4.7/5 - (67 votes)

In 1985, three psychologists proved that basketball’s famous “hot hand” was an illusion, and researchers dismissed the idea as a fallacy for years. Then two researchers found a flaw hiding in the way the shots had been counted.

The myth everyone loved to debunk

Every basketball fan knows the feeling. A player sinks two, three, four shots in a row, and suddenly everyone in the arena is sure the next one is going in too. That player has the “hot hand”.

Thomas Gilovich, Amos Tversky and Robert Vallone decided to test it. Their 1985 paper, “The Hot Hand in Basketball”, attacked the question from four directions, each one stripping away more outside noise.

First, they surveyed 100 basketball fans from Cornell and Stanford. Then they studied the shot records of the 1980-81 Philadelphia 76ers, then free throw data, and finally a controlled shooting experiment, far from the game setting and its distractions.

The verdict was the same every time. Hits and misses looked independent of each other, and feeling “hot” predicted nothing. The “hot hand fallacy” was born, held up as proof of how badly humans read randomness.

Why our brains see streaks that are not there

Gilovich offered two explanations. The first is confirmation bias: fans go into a game already looking for streaks, so they remember the ones that fit.

The second runs deeper. People expect chance to alternate far more often than it really does. Genuine random sequences look too lumpy, so we decide they cannot be random. This is known as the clustering illusion.

See also  Guinness forbade his real name, so he signed his math as Student

It is the mirror image of the gambler’s fallacy. The gambler expects a streak to reverse. The hot hand believer expects it to continue. Both are betting that a short sequence carries a message.

The hidden flaw in the counting

The 1985 test looked simple. If the hot hand is real, a player should hit more often right after a streak of hits than right after a streak of misses. For the 76ers, those two rates were not significantly different. Case closed.

In 2015, Joshua Miller and Adam Sanjurjo reopened it. They showed that the test itself was tilted. When you look at a finite run of shots and only start counting after a streak, the shots you end up examining are biased toward misses.

Their own words are striking: “it is incorrect to expect a consistent 50 percent (Bernoulli i.i.d.) shooter who has taken 100 shots to make half of the shots that immediately follow a streak of three hits”.

How big is the effect? We ran a quick simulation of 200,000 seasons of a perfectly average 50 percent shooter taking 100 shots. Right after three straight hits, that shooter makes only about 46 percent of the next shots on average, even though nothing about them has changed.

So a truly random shooter should look slightly cold after a streak. If a real player shoots just as well after hits as after misses, that is not evidence against the hot hand. It is evidence for it. Reanalysing the 76ers data with this correction, Miller and Sanjurjo reported “significant evidence of streak shooting”.

A wave of new evidence

The doubts had actually started earlier. In 2003, researchers at Monash University pointed out that the 1985 study never checked its statistical power. Even if the 76ers had shot in streaks, the original tests were very unlikely to catch it.

See also  A 1936 poll heard from 2.4 million voters and still got it wrong

Bigger datasets followed. A 2011 study of more than 300,000 NBA free throws, covering five regular seasons from 2005 to 2010, found that players were more likely to hit a second free throw after making the first than after missing it.

In 2014, three Harvard graduates used tracking data that could control for shot location and defender position for the first time. They found a “small yet significant” hot hand effect. A 2021 study of NBA Three-Point Contests from 1986 to 2020 found “considerable evidence” of hot shooting.

So, is the hot hand real?

Partly, and less dramatically than fans imagine. Other recent studies still find nothing. The evidence suggests that only a small subset of players shows a hot hand, and that the effect is small when it appears.

One reanalysis of the three-point contest data even found the effect was limited to shots taken from the same spot. That would fit free throws and shooting contests, but not the chaos of a real game.

The lesson goes beyond basketball. A study about human bias turned out to contain a statistical trap of its own, one that went unnoticed from 1985 to 2015. Randomness fools fans, and sometimes it fools the people studying the fans too.