference on one play between offenses is tiny. It's only over time that they add up. To use your 76 Bucs vs 2000 Ravens argument, would you rather bet on the 76 Bucs outplaying the 2000 Ravens defense on 1 play as opposed to 1 drive?

39
by bigpoppapump :: Tue, 01/26/2016 - 7:40am

But this makes the mistake he mentions in his article: The same set of players (who you say are handicapped by lack of ability in going for the two) are being trusted to win the OT which comes from only going for one extra point.

so if you say - these players are not equipped to play to league averages used for the math-model - you also have to apply their inadequate play to their chances in OT.

So use of league averages is fine.

and yes you would advise the 76 Bucs to roll the dice to avoid OT. you just would.

46
by Bright Blue Shorts :: Tue, 01/26/2016 - 4:40pm

But when you go to overtime, you no longer just have your offense ... you can also win the game through defense or special teams.

What do you do if you're matching the 2000 Ravens vs 2003 Bucs? Trent Dilfer vs Brad Johnson. I'd be taking the game to overtime and looking to win the field position/turnover battle.

3
by dafrk3in :: Sat, 01/23/2016 - 8:56am

The idea that a coach's only objective is to maximize the probability of winning in situations like this is obnoxious.

4
by PatsFan :: Sat, 01/23/2016 - 9:18am

Why, exactly, in a do or die game, should the coach not be maximizing the probability of winning??

8
by dafrk3in :: Sat, 01/23/2016 - 12:00pm

I said nothing about what coaches should do.

One response to behavior we find confusing is to figure out what lead to the decision. Another response is to call the decision-maker stupid. The latter isn't all that helpful.

9
by Alex51 :: Sat, 01/23/2016 - 12:38pm

One response to behavior we find confusing is to figure out what lead to the decision.

It's not confusing. It's just extremely frustrating. We know why coaches do this: they make conservative choices to avoid blame for bad results and enhance their job security, at the expense of wins. So, it's not so much that they're bad at math, it's that the people they answer to are bad at math.

22
by Jimmy Oz :: Sun, 01/24/2016 - 6:29pm

Alternatively, the coach thinks that their side is better. By taking the PAT, the coach seeks to remove the 50-50 2 point contest as it will disproportionately influence the outcome of the game.

By attempting the 2 point conversion in that situation, the coach admits their chance of winning the football game is, at best, a 50-50 proposition.

6
by billsfan :: Sat, 01/23/2016 - 11:09am

The coach's only objective is to maximize the probability of winning. In all situations.

11
by PaddyPat :: Sat, 01/23/2016 - 1:39pm

You guys are funny. I assumed the obnoxious comment was a very dry joke. How else can you possibly read that? What other objective could a football coach have? Trying to make the play look "pretty"? Trying to impress his dad or his wife? Running a play to feature a certain player to make a contract incentive? I mean--what??

12
by RickD :: Sat, 01/23/2016 - 2:30pm

The coach's primary motivation is to keep his job. Impressing the GM and/or owner are part of that.

Just a few weeks ago, Bill Belichick chose to kick the ball off in OT. He understood that, generally speaking, OT is a 50-50 proposition regardless of who kicks off. And if you have more confidence in your defense than your offense that day, kicking off might be the right choice. (After all, in the previous OT game the Patriots had received, had a 3-and-out, and punted to the Broncos who scored and won.)

The amount of scorn and criticism heaped on Belichick was something to behold. Every armchair pundit "understands" the the only defensible decision there is to receive the first kickoff.

Many NFL coaches follow convention because they don't want to expose themselves to the kind of shitstorm that falls on any coach who bucks it. Belichick has proven himself to the point that he doesn't care what the media says: he knows that his owner has given him the freedom to do whatever he wants to do. Belichick is also the longest-tenured coach in the NFL. A coach in Tampa or Cleveland won't feel anywhere near as secure.

15
by Will Allen :: Sat, 01/23/2016 - 3:18pm

One of the nice aspects of Zimmer as head coach is that age 59 he's decided to not give a damn about the noise. He elected to kickoff against the Rams in ot this year, for the simple reason that his roster and the opponent's roster had better defensive personnel. He was right, all yakking to the contrary.

The other evidence that supports the notion that Zimmer is very unusually secure about himself, and unwilling to play the pr game, is how unafraid he is to admit that he made a mistake, whether it be a defensive call, or in clock management. How often do you see that in any NFL head coach, much less a guy in his 1st or 2nd year? It's pretty obvious why his players buy in the way they do; a coach who is so willing to hold himself publicly accountable is going to have players who are likely to hold themselves accountable.

