Is Kobe Bryant clutch?
Are Kevin Durant or Dirk Nowitzki? Was Michael Jordan ever clutch?
These questions may seem ridiculous, as each of these players has made many “big shots” in the dwindling moments of close games. But how many of those shots are attributable to “clutchness,” and do they matter as much as we think they do?
All of those people mentioned are/were professional basketball players. They spend many hours each day perfecting their shot form on different plays from different areas of the floor. A specific isolation or post-up play, when run in a game, in the beginning or end, has probably been run multiple times during practice or scrimmages, and the offensive players involved have mastered these situations to the best of their respective abilities. The chances of those shots then going in are then left to just that – chance. Taken from a distribution of an uber-complicated probability model just like every other shot in the game.
Consider this: players like Kobe and Durant may be able to make so many shots in “crunch time” simply because they’re better players, not because they’re clutcher players. A play is run, and the abilities of those players are put on display in an effort to score.
Take a look at the top eight point-scorers in crunch time last season, courtesy of 82games.com:
The first thing you may notice is that perhaps Kobe and Durant weren’t as clutch as you thought, both shooting well below their season averages. But what I’d like to point your attention to is the minutes column. These numbers fall between 104 and 161 minutes over the course of the season, not much more than the course of a few games. This is a statistical concept called small sample size – there simply is not enough data to make an overarching prediction about any of these players. If James Harden were to go out on a three-game stretch and shoot .402, not very much would be made of that. So why does this same amount of minutes – 143 out of 3936 minutes in a season – over the course of an entire season draw such attention from fans and media alike? A lot is made of players who perform “when it matters.” Here’s a concept: the first three quarters of the basketball game actually matter three times more than than the fourth – 36 minutes compared to 12. All points count for the same amount from quarter to quarter, and the first five minutes matter just as much as the last five.
According to Rob Mahoney of the New York Times, “No player can be [clutch]; the word itself describes but a tiny slice of past performance, and indicates the timing and importance of a particular play rather than a fundamental attribute of any one player… Jordan wasn’t a winner in crunch time. He was just a winner.”
Shall we take a look at who was best at making “clutch” shots? Here’s the same data, sorted by field goal percentage:
As you can see, the top five most efficient scorers in these situations were all centers, players who normally shoot better percentages then the rest of their teams due to the nature of their close-to-the-basket shots. Were they more efficient in the clutch than their peers, or simply more efficient than their peers in general?
Statistics has a test specifically designed for situations like this. Known as the “two-sample t test,” this tests takes two sets of data and provides, with 95% confidence, whether or not they come from different probability distributions. In this case, we’d want to decide whether players are actually performing differently (better or worse) in the clutch, or if they’re just as good as they are for the rest of the game.
This test would be most effective with the most data possible, so let’s start with Kyrie Irving, who took the most crunch-time shots of anyone else in the NBA last season. A comparison of his .467 shooting on 38.8 attempts to his .452 shooting that season says that the two numbers are too similar to say they’ve come from separate distributions.
This is called rejecting the alternative hypothesis: just like how a criminal is innocent until proven guilty in the courtroom, this test assumes there is no clutch factor changing the data, unless sufficient data says otherwise.
For Kobe, his .426 in the clutch, compared to his .463 percentage that season, although seemingly very different, comes from a small enough sample size that the test detects no significant clutch factor. Even if it did, it would say Kobe – who is widely hailed as being clutch – is actually a less efficient player in those crunch-time moments. Conflicting evidence for widely held opinions make the clutch argument a difficult one.
Despite all of that, the fact remains that NBA players are people, and any psychologist will tell you that their performance would be affected by their surroundings and situation. NBA players themselves refer to the concept of clutch as fact all the time, and talk about their nerves in late-game situations. Does a player’s personal confidence, or belief in clutch, affect his performance in such situations? It’s certainly possible, and there are many parts of this discussion that statisticians might never be able to solve or agree upon.
Another possibility is that the minus for defensive focus on star players, combined with the plus from their clutch, causes the stats to be such a wash, in which case, the data could be deceiving. What’s deeper behind the numbers?
With such small sample sizes for clutch shots, alongside some conflicting evidence, it is very difficult to make a concrete decision either way on whether or not clutch exists. If it does, though, its effect is many times smaller than most people assume. Not only does it have little effect on the efficiency distributions, but in terms of number of shots over the course of the game, having a player who’s clutch would only help for a small amount of time, only doing so if the game were close.
So, what do you think? Does clutch exist? Does it matter?
by Derek Reifer, Northwestern University