Lucky Misses: Contesting Three Point Shots

Mandatory Credit: Soobum Im-USA TODAY Sports

In any sport, luck plays a role in the outcome, which is part of the fun. Los Angeles Clippers coach Doc Rivers captures this with a favorite saying: that the NBA, “Is a make or miss league,” conveying, among other things, that even when a team plays “the right way” sometimes the shots just fall more for one team than the other.

Three-point shooting, rightfully, is often cited as a particularly noisy number. Though in no way should that imply that three-point shooting doesn’t take skill.  It just takes a large sample to separate out the skill from luck.

Now, thanks to SportVU data released by the NBA, we can see the number of shots contested by each player from different part of the court. This data describes how well opponents have fared when the player contests shots and gives a ‘Plus/Minus’ stat that calculates how well the opponents performed compared to their average otherwise from that location. Unfortunately, when it comes to contesting three0point shots of opponents by individual players things get even more noisy than shooting threes.

Using data from 2014, which is also available from NBA.com, I did a simply year to year correlation of opponent field goal percentage from three when a player is contesting the shot. I used 50 three-point contests as the minimum for this one, but is makes little difference as, basically, there is no relationship. As you can see in this scatter plot.

We can also see that looking at the top ten from this year in terms of lowest percentage of three point shots made when they contested them and the where they ranked last year.

Only Damian Lillard from the top ten this year even sniffs the top ten percent last year. Andre Roberson has become a defense plus/minus star, but last year his three-point contest numbers were one of the worst in the league. Part of this is a function of a small sample size for individual players, but the lack of convergence to any ‘real’ ability over a season and a half is enough to question the reliability of this number as a predictor of anything.

Luck is often described as noise in analysis, because it represents what has happened that isn’t necessarily a good predictor of what will happen. There is nothing wrong with descriptive stats that tell us what happened, as Ian Levy noted in an off season piece here at Nylon Calculus. The problem comes when we try to use noisy descriptive statistics as predictive statistics.

That individual three-point defense is noisy shouldn’t come as too much of a surprise. For example, Ken Pomeroy did an