I'm pretty convinced that NFL head coaching hires are about 80% random, in terms of being successful or unsuccessful, so my view is that Vikings ownership got pretty lucky with Zimmer.

19
by Duff Soviet Union :: Sat, 01/23/2016 - 9:15pm

"One of the nice aspects of Zimmer as head coach is that age 59 he's decided to not give a damn about the noise. He elected to kickoff against the Rams in ot this year, for the simple reason that his roster and the opponent's roster had better defensive personnel. He was right, all yakking to the contrary."

It's hilarious that, for all the crap Bellichick took for his decision, that someone did exactly the same thing and I never even knew about it because it worked. Hindsight bias is so ridiculous sometimes.

43
by bigpoppapump :: Tue, 01/26/2016 - 1:18pm

It's not factually correct that Zimmer chose to kick versus the Rams.

He chose to "defend this end" in the words of his toss wining captain. The Rams then chose to receive.

Decision was driven by it being very windy along the field. Zuerlein had tied up the game with a 53 yarder down wind. It's true about the Vikes D being good that day and neither team having much aerial threat. But Zimmer chose an end.

16
by mehllageman56 :: Sat, 01/23/2016 - 4:18pm

While I don't think that was the best decision by Belichick (it isn't a 50-50 proposition irregardless of who kicks off in overtime), it made sense given what had happened that day. You are correct he received a deluge of criticism from people who should have learned to give Belichick the benefit of the doubt by now.
A similar case resulted from the first Jets/Bills game. Bowles decided to go for it on 4th down 3 separates time while trying to come back from a large deficit, with the end result being a 5 point loss, and everyone in the media claiming he should have just taken the field goals every time and that would have resulted in a win. The irony is the NYT 4th down bot liked every attempt but the last one, when they were down 5 in the red zone with 3 minutes left, which makes no sense to me. http://nyt4thdownbot.com/team/jets

It isn't just the coaches that are clueless when it comes to strategy, it's the fans and the owners as well. In order to become a Riverboat Ron, you need an owner that will think independently about strategy and accept decisions like Belichick's and Rivera's.

48
by RickD :: Wed, 01/27/2016 - 3:42pm

Under the previous system, the first team getting the ball won 60% of the time. It's now down to roughly 53%, but that's considering all teams. If you feel more confident in your defense and less confident in your offense, I could see viewing it as 50-50 (or lower) if you took the kickoff.

50
by Pat :: Thu, 01/28/2016 - 2:41pm

The new system hasn't been around long enough for you to tell the difference between 53% and 50%.

At this point you can "barely" tell the difference between the old system and the new system: it's about a 1-sigma fluctuation. If you hadn't changed the rules at all, about 9% of the time you'd expect to have this few (or fewer) wins from a team that receives the ball.

If the new system is truly fair (50% either way), over 40% of the time you'd expect to get this many (or more) wins from a team that receives the ball.

14
by tunesmith :: Sat, 01/23/2016 - 2:51pm

I'm not sure it's that simple. Aside from cynical motivations like "look good to the GM", you also have variance to think about. Sort of like the lottery - if, due to a jackpot size, the expected value of a lottery ticket were actually above a ticket price after taking into account taxes, multiple winners, etc, you still shouldn't buy any tickets for investment purposes. In gambling theory there's this thing called the Kelly Criterion where if the odds are in your favor, you still limit your bet to some percentage of your bankroll - otherwise you could easily find yourself bankrupt even if you took the proper maximization bet every time. I'm not sure what the "bankroll" would be in a football game though.

25
by Mikey Benny :: Mon, 01/25/2016 - 11:45am

A coach's job is to put their players in the best possible position to win. How is this obnoxious?

26
by bravehoptoad :: Mon, 01/25/2016 - 11:47am

It's funny? I laughed.

5
by James-London :: Sat, 01/23/2016 - 10:42am

I'll bite.
A coach is maximising the possibility of avoiding blame for the loss. A coach that los3s conventionally avoids blame for doing so, while losing because of two-poinr attempts land the coach in the shit.

While that's a shame it's rational from the coach's perspective. McCarthy would have been strung up if he'd gone for two and failed at the end of the game. Andy Reid should have gone for two, but as Vince says, who's talking about it?

Job security > winning where the two don't align. Belichick aside, what other coaches could avoid a firestorm?

