Tipping Pitches: Sports: Why the Stat Heads Win


Monday, November 16, 2009

Sports: Why the Stat Heads Win

email to friend edit
Bill Belichick's controversial 4th down call against the Colts made for a fun time to be a sports fan, particularly for one who is a bit of a stat nerd.  As I wrote following the Indianapolis Colts' comeback win over the New England Patriots, I support Belichick's decision to go for it on fourth and two from his own 28 with a six point lead and about two minutes remaining in the game.

I know!  It's crazy.  While I understand the risks involved, I think it was the right move.  There is some subjectivity involved.  But objectively speaking, there is plenty of data to support this decision.

Of course, I expected to have very little support.  That, as apparent on ESPN.com, ESPN and ESPN Radio, seems to be the case.

Fans are separated into two groups: 1) the stat heads and 2) meat heads who don't care what the stats say.

The stat heads, I admit, are prone to see football in a vacuum.  Use league-wide stats, ignore who is home or away, ignore momentum.

On the flip side, meat heads are impossible to argue with.  You punt there.  It's just what you do.  Conventional wisdom wins.  No other coach goes for it.  Flawed arguments, omitting important factors, and completely ignoring stats.

Mainstream sports media was largely dominated by the meat heads.  But the stat heads were able to gain some steam.

A couple of the more prominent pieces in support of going for it on fourth down were written by Brian Burke for the New York Times Blog and Joe Posnanski for his own blog.  Many of the same arguments in both.

The core argument for Bill Belichick is that the odds of successfully converting fourth down and two was 60%.  The odds that the Colts would score a touchdown from their own 29 was 53%.  And the odds that the Colts would score a touchdown following a punt was 30%.

The meat heads would counter this by saying that the Patriots had a 70% chance of stopping the Colts after a punt compared to a 60% chance of succeeding on fourth down.  You have to punt!

Uh, that's not how it works.  If you understand statistics, the result is that the probability of winning if the Patriots go for it on fourth down was 79% versus 70% if they punt.

I applaud the work, but think it's dangerous to use stats as absolutes.  It isn't black and white.  Those odds apply to all teams in similar situations.  You need to account for these two offensively superior teams and the situation.

The likelihood of the Colts scoring from the Patriots' 29 has to be higher than the norm.  Similarly, the likelihood they score after a punt must also be higher.

Additionally, I feel the 60% number for successfully converting on fourth down may be a little generous.  Even for the Patriots, who have been successful going for it on fourth down with a pass 70% of the time dating back to 2004.

However, keep in mind that most fourth down plays do not decide a game.  In this case, the Colts could focus entirely on stopping the short play.  Not the case in a typical fourth down play that must account for stopping the long pass.

What is the correct probability of each event occurring?  There isn't one.  But we can look at ranges.

So when using stats in this case, I wanted to consider a range for each event.  Probability of success would be determined by ranges in the following scenarios:
  • Probability of converting on fourth down between 40 and 65%
  • Probability of Colts scoring from the Patriots' 30 between 55 and 80%
  • Probability of Colts scoring after a punt between 30 and 55%
I think these are unbiased and reasonable.  In fact, I'd expect them to favor the likelihood of supporting a punt.  Or it will at least be difficult for someone to make the claim that I left out a likely scenario.

Then I applied Brian Burke's formula to every possible scenario at multiples of five (for example, probability of converting on fourth down at 40, 45, 50, 55, 60 and 65%).  Total of 216 scenarios.

I know, I'm nuts.  And a nerd.

This is what I found:
  • Of the 216 scenarios, 177 indicate odds of victory when going for it on fourth down were equal to or greater than odds of victory when punting.
  • 39 of the 216 scenarios indicate that punting was the better decision.
  • In order for punting to be the best option, the most common factor was probability to convert on fourth down.  In 15 of those 39 scenarios, the chance of successfully converting on fourth down was 40% - a number unlikely to be supported by many.
  • In 32 of the 39 pro-punt scenarios, the assumption is made that the Colts would have had only a 30 or 35% chance of scoring after a punt.
I provided ranges to cover any reasonable odds.  If you  believe that the Patriots had a 60% chance of succeeding on fourth down, there is no reasonable reason to punt.  If you believe the Colts had greater than a 35% chance of scoring a touchdown with under two minutes from their side of the field (as they had already done twice in the quarter), there are few reasonable explanations to prefer the punt.

Reasonable football fans will grant the Patriots had at least a 50% chance of converting on that fourth down.  Reasonable fans would acknowledge the Colts had a 40% or better chance of scoring on a drive after a punt.

If you grant those things, you acknowledge -- intentionally or not -- that going for it on fourth down was the right move.

So why, then, would someone prefer to go against the odds?

This is the problem I have with people who say you can't use stats and probability when judging this move: Then on what do you base your decision?

Aren't all moves based on probability, even if they aren't based on absolutes?  For example, what should Belichick be thinking as he is making his decision?  Should he do what gives him the highest probability of winning, or should he do what seems obvious?

I'd argue that everyone is basing their strategy on some sort of probability, even if it's faulty logic or missing valuable information.  An example is the argument I've heard repeatedly that you don't give Peyton Manning a short field.  You make him drive 70 yards.

Of course that's the preference!  But that is missing a very important factor.  The Patriots didn't just hand the ball to the Colts at the 29 and say they prefer to give the Colts the ball with a short field.  There was a decision before that to go for it on fourth down.  The exact odds are up for debate, but it's reasonable to say that there was a 50/50 chance that Peyton Manning even gets the ball in the first place.

You can't omit the odds of converting on fourth down from the equation.  That, unfortunately, is what many are doing.  Or they are assuming that the Patriots were never going to convert, which is clearly bad logic.

Not to mention, many are assuming that it's an automatic that the Colts win if they take over at the 29.  Not only do they first have to score a touchdown, but if not for some poor strategy by the Patriots defense, New England may have gotten the ball back one more time to go for a winning field goal.

And to be honest, considering the possibility of getting the ball back also needs to be considered in the odds when deciding whether or not to go for it.  Guess what?  Including that factor would only improve the probability of winning when going for it on fourth down.

In other words, I'm doing everything I can to help the meat heads.  I'm allowing for the possibility that the odds the Patriots convert is below 50% (no matter how irrational that may seem).  I am allowing for the possibility that the Colts had below a 40% chance of scoring a touchdown if the Patriots punted.  And I also omit a scenario which only would have helped the pro-go for it argument.

I'm generous in that way.  But the meat heads are wrong.

And let me be clear, I am not saying that this means going for it is absolutely the right move.  As I've said before, I hate absolutes.  I like the move.  Depending on how you view the different probabilities, it may be the right move.

But many are behaving as if making such a move was stupid.  It wasn't.  It was well thought out.  It was logical.  And it had merit.


Post a Comment