Myth of the Hot Hand ... no myth?

Are 'Hot Hands' in Sports a Real Thing? Maybe there's something to it, according to today's New York Times (5/8/13).

The old news (imagine the hot hand involved tests on Cornell hoops in the 1980s)
... our faith in hot hands is challenged by a rich and well-regarded body of science over the past 30 years, much of it focused on basketball, that tells us our belief is mostly fallacious. In one of the first and best-known of these studies, published in 1985, scientists parsed records from the Philadelphia 76ers, the Boston Celtics and the Cornell University varsity squad and concluded that players statistically were not more likely to hit a second basket after sinking a first. But players and fans believed that they were, so a player who had hit one shot would be likely to take the team’s next, and teammates would feed this “hot” player the ball.

New news
Now, however, some new studies that use huge, previously unavailable data sets are suggesting that, in some instances, hands can ignite, and the success of one play can indeed affect the outcome of the next.
... In these big sets of data [entire NBA season, 50 pro bowling matches], which were far larger than those used in, for instance, the 1985 basketball study, success did slightly increase the chances of subsequent success — though generally over a longer time frame than the next shot. Basketball players experienced statistically significant and recognizable hot periods over an entire game or two, during which they would hit more free throws than random chance would suggest. But they would not necessarily hit one free throw immediately after the last.

... and the anti-hot hand
But if winning streaks have some rational basis, then by inference so would losing streaks, which makes the latest of the new studies, of basketball game play, particularly noteworthy. In that analysis, published last month in the journal Psychological Science, Yigal Attali, who holds a doctorate in cognitive psychology, scrutinized all available shooting statistics from the 2010-11 N.B.A. season.

He found that a player who drained one shot was more likely than chance would suggest to take the team’s next shot — and also more likely than chance would suggest to miss it. Essentially, he found that in real games, players developed anti-hot hands. A momentary success bred immediate subsequent failure.

billhoward
Are 'Hot Hands' in Sports a Real Thing? Maybe there's something to it, according to today's New York Times (5/8/13).

The old news (imagine the hot hand involved tests on Cornell hoops in the 1980s)
... our faith in hot hands is challenged by a rich and well-regarded body of science over the past 30 years, much of it focused on basketball, that tells us our belief is mostly fallacious. In one of the first and best-known of these studies, published in 1985, scientists parsed records from the Philadelphia 76ers, the Boston Celtics and the Cornell University varsity squad and concluded that players statistically were not more likely to hit a second basket after sinking a first. But players and fans believed that they were, so a player who had hit one shot would be likely to take the team’s next, and teammates would feed this “hot” player the ball.

New news
Now, however, some new studies that use huge, previously unavailable data sets are suggesting that, in some instances, hands can ignite, and the success of one play can indeed affect the outcome of the next.
... In these big sets of data [entire NBA season, 50 pro bowling matches], which were far larger than those used in, for instance, the 1985 basketball study, success did slightly increase the chances of subsequent success — though generally over a longer time frame than the next shot. Basketball players experienced statistically significant and recognizable hot periods over an entire game or two, during which they would hit more free throws than random chance would suggest. But they would not necessarily hit one free throw immediately after the last.

... and the anti-hot hand
But if winning streaks have some rational basis, then by inference so would losing streaks, which makes the latest of the new studies, of basketball game play, particularly noteworthy. In that analysis, published last month in the journal Psychological Science, Yigal Attali, who holds a doctorate in cognitive psychology, scrutinized all available shooting statistics from the 2010-11 N.B.A. season.

He found that a player who drained one shot was more likely than chance would suggest to take the team’s next shot — and also more likely than chance would suggest to miss it. Essentially, he found that in real games, players developed anti-hot hands. A momentary success bred immediate subsequent failure.

This all seems so dumb, at least as written up in the Times. Maybe it sounds less dumb in the paper. Of course there are hot and cold streaks. That's what random chance produces. The question is whether those streaks are self-generating. I'd be shocked if someone doesn't write a counterpoint to this eviscerating it within the year.

ugarte
This all seems so dumb, at least as written up in the Times. Maybe it sounds less dumb in the paper. Of course there are hot and cold streaks. That's what random chance produces. The question is whether those streaks are self-generating. I'd be shocked if someone doesn't write a counterpoint to this eviscerating it within the year.

The intersection of people who understand methodology and people who write about studies in the media is a null set.*

(* this could be simplified to The intersection of people who understand anything and people who write about anything in the media is a null set.)

Speaking as a real statistician (B.S. Statistics, CALS '93), I didn't find the overall article terribly surprising. But we need to step back and really understand the greater context in which these streaks occur.

Big data can reveal patterns that simplistic analysis will never reveal. However is the correct conclusion that "success did slightly increase the chances of subsequent success?" I don't think so. I think that the factors surrounding the success that day (i.e., a defensive scheme that makes it easier to see the basket and control the shot, subtle differences in how a player is seeing / feeling / has precision physical control that day, etc.) are more legitimate explanations. This would help explain why, in the course of a game, a player could be considered to have a "hot hand" through the course of a game. If a player makes three of five baskets from their last five shots, missing the (middle) third shot, they would still be considered to be hot, even though they only have two two-shot "streaks."

The reality is, each shot is an independent event, but in the context of a game, may and do share contributing factors. I believe that because of this, streaks absolutely do occur, but strongly disagree with the conclusion that the success of one play impacts another. Having said that, if I see a pattern that a player is making a higher percentage of shots than normal, I would pass to him or her, because those other contextual factors I mentioned (or others) may be coming into play.

Identifying patterns is easy. Explaining why they exist needs to be more thoughtful and logical.

I had a little trouble following how there's a hot hand and a concurrent (?) cold hand. Is this like anti-matter? Are there other factors such as the defender switched on the player has a twisted ankle and so the hot player is getting off easier shots, or the shortstop is playing the hitter two steps out of position, or ? But massive data crunching makes this fascinating. I was equally fascinated by the Flyers fanatic who charted every offensive rush of the season to find that carrying the puck in is more successful than giving up the puck in a dump and chase offense.

How long till pressbox or ceiling cameras can track every play, ID every player and defender, see where the puck/ball is, auto-determine a zone vs. man to man, and show the coach and players incredibly detailed tendencies. Next year, only the Cowbox can afford; in five years, it's part of the GoPro camera.

I would argue that the same data set should also be able to indicate the presence of a cold hand / "cold streaks." A player playing with the flu, for example, would likely not perform as well (not really rocket science) and through the course of a game (or the illness) may have what is seen as a "cold hand" or "cold streak." The same rationality that works for explaining streaks of successes should have a related explanation correlated to failures.

Of course, some observed streaks are simply a function of the laws of probability. In the course of any sport, some crazy, improbable streak will happen. And in life, people have been hit by lightening twice, or won the lottery twice. Take a large enough data set and you will find those incredibly improbable occurrances. It is part of the law of truly large numbers, attributed to Persi Diaconis.

I could see instances where a player started off with a "hot streak", then, due to other circumstances (excessive fatigue, sweat getting into the eyes obscuring vision, playing with pain that occurred during the game, equipment failure or issue, changes in defense schemes) became "cold." But logic would dictate that there was some underlying reason for this change. A successful shot is not the predictor for the next shot, but rather, the indicator of some other underlying explanation. In the course of a game, you need to play now, analyze later.

Remember, a player who typically makes 90% of his shots still only has about a 12% of making 20 in a row.

billhoward
How long till pressbox or ceiling cameras can track every play, ID every player and defender, see where the puck/ball is, auto-determine a zone vs. man to man, and show the coach and players incredibly detailed tendencies.

The EPL is pretty damn close to this already, and the metrics are hysterical.

It's only a matter of time until somebody slaps a skating pedometer on Sidney Crosby.

I dont equate hot hand with making a shot and then making the next one.. i think its when a guy makes a crazy/tough and then makes another tough shot.

this study is more along the lines of making a FT and then making another.. and does making a layup lead to making another shot?

you cant see a hot hand with stats, you have to be watching a game to see it happen, and when you are on the court playing you can feel it as well. when someone who cant throw it in the ocean makes 6 of 8.. and how do you know when the hot hand ends.. is missing a shot part the end? not in real life.

C'mon, everything is always like "NBA Jam," where if you make 3 straight shots, you go "on fire" and can do things like dunk from mid-court while doing 800 flips.

This is a favorite topic of mine because it's the only time I really deviate from the SABR style of thinking. The human body and human psyche isn't random. Things like health or confidence (or lack thereof), or an alteration of your form/stance aren't randomly generated to fall into a bin of statistics. If a player feels confident for whatever reason: the previous result, or an extra session of practice, or from the way a cheerleader smiles, he/she might approach a play differently or more aggressively than normal. On the flip side, think of a case of the "yips" in golf or a baseball player thinking "shit, I'm 0-12 this series. Maybe I'll let this 3-1 pitch go by this time." Sports psychology exists. I think hot/cold streaks also exist and are directly a result of very human sources.

I don't believe EVERY hot/cold streak is human-generated...sometimes you get "the breaks" or a "lucky bounce," but those tend to average out over the course of a season depending on talent level, as SABR folk argue.

One reason I love to see some hard data behind performance claims is because we're awash in meaningless statistics

• The Penguins are say 18-1-1 when leading entering the third period. But never a comparison to how all teams fare when entering the third period, or to when it was a one-goal lead for the penguins.
• The batter hit safely in 6 of his 11 at-bats, which means he was also 6 of 12 and maybe 6 of 13.
• Team got the tying goal with 48.2 seconds left, as if the 0.2 seconds was important.

I thought a meaningful stat was Cornell's incredible record in one-goal games (not counting that we've lost the last, ah, three?). Except the comparison shouldn't be against it being 50-50, but against our record over the same period, which is also overwhelmingly good.

Nice to see the topic gets some attention. Give Nate Silver credit for making statistics a little sexier, too.

RichH
This is a favorite topic of mine because it's the only time I really deviate from the SABR style of thinking. The human body and human psyche isn't random. Things like health or confidence (or lack thereof), or an alteration of your form/stance aren't randomly generated to fall into a bin of statistics. If a player feels confident for whatever reason: the previous result, or an extra session of practice, or from the way a cheerleader smiles, he/she might approach a play differently or more aggressively than normal. On the flip side, think of a case of the "yips" in golf or a baseball player thinking "shit, I'm 0-12 this series. Maybe I'll let this 3-1 pitch go by this time." Sports psychology exists.

William James, white courtesy telephone please.

Still, I suspect almost all of what we think we are seeing as "hot hand" is confirmation bias. The SABR people aren't all that bright, they're just brighter than Joe Morgan, but on this when the empirical dust settles I'll bet they're correct that "streakiness" is just the tail of good ol' random variation.

When the scoreboard clock had no tenths, Bill Russell could speak at length about whether at 0:01 and the Celtics down by one, was that a long or short one second left. Then, only, did I care about tenths on the shot clock.

Any set of numbers in the hands of Barry Melrose is like having a loaded .22 pistol to a five-year-old. Hearing Barry try to figure aloud the differential in two penalties that have, say, 1:14 and :57 left ... that's why McDonald's has machines that make change on an order for $4.51 when you hand over $5.05.

billhoward
When the scoreboard clock had no tenths, Bill Russell could speak at length about whether at 0:01 and the Celtics down by one, was that a long or short one second left. Then, only, did I care about tenths on the shot clock.

Any set of numbers in the hands of Barry Melrose is like having a loaded .22 pistol to a five-year-old. Hearing Barry try to figure aloud the differential in two penalties that have, say, 1:14 and :57 left ... that's why McDonald's has machines that make change on an order for $4.51 when you hand over $5.05.

Or how about it's $15.65, you give them $21, and they want to give you the $1 back saying they don't need it.

Jim Hyla

billhoward
When the scoreboard clock had no tenths, Bill Russell could speak at length about whether at 0:01 and the Celtics down by one, was that a long or short one second left. Then, only, did I care about tenths on the shot clock.

Any set of numbers in the hands of Barry Melrose is like having a loaded .22 pistol to a five-year-old. Hearing Barry try to figure aloud the differential in two penalties that have, say, 1:14 and :57 left ... that's why McDonald's has machines that make change on an order for $4.51 when you hand over $5.05.

Or how about it's $15.65, you give them $21, and they want to give you the $1 back saying they don't need it.

Or give them a $20, then after they enter it in the cash register hand them the $.65 and watch the wheels spin.

George64

Jim Hyla

billhoward
When the scoreboard clock had no tenths, Bill Russell could speak at length about whether at 0:01 and the Celtics down by one, was that a long or short one second left. Then, only, did I care about tenths on the shot clock.

Any set of numbers in the hands of Barry Melrose is like having a loaded .22 pistol to a five-year-old. Hearing Barry try to figure aloud the differential in two penalties that have, say, 1:14 and :57 left ... that's why McDonald's has machines that make change on an order for $4.51 when you hand over $5.05.

Or how about it's $15.65, you give them $21, and they want to give you the $1 back saying they don't need it.

Or give them a $20, then after they enter it in the cash register hand them the $.65 and watch the wheels spin.

Oh that's just mean.