Phil Simms is a Cretin.

7
by Dired :: Sat, 01/23/2016 - 11:21am

No one wants to be the next Chip Kelley who tried to do things wrong.

20
by Alternator :: Sat, 01/23/2016 - 9:48pm

Coughlin a few years ago could have avoided any series firestorm (as in, the media would buzz, but ownership wouldn't bat an eye). Parcells could have gotten away with it when he was on the sideline. I suspect Marvin Lewis could get away with it now, as long as he can make a good case to ownership (who seem to really buy into him.)

Really needs to be a veteran coach, though, with a strong history. If more did it, however, the bar would slowly lower over time as coaches taking risks would be looked upon as, well, risk-takers rather than lunatics, and some owners would appreciate that.

49
by RickD :: Wed, 01/27/2016 - 3:45pm

Who else?
Pete Carroll, Sean Payton, John Fox, J* Harbaugh, Mike Tomlin, maybe Rex Ryan...

Parcells would have done whatever he wanted to do and openly mocked any reporter who dared question him.

21
by mathesond :: Sun, 01/24/2016 - 6:21pm

Has any coach been fired for going for 2?

32
by Bobman :: Mon, 01/25/2016 - 10:08pm

Depends on coaches, age, philosophy, win %, recent superbowl success, and ownership intellect (among other variables).

I bet Arians could get away with it in a heartbeat, Rivera, BB of course, Zimmer (I guess... I didn't know either), Pete Carroll, Payton (think of his SB onside kick in 2009). Jeff Fisher ten years ago, yes. Today, not so much. Young guys, probably not. Mike Tomlin, yes. Guys with iffy records and no SB wins, probably not, guys who have absolutely no chance of pulling it off because of personnel or the opponent (using the 76 Bucs vs 85 Bears analogy) probably not.

I think most of us would love to see more logical aggression, but you need a bunch of things to align for that to happen. Like the opening line of Anna Karenina. Tolstoy knew what he was yakking about. Paraphrased: all happy families alike, all unhappy family unhappy in their own way. Meaning, for everything to be good, everything has to work and mesh well. But throw one wrench into the works, no matter what it is, and it messes things up in various ways. The problem with the NFL is it's tough to align all the things you need for a coach to go for it. SO we're all unhappy with the illogical conservatism on the sidelines. Harrumph.

10
by wiesengrund :: Sat, 01/23/2016 - 1:04pm

I'm really skeptical about the math behind this article.

For one, it splits OT success rate into home and away, but neglects to do so for the 2PT rate. McCarthy was probably thinking about a 50/50 chance if he gets it to OT, which seemed like an upgrade over any 45-47 proposition that a 2PT attempt would entail.

Additionally, McCarthy was probably using school math in trying not to get carried away by the old pitfalls of small sample size and recency bias. Yes, the two biggest plays came on that drive, but outside of those plays Rodgers was I think 22 of 41 for 161 a gruesome 3.9 YPA. Faulting McCarthy for underestimating Rodgers seems backwards to me. I think McCarthy pretty reasonably used school math to guess Rodgers success rate closer to 40% (taking into account his whole day), and was taking the coin flip instead.

Also, GBs Def was not bad at forcing turnovers up until that point. 2 INTs and a Forced Fumble, plus a good 3rd down performance, seemed to give the Packers a good enough chance even if they should lose the toss. McCarthy gambled on the better unit on that day: The Defense.

And the Defense lost that game, not McCarthy.

13
by RickD :: Sat, 01/23/2016 - 2:41pm

Yes, leaving Fitzgerald uncovered was a bigger mistake than kicking the extra point.

24
by justanothersteve :: Sun, 01/24/2016 - 7:25pm

I agree it was a mistake. But it took Carson Palmer channeling his inner Russell Wilson to elude the pass rush and pass across the field. Nobody gives Palmer any credit for his effort on that play. If Palmer is sacked, nobody notices the breakdown in coverage.

23
by Karl Cuba :: Sun, 01/24/2016 - 7:21pm

Talking about high school maths:

The Pats listed 16 players as questionable, meaning by NFL rules they had a 50:50 chance to play.

They were all active.

The odds of 16 players with a 50% chance to play all being active is 1 in 65,536. If you assume they are playing 18 games a season, because they play a lot of playoff games, then this should happen once every 3,640 years.

In other words the Pats are taking the piss.

