Sep 23 2013

The Hot Hand Exists

Probably. But not really in the way you think. It’s kind of a long story, actually.

Of the many triumphs data has enjoyed over intuition in the last decade or so, none has been as rousing or total as the debunking of the notion of the “hot hand”—the phenomena wherein a player who’s hit a handful of consecutive shots supposedly becomes more likely to hit the next one. Thanks to the work of economists like Sandy Weil, hoop fans haven’t been disabused of the notion of the “hot hand” as much as we’ve been hit upside the head with analytic 2x4s, then shuttled away to Soviet-style reeducation camps where we’ve recited, in automatonic unison, “There is no hot hand. There is no hot hand. There is no hot hand,” until we’re cured.


32886996-media_httpwwwfuturity_zdzEcIn a must-read over at Grantland (“must-read” even by Zach Lowe standards), Zach Lowe broke down a paper presented at the New England Symposium on Statistics in Sports this weekend that persuasively debunked the debunking of the hot hand. Sort of.

The paper’s authors—former Suns consultant John Ezekowitz and a handful of Harvard grad students—argued that the “hot hand” exists, but is camouflaged by the extent to which a player’s perception of his own “hotness” affects his shot selection.  

To wit: When a player hits his last shot, he is not only more likely to take his team’s next shot, but is more likely to miss it. This is the logic by which the “hot hand” has been argued out of existence, and Ezekowitz concedes that it’s sound. But—and this is an enormously consequential but—the “hot” player’s field goal percentage plunges only because when he thinks he’s hot, he tends to take more difficult shots. And here’s the really interesting part: if you adjust for the difficulty of a shot that follows a make, the “hot” player is more likely to hit that particular shot than the averages would suggest.

The methodology is as cool as the conclusion. Ezekowitz and Co. used SportsVU cameras to determine with (until very recently, impossible) precision, the relative difficulty of the shots ostensibly hot-handed players were taking. Looking back at 83,000 shots from the 2012-13 season, the cameras allowed the study’s authors to account not only for the location from which hot players launched shots, but also for how opposing defenses responded (defenders moving closer to the shooter, height of defender, etc.)—an obviously crucial determinant of a shot’s degree of difficulty. Once they controlled for a shot’s DoD, they found a “hot” player was somewhere between 1.4 and 3 percent more likely to hit it.

In other words: let’s say Thad Young sinks a pair of 15 footers on consecutive possessions and, on his next trip down the floor, chalk full of confidence, heaves up a contested 18-footer with Chris Bosh in his face.  Due to the location of the shot, and the fact that an enormous human is on top of him, this is a bad shot: let’s say, in a vacuum, it has a 27 percent chance of going on. But, because Thad Young is “hot,” the paper’s authors might say he has a 30 percent chance of hitting it. Still not great, but not nothing either.  

I’m not sure this paper will (or could) have much of an impact on game strategy given how modest the “hot hand” effect it identifies is, but it should serve as a lesson for those of us (myself included) who follow the numbers wherever they go and thumb our noses at those who resist: be humble. Science is a winding road, and sometimes it drops us off exactly where we started.