Monday, November 16, 2009

On Stats (and Football)

I cannot understand the outcry today over Bill Belichick's decision to 'go for it' in last night's game.  Not a single media member that I have seen on TV has stopped to ponder whether it was a decent decision.  Nope, they've been quick to pile on.  Yes, the decision didn't work out.  But let's not kid ourselves, it was unequivocally the right one.

As Nate Silver observes (yes, of PECOTA and 538 fame),

Bill Belichick is not dumb, provided that his goal is to help the New England Patriots win football games. Instead, much of the NFL's conventional wisdom on when to go for it on fourth down is horribly, horribly wrong -- teams are way too conservative and punt way too often. This is the one case where 9-year olds playing Madden -- it's no fun to punt in a video game -- quite literally make better decisions than most NFL head coaches. With that said, since the same flawed conventional wisdom can govern hiring and firing decisions, there may be a price to be paid for unconventional (if statistically correct) playcalling; see also Marty Mornhinweg.

If you follow that first link, it affirms--the Pats had a 79% chance of winning doing what they did, versus 70% by punting. So why is there outrage?

8 comments:

Greg said...

Westy, I couldn't agree more.

Chairman said...

Keep your eyes out for a post over at the Board Room on this topic. I'll throw out some thoughts for you guys to chew on, and I'd bet that Greg will reconsider and become more agnostic about this issue. I don't know if I can really build a case for this particular scenario (I don't have enough data), but I think that you're using a faulty set of assumptions for this type of analysis. Basically, the answer derived by the method for this particular question may be OK, but I don't buy into this method of analysis.

Here's a teaser question. What are those win probabilities based on, and how does the skill/performance of Peyton Manning compare to the skill/performance of the players upon which those win probabilities are based?

Robby said...

Win probabilities are based on historical average performance. But I believe for the specific Peyton/Patriots case it is skewed even more towards going for it on 4th.

Greg said...

Chairman, I'll keep an eye out for your post. Just to clarify, though, I was agreeing that the criticism of Belichick was way out of proportion. I'll stay tuned for an analysis of the specific numbers.

Chairman said...

Not only would historical average performance, not account for Peyton, but I believe that does so w/out enough sophistication with regard to situation.

The more pertinent question: What is the probability of getting 2 yards with a short pass, when the defense knows it's coming (i.e., the empty backfield)? This is more like a 2-pt conversion, which succeeds overall at a 50-55% clip, but I don't know exactly where to look for numbers about situational 2-pt conversions. But I suspect that the stats used aren't the most appropriate for this question.

Chairman said...

Interestingly, the latest Bill Simmons column tackles this issue, and actually brings up the exact point that I left in the previous comment. Simmons throws out numbers that suggest that the odds of the Pats actually getting the 2 yards on that play call is more like 40%.

I think that he covers a lot of the arguments on the rationality of the decision well, and combined with the other commentary, I don't have much to add there. However, I do have a some thoughts on the method of analysis that folks quickly throw out there, so my post over on the Board Room will focus more in on that angle.

Robby said...

I felt as though the Simmons article was pretty horrible, actually. Throwing out numbers and disregarding the baseline just results in fluff.

Despite not going into incredible detail, there just really isn't that much to argue against that the AdvancedNFLStats.com write-up doesn't provide. He also has some follow-ups on his site.

Robby said...

http://belichick-decision.heroku.com/

There's a calculator to help in your analysis. I will admit, when I threw in my random guesses on percentages it came out as a clear punt. However, when you look at the realities of the situation and the historical numbers you really, really have to skew them to not go for it on 4th.