33
by Bobman :: Mon, 01/25/2016 - 10:10pm

I wondered about the same thing, Karl Cuba, and found that list very dubious. I also love that expression.

42
by Hoodie_Sleeves :: Tue, 01/26/2016 - 1:12pm

The problem here is that your starting assumptions are wrong, particularly this one:

"The Pats listed 16 players as questionable, meaning by NFL rules they had a 50:50 chance to play."

Here's the NFL injury report rules:
http://uaasnfl.blob.core.windows.net/live/1818/2015-injury-report-policy...

Pay particular to the description of "Probable" - it says that the player is a virtual certainty to be available for 'normal duty'- and any players that don't, must be explained:

"If a player listed as “probable” does not play, it raises serious questions about the integrity of the injury report. If this situation occurs, clubs will be required to submit a written explanation to the league office within 48 hours."

There's nothing between "virtually certain" and 50/50 - if a player normally plays 50 plays a game, and you think he'll be ready to play 35, he gets listed as questionable.

27
by Pat :: Mon, 01/25/2016 - 3:30pm

I really, really disagree with this.

The article really should be called "Guy Criticizes NFL Coaches Using Middle School Math," because that's what's really going on. It suddenly doesn't sound so impressive.

The problem here is that no matter what you do, there's no real way to figure out what the underlying probabilities are. Saying "teams going for 2 succeed ~50% of the time, therefore the probability of a successful 2 point conversion is 50%" is nuts, because teams choose to go for 2.

If a coach doesn't think he has a play that will succeed 50% of the time or more, the coach doesn't go for 2.

Coaches tend to be conservative in the NFL because the NFL is all about small advantages. You see a mismatch that you like, and you try for it. Sometimes it works, sometimes it doesn't. But you don't gamble the whole damn game unless you have to.

It's the same reason why rushing persists in the NFL, even though statistically it looks terrible. It's a low-risk play that gains you information, and changes the way you'll call future plays.

Extending the game is key in the NFL, because it gives you more opportunities to look for a win. In the end, it looks like a coin flip, but it's not the play that's the coin flip, it's who can find the better play faster.

28
by Duff Soviet Union :: Mon, 01/25/2016 - 6:16pm

"The problem here is that no matter what you do, there's no real way to figure out what the underlying probabilities are. Saying "teams going for 2 succeed ~50% of the time, therefore the probability of a successful 2 point conversion is 50%" is nuts, because teams choose to go for 2."

But most teams don't "choose" to go for 2. Most teams only go for 2 when they have to. As the article points out, the teams that go for 2 the most tend to be bad teams and bad offenses because they're the teams that have to go for two the most due to being behind and needing to catch up.

29
by Jerry :: Mon, 01/25/2016 - 7:05pm

The Steelers set a record for 2-point conversions this year, and they were neither a bad team nor a bad offense. Apparently, Tomlin was going to go for two if a touchdown brought them within a point in Denver.

But that's Pat's point - Pittsburgh went for two a lot because they were good at it. If Mike McCarthy couldn't feel confident about a play there, he's better off kicking and keeping the game going. And, of course, coaches who know their teams well, and have spent a week doing film study of their opponents, have a lot of information we don't.

37
by Pat :: Tue, 01/26/2016 - 1:15am

The other thing is that really, a 2 point conversion is a crappy play to hang an entire football game on. It's totally forced. The defense has no worries, they're compressed into a tight area, and they only have one thing to worry about. It's just not the same as the rest of the field.

Suppose a 2-point conversion wasn't actually a conversion. Suppose it was just a flat coin flip. You walk out, and flip a coin, and if you win, you get 2 points. If you lose, you get nothing.

You think coaches would choose to go out and flip a coin to win the game? No, of course not. They came to play football.

Personally, I think that's a good part of the reason coaches don't go for 2 to win the game. It's just not football to them. I think it's football to the *fans* - it's one play, it's exciting, wooo! But I don't think it's football to the coaches.

30
by The Powers That Be :: Mon, 01/25/2016 - 7:11pm

The only team that "chooses" to go for 2, to my knowledge, is the Steelers. All other teams go for 2 only when they feel they have to.

The Steelers went for it far more than anyone this year (11 times to the Packers 6), and had a higher success rate than anyone else (73%). The rest of the league had a 44.5% success rate.

35
by Pat :: Mon, 01/25/2016 - 11:08pm

You think the Steelers would be going for 2 all the time if, in practice, they were only beating the defense 10% of the time?

