Thursday, February 10, 2011

On Win Probabilities and Watching Sports

Brian Burke's Advanced NFL Stats "blog" is one of the best sports websites on the planet. It's what happens when you turn a rocket science loose on raw sports data. That said, Burke has also greatly diminished my enjoyment of watching NFL football. For example, he has a wonderful win probabilities equation that follows live play-by-play action of all the games. It uses game score, offense field position, and time left on the clock to calculate a team's win probability (WP) based on all historical games played since 2003. I always follow along when watching the Super Bowl.

Here is Super Bowl 45's WP graphed:

Note that huge drop on the left hand side, that's when the Packers went up 14-0 with 3:34 left in the first quarter. At that point, Green Bay had a 90% probability of winning.

If you're like me, you think "It's only a two possession game. There are still more than three quarters of football left to play. Anything can happen." But history says only 10% of teams overcome a 14 point deficit even that early in the game. When you tell me there is a 90% chance of rain, I take my umbrella. Tell me there's a 90% probability of a team winning, I turn the TV off and find something else to do (though I didn't on Sunday). You can see that according to the WP graph, the probabilistic outcome of the game wasn't in doubt very often.

You can read his analysis of the game-- the key plays and coaching decisions. It is now too hard to watch a coach punt on 4th-and-short when the data say he'd increase chances of winning by going for it. It's too hard to hear Phil Simms say something stupid like "It's too early to go for two," when the probabilities of that team winning are higher with a 2-point attempt than an extra point try. You wish that Burke was on the sideline with his computer so the coach could ask him "should we run or pass here?"

I imagine that since college offenses are much more varied and wide-open than NFL, average scores are higher and there is much more variance in outcomes. Maybe that's why I enjoy watching NCAA so much-- some ignorance is bliss.

No comments: