NFL coaches commonly call a time out when the game is on the line and the opponents are setting up a last-minute field goal attempt. They are "icing the kicker," attempting to disrupt his preparation prior to the crucial play. But does it work? There are reasons to be skeptical. Players know that such a time out is highly probable, and should thus know what's coming. The time out doesn't fool anybody, and kickers are highly specialized professionals that learn how to make plays in the face of fierce opposition. A time out is not very fierce opposition.
On the other hand, kickers are creatures of habit. They fine tune a routine and presumably employ that routine for the typical field goal; extra time takes them out of this routine, even if they know what's in store. My personal view (words to suggest that this next idea may be really screwed up) is that kicking a football is similar to hitting a golf ball. The less you think about either, the more likely you are to take a swinging leg or club and actually make solid contact with an oddly placed and perhaps oddly shaped object on the ground.
What do the data say? Here's Science News' summary of a recent analysis of icing by Scott Berry and Craig Wood in Chance.
Berry and Wood analyzed data about field goal attempts during the 2002 and 2003 NFL seasons (including playoffs). They recorded the kicker, the length of the kick, the score of the game, the time left in the game, and whether a timeout was called by the defense before the kick. They even noted whether the field was grass or artificial turf and the weather conditions (sun, clouds, rain, snow, average wind speed, temperature--unless the games were indoors).
In these two seasons, there were 52 different field goal kickers, combining for a total of 2,003 attempts. Of these kicks, 1,565 (78.1 percent) were successful.
Berry and Wood then looked at what they defined as "pressure" kicks--those that occurred with 3 minutes or less remaining in the game (or overtime) and would create a lead or a tie for the team attempting the kick.
There were 139 such pressure kicks, and 101 (73 percent) were successful. The defense called a timeout 38 times before the pressure kick, and 24 (63 percent) of these kicks succeeded.
Berry and Wood also estimate a probability model which allows for the influence of factors like weather to be captured. The results appear sensible, and the "icing effect" remains.
A kick made indoors is more likely to be successful. Clouds also have a small beneficial effect on kicks. Rain or snow, on the other hand, reduces the chances of success. High winds also reduce the probability of success, but not as much as rain or snow.
In pressure situations, the odds of success change very little (a mean decrease of 1.8 percent). However, icing the kicker in such a situation has a pretty strong negative effect.
Using their model, Berry and Wood calculate that, for an average kicker, the estimated probability of a successful 40-yard kick in sunny weather is 0.759. The estimated probability under the same conditions for an average kicker who has been iced is 0.659. "Reducing the probability of a successful kick from 0.759 to 0.659 is a very important difference," Berry and Wood report.
I concur on the magnitude - it's very large, and suggests that icing is a critical strategy. There aren't many "free calls" that a coach can make that increase the probability of winning a game by as much as 0.1.
I have two quick observations. I've often wondered whether a mixed strategy is the appropriate approach here - i.e. call a time out with probability of, say 0.65. This should put the thought in the kickers mind that he might or might not have to execute, rather than the near certainty that he will be iced. But 0.1 is perhaps too large to give up in return for the benefits of additional uncertainty. If there are diminishing returns to icing, then perhaps the spot to play a mixed strategy would be when you have two or three time outs left. Finally, the sample size is small for my tastes - the effect is identified from just 139 "pressure" kicks, and with 30+ idiosyncratic kickers in the league, 139 seems a tad small. I'd like to see the sample expanded to a decade's worth of games.
Thanks to Blake Linney for the link.