Death of the Hot Hand

Dateline: 12/18/96

In 1985, the academic psychological journal, Cognitive Psychology, published some research on whether there is such a thing as a "Hot Hand" or whether basketball players are really streaky. Since very few members of the mainstream press read Cognitive Psychology (and I don't blame them), one of the paper's co-authors, Dr. Amos Tversky, took the results of that paper to the public a few years later. What his work (originally authored with Thomas Gilovich) demonstrated was that streaks do not exist, regardless of what Hubie Brown says.

The work Tversky did cannot be ignored as psychobabble. It used real NBA data, taking an in-depth look at the Philadelphia 76ers over the course of the 1980-81 season. Tversky and his colleagues tracked every shot the Sixers took and looked at whether any players were more likely to make a shot after they made their previous one. For example, the Sixers were often lead by the so-called streaky shooter, Andrew Toney, but Toney was statistically more likely to make a shot after missing one than after making one. Although most people considered Toney streaky, the statistics that people like Hubie Brown cling to didn't support that conclusion.

It is all a matter of perception, according to Tversky, who was a Professor of Psychology at Stanford University. He said that we look for streaks even when they are not there or when what is occurring is nothing more than random variation. By believing in streaks and trying to feed players on hot streaks, we are just deluding ourselves, he said.

As someone who has coached and played basketball, I have a hard time believing this result. As someone who is familiar with the statistics used in Tversky's work, I find it hard to argue with. So what is right?

To answer this, let's first take a psychological test to see whether you can pick out a streaky player vs. a non-streaky player. Then I will explain why I think streaky players may truly exist, but cannot be identified.

Can You Identify Streaks?

Here is the test. Below I simulate two players who shoot 50% on average. Of these two players, one, both, or neither is "streaky". A "streaky" player is one who shoots 60% after he makes a shot and 40% after he misses a shot. A "non-streaky" player always shoots 50% regardless of whether he made the previous shot. You can see these simulations by pressing the Run button below. In the sequence of shots that appear, "X" denotes a made shot and "O" denotes a missed shot. After the two sequences, you must choose which of the two sequences, if any, represents true streakiness. (Note that the answer changes each time you press "Run".)

1.
2.

Which of the above two sequences represents true streaky performance?

First Sequence is Streaky
Second Sequence is Streaky
Both Sequences are Streaky
Neither Sequence is Streaky

I personally could not identify streakiness in these sequences. I doubt that you could.

But take a look at the sequences you simulated. I'm sure you can find a run of made shots that was four or five in a row even if the simulation was completely random, not streaky. In a game, it may be very easy to call that series of four or five in a row a "hot streak", even though it is not. It is just that player going through a normal biorhythm, if you believe that sort of thing.

Now let's say that a streaky player shoots 80% after a made shot and 20% after a miss. Can you identify the streaky player?

1.
2.

Which of the above two sequences represents true streaky performance?

First Sequence is Streaky
Second Sequence is Streaky
Both Sequences are Streaky
Neither Sequence is Streaky

You should have done better with this one. I did. Of course, I started seeing players hitting 10 or 11 in a row, once seeing 19 in a row. That made 4 in a row seem pretty random. Of course, 11 in a row is also extremely rare in the NBA (the record is 35 in a row by Wilt Chamberlain), certainly not as common as you see in this simulation, indicating that streaks of this type do not exist.

So How Does This Matter?

Next time you're watching a basketball game and the announcer says Dell Curry is hot, think about what that means for a strategy. For Charlotte, Muggsy Bogues is going to be trying to get Curry the ball as much as possible and Curry will be shooting because he "feels it". For Charlotte's opponents, they also sense that Curry is "hot" and will try to "cool him off" by playing tighter defense on him. Ultimately, whether Curry is truly streaky or not, because his opponents should recognize this as quickly as his own teammates, that competition should be enough to end a streak.

This is very much my opinion. Even if players can get hot or cold in reality, having opponents who can recognize that streakiness will cause an end to that streak and, as a result, prevent studies like Tversky's from showing it. If a team is playing against a stupid defense (the Clippers?), we may find more evidence of streaks.

Post Script

In June, Dr. Amos Tversky, the principal author of this and other sports-related psychological work, died. This work was probably the most prominent of his research into popular topics. A short biography of this rare scientist who made his science accessible to sports fans can be found here .