31
by Alex51 :: Mon, 01/25/2016 - 7:57pm

The problem here is that no matter what you do, there's no real way to figure out what the underlying probabilities are. Saying "teams going for 2 succeed ~50% of the time, therefore the probability of a successful 2 point conversion is 50%" is nuts, because teams choose to go for 2.

If a coach doesn't think he has a play that will succeed 50% of the time or more, the coach doesn't go for 2.

I don’t think that’s true. In my (admittedly limited) experience, I don’t recall many NFL coaches going for 2 unless kicking would do them no good (except, apparently, Mike Tomlin).

That sort of situation might not cover all 2 pt conversions in the NFL, but it probably covers the vast majority of them (feel free to correct me if I’m wrong). I suspect that most of the rest are botched snaps/holds on kick attempts, like Tony Romo vs. the Seahawks in the playoffs. Which, btw, means the 50% figure is probably a significant underestimate of the chances of success, because it includes a bunch of failed kicks that became 2 pt attempts that the offense was not preparing to attempt.

Either way, if you just look at the times that NFL teams went for 2 when the score/time situation forced them to, you should get a good estimate of the underlying probability of success. Now, it probably varies somewhat from one team to another, but that doesn’t explain why all NFL coaches are so conservative in that situation. Surely some of them have teams that are considerably more likely to succeed on a 2 pt attempt than the average team. But among NFL coaches, Mike Tomlin is apparently the only one who goes for 2 very often when not forced to.

It's the same reason why rushing persists in the NFL, even though statistically it looks terrible. It's a low-risk play that gains you information, and changes the way you'll call future plays.

That’s a bit of an oversimplification. While rushing plays are valuable for the low-risk information gain, that’s not the only reason they persist in the NFL. Even if they gave you no information at all, they would still persist, because rushing plays are the best way to convert in short-yardage situations, where you need a play that will reliably get you at least a short gain, with a low risk of a turnover. They're also the most effective way to preserve a lead at the end of a game, because they keep the clock running and minimize risk of a turnover. Not to mention setting up play action passes.

Extending the game is key in the NFL, because it gives you more opportunities to look for a win. In the end, it looks like a coin flip, but it's not the play that's the coin flip, it's who can find the better play faster.

This doesn’t necessarily suggest caution. Extending the game gives you more opportunities to find a strategic advantage, sure. But it also gives the other team more opportunities to do that. And if the other coach is a better in-game strategist than you are, then all else being equal, you should try to shorten the game, to diminish their advantage. Otherwise, you’re just giving them more opportunities to outpace you. Now, Andy Reid is a great coach, but I’m not sure he’s as good at in-game adjustments as Bill Belichick. And unless he’s better than Belichick, there’s no good reason to prefer overtime to a 50/50 shot at a 2 pt conversion.

You see a mismatch that you like, and you try for it. Sometimes it works, sometimes it doesn't. But you don't gamble the whole damn game unless you have to.

Why not? There’s no good reason to prefer a bunch of tiny risks spread over several plays to one big risk on a single play, unless the overall likelihood of success is higher with the risk spread out. And that’s not always the case.

For instance, say you line up to kick, but the other team is called for encroachment a few times. Now, the ball is on the half yard line. Do you go for 2? I certainly would, riverboat gambler that I am. And I suspect many NFL coaches would even go for it then. And really, why not? It doesn’t take a strategic genius to figure out the playcall. You put your best blockers on the line, and then run it up the middle. Even if the defense knows exactly what’s coming, they’re not likely to stop you from getting half a yard.

Also, getting back to the original scenario, you’re gambling the whole damn game either way. Either you’re gambling that you’ll convert the 2 pt attempt, or you’re gambling that, in overtime, you’ll (a) have the opportunity to find a mismatch on offense before the other team scores a TD, and (b) that you will successfully exploit that mismatch to win the game – neither of which can be estimated with any more certainty than the result of a 2 pt attempt.

36
by Pat :: Tue, 01/26/2016 - 1:04am

I don’t think that’s true. In my (admittedly limited) experience, I don’t recall many NFL coaches going for 2 unless kicking would do them no good (except, apparently, Mike Tomlin).

That's the point. They have a choice to go for 2. That automatically biases the end result. Even if you think "oh, well, I'll just look at cases where it doesn't make any sense for the team to kick the point-after", those still aren't representative cases, because now you're looking at specific corner cases. Not the generic "go for 2" percentage.

Heck, this *specific* case isn't even a generic "go for 2". It's "go for win." Which means the defense knows this play is everything. They don't have to worry about saving something for the next play. Or even overtime. They don't have to worry about anything except stopping this play.

So now instead of looking at all 2 point conversions, in this case you'd have to look at all cases where teams went for 2 instead of going to overtime. Except, oh wait... you don't really have any of those. Because most people would kick the point-after.

While rushing plays are valuable for the low-risk information gain, that’s not the only reason they persist in the NFL.

Yeah 'persist' wasn't the right word. I meant it's the reason why they're so prevalent. And setting up play-action is exactly what I was saying. You're finding out how a defense will react to a given formation and movement.

Really, the entire existence of play-action does fundamentally just screw over most football analytics entirely. Play-action exists because plays aren't independent events - their success is contingent upon previous plays. Which is why all the attempts to analyze a decision as a simple "X% this happens, Y% this happens" are super-naive.

Extending the game gives you more opportunities to find a strategic advantage, sure. But it also gives the other team more opportunities to do that.

Sure. It's a game. There's never any magic to win.

Why not? There’s no good reason to prefer a bunch of tiny risks spread over several plays to one big risk on a single play,

Ah ha! That's the mistake. Yes, there is a good reason!

It's the same reason why poker players wouldn't risk things big on the first hand with a bunch of new players at their table. Waiting gives you more information. More information lets you make a better decision as to when to gamble.

Going for a 2-point conversion to win it all is a very forced circumstance. Unless you've practiced it a ton, and are confident in it, and recognize what the defense is going to do, it's not worth it. Offenses don't have as much advantage when going for 2: it's just one play, and it's a "succeed/fail" try. You can't force defenses to choose between "not so bad" and "awful" options, because, well, there's only one option, anyway. You can't probe the defense with a 'safe' play, trying to find a weakness for a knockout.

I have *never* been surprised by people kicking the PAT to go to overtime. Going for 2 always seemed like a coin flip. It never seemed like football. I mean, I know it's bizarre, as most people look at the "one play to win it all!" as being awesome. They're exciting, sure. But as a coach I would *never* choose them.

44
by Alex51 :: Tue, 01/26/2016 - 2:56pm

Even if you think "oh, well, I'll just look at cases where it doesn't make any sense for the team to kick the point-after", those still aren't representative cases, because now you're looking at specific corner cases. Not the generic "go for 2" percentage.

I’m not sure what the phrase “specific corner cases” refers to. Are you talking about corners of the field or something?

Heck, this *specific* case isn't even a generic "go for 2". It's "go for win." Which means the defense knows this play is everything. They don't have to worry about saving something for the next play. Or even overtime. They don't have to worry about anything except stopping this play.

Okay, but that cuts both ways. The offense also knows that this play is everything, that they don’t have to save anything for the next play, etc. All they have to worry about is succeeding on this one play. So why should the fact that the stakes are higher for both teams benefit the defense more than it benefits the offense?

“Extending the game gives you more opportunities to find a strategic advantage, sure. But it also gives the other team more opportunities to do that.”

Sure. It's a game. There's never any magic to win.

Right, but my point is, if extending the game gives both teams more opportunities to gain the upper hand, why would extending the game help your team any more than it helps theirs? Why would you expect to have a better chance of winning than the historical average of similar teams going to overtime, given that you can’t possibly know which team will gain the upper hand first?

I get that overtime is not a neat, tidy statistical exercise. Things are more messy. But when you’re making the decision about whether to go to overtime, you don’t know what kind of messy they’re going to be. So you might as well treat overtime as if it were a neat statistical exercise, for the purposes of deciding whether to go to overtime, since you have no way to guess the outcome with any more accuracy than simply looking at the historical average.

“Why not? There’s no good reason to prefer a bunch of tiny risks spread over several plays to one big risk on a single play,”

Ah ha! That's the mistake. Yes, there is a good reason!

It's the same reason why poker players wouldn't risk things big on the first hand with a bunch of new players at their table. Waiting gives you more information.

Sorry, I should have been more careful with how I said that. My point was that, assuming you somehow knew for a fact that your overall probability of success would be the same if you spread out the risks over a bunch of plays as it would be if you took all your risk on a single play, then there would be no good reason to prefer one over the other.

So, if an omniscient superbeing told you that you would have a 50% chance of winning in overtime, and a 50% chance of winning on the 2 pt conversion, then there’s no good reason to prefer overtime over the 2 pt conversion.

Your argument was that (in the absence of magical coaching advice), you should assume that spreading out the risk over several plays is going to increase your overall probability of winning the game, compared with putting all the risk on one play. Here’s the problem with that argument. You’re saying that:

(1) Waiting results in more information
(2) More information is always good
Therefore
(3) You should always wait

(1) is clearly true, unless you take a nap in overtime or something. But (2) is false. More information is only good for you if you are better at using it than the other side. You don’t get more information in isolation (at least not in football) - the other team gets the same information you do. So, if they are able to make better use of it, then more information will work to their advantage, not to yours.

And even if we assume that more information is always good, that doesn’t end the analysis. If the probability of success is high enough, taking a big risk on a single play would be better than spreading the risk out on several plays, even if you were better than the other coach at using the information gained in the latter option.

For instance, let’s say that, instead of kicking from the 15 yard line for an extra point, you could kick from the 25 yard line and get 2 points if you made it (this example would sound less awkward and contrived if they hadn’t moved the extra point kick back from the 2, but just work with me here). Assume you’re down by one point, with no time left, and you just scored a TD. I am quite convinced that just about every coach in the NFL would go for 2 in that situation.

In fact, I bet you would even go for 2 in that situation, assuming your job depended on winning that game. I could be wrong, but I think you’d go for it, even though it would be putting all the risk on a single play, even though it arguably “doesn’t seem like football”, and even though you’d be giving up the chance to gain more information in overtime. The ridiculously high chance of success on a kick from the 25 overwhelms any advantage you could plausibly gain by getting more information in overtime.

Unless you've practiced it a ton, and are confident in it, and recognize what the defense is going to do, it's not worth it.

I know that it’s not perfectly analogous, but from the perspective of the offense, a 2 pt conversion is not that radically different than a garden variety 3rd and short, and I’m pretty sure every offense practices those quite a bit. As for what the defense is going to do? It’s not going to vary a whole lot from one defense to another. The compressed field makes it easier to defend, but it also simplifies the playcalling strategy considerably. I mean, it’s not like they’re going to drop into a Tampa-2 or bring out their dime package.

They’ll single cover any receivers that line up wide of the formation, then leave maybe 1 or 2 other defenders in the middle of the end zone in case a TE runs a route, and send everyone else crashing towards the line of scrimmage as hard and fast as they can as soon as the ball is snapped. Anything else would give you an easy walk in, or an uncovered receiver to throw to. At that point, it’s either who gets more push, or whether one of your TEs gets open fast enough. Sure, it’s a gamble, but it’s just as much of a gamble for the defense, and there’s very little they can do formation/strategy wise to get a big advantage.

You can't force defenses to choose between "not so bad" and "awful" options, because, well, there's only one option, anyway. You can't probe the defense with a 'safe' play, trying to find a weakness for a knockout.

True, but I think the main weakness for any defense on a 2 pt conversion is that you only need to gain 2 yards.

Going for 2 always seemed like a coin flip. It never seemed like football.

Okay, but assuming it gave you a better chance at winning the game, don’t you think you have an obligation to the players to give them the best shot at a win, even if it doesn’t seem like football?

Me, personally, I don’t care if the last, deciding play seems anything like football. If they want to put on midnight green tutus and pirouette into the end zone, then as long as they win, I’m not going to judge.

45
by Pat :: Tue, 01/26/2016 - 4:01pm

I’m not sure what the phrase “specific corner cases” refers to.

A corner case is something that isn't "normal" for some reason. What I mean is that coaches go for 2 typically when there's a slight advantage to adding more than 2 points. Which means that you're in situations like you're down multiple TDs, or up multiple, etc.

So why should the fact that the stakes are higher for both teams benefit the defense more than it benefits the offense?

Because defense always has a numbers advantage. Magnifying that helps them more. Doesn't matter if the WR on the far side is amped up, if he's not in the play it's probably not going to make a difference.

But (2) is false. More information is only good for you if you are better at using it than the other side. You don’t get more information in isolation (at least not in football) - the other team gets the same information you do. So, if they are able to make better use of it, then more information will work to their advantage, not to yours.

It's a game! You don't make decisions with the belief that you're stupider than the other guy. If you start off with that belief, you've already lost.

Okay, but assuming it gave you a better chance at winning the game, don’t you think you have an obligation to the players to give them the best shot at a win, even if it doesn’t seem like football?

If you offered the Broncos a coin flip on Superbowl Sunday at the beginning to win the game, you think they would take it?

It's probably their best chance to win the game. Think they care?

47
by Pat :: Wed, 01/27/2016 - 11:01am

Honestly, this comment here is still driving me nuts:

More information is only good for you if you are better at using it than the other side. You don’t get more information in isolation (at least not in football) - the other team gets the same information you do. So, if they are able to make better use of it, then more information will work to their advantage, not to yours.

I just don't get how you can't see what the problem with this thinking is here. It's a game. You don't approach your decisions thinking "well, the other team will probably out-think me for the rest of the game, so I better gamble the whole damn thing here."

Or, if you do, you're screwed no matter what you do. All the decisions are bad. Going for 2, kicking the PAT, you're going to lose. If you figure you can't out-think the opponent, then you're not going to convert at 40-some-%. It's going to be less than that. So the entire reasoning is shot anyway.

And if you do think that way - that the opponent is better than you and has more information (or processed it better) than you - why would you gamble everything in a situation where the defense has the advantage and knows what you're going to do? Why wouldn't you do something more insane, like kick the PAT, and then onsides-kick to start OT?

Okay, but assuming it gave you a better chance at winning the game

This is really the entire crux of your argument. You have to assume that going for the PAT gives you a better chance of winning.

And read the article. They're claiming that going for it gives you a ~48% chance of winning. And kicking the PAT gives you a ~43% chance of winning.

In other words, if you go for 2, after you've played this kind of game 20 times, you'll win one more game, on average.

This is not a serious advantage, and the numbers backing it up have huge error. The idea that it's clearly the better thing to do is just insane. Even a slight change in those 2-point odds tilts things in favor of overtime.

And remember, we've already said that going for 2 makes sense when you don't think you can use the extra information better than your opponent. So you think you're the idiot. And if you think you're the idiot, what makes you think you'll get the 2-point conversion at league-average? Again, in this scenario, you're the idiot coach - a 5% shift down makes more sense to go for overtime.

38
by tuluse :: Tue, 01/26/2016 - 2:21am

I always love reading your anti-anti-establishment posts, Pat.

I think probably most coaches aren't actually thinking about the stuff you're saying, but they just intuit it and thus it becomes "what football coaches do".

The point about 2pt conversions not being "football" is rather interesting. For a quite a few NFL coaches who played, the 2pt conversion probably didn't even exist.

40
by Pat :: Tue, 01/26/2016 - 11:23am

I completely agree. Most coaches actually say things that completely lead you to conclude that they're thinking along these lines.

Like I implied, the thing I hate about articles like this is that they simplify the situation. And the *biggest* simplification that they do is that they treat it like a single-player game. A two-player game has completely different dynamics than a single-player version.

The point about 2pt conversions not being "football" is rather interesting. For a quite a few NFL coaches who played, the 2pt conversion probably didn't even exist.

That's a good point, too. Not quite what I meant, though. It's just that it's a totally artificial setup from the rest of the game.

Football is poker. If you're evenly matched, you take little bets, probe here and there, and then look for that opportunity to go "all in" and take a shot. If it's not there, you fold (punt). And the 2-point conversion doesn't really give you that option. You can't punt to a safe play - not without a timeout, at least. It's just a bad option for a coach to make.

And that's a point that people don't think about much at all. If a coach doesn't feel comfortable about a 2-point play because it's "not a normal football situation," he shouldn't do it. Because his decision-making isn't clear. Again, poker: if something *feels* wrong and you can fold, you fold.

34
by Bobman :: Mon, 01/25/2016 - 10:15pm

At the very, very least, you line up for two and the other team is in kick-blocking package and has to burn a TO or risk giving up an easy two. Then you come out after the TO and either kick or run it. Whatever. Won't work often, but it's gamesmanship that could get inside your opponents' heads.

Shit, this guy's crazy.
Shit, this guy is outcoaching us even on extra points.
Shit, we have no idea what we're doing against this guy.

I advocate trying that on short 4th downs as well, though you might have to take a last second TO if you're not prepared to go for it and the other team has not called for one.

41
by Pat :: Tue, 01/26/2016 - 11:26am

Before they pushed the PAT back, I completely would've agreed. Get the kicker used to being in "simple" 2-point plays. Get a QB in there as a holder, have him focus the hell on simple 2-point plays, too.

Then *actually have the routine* for the PAT be line up to go for two, then have the kicker and backup QB shift to their normal positions if you want to kick. But don't do it until the defense sets.

Obviously, now, that idea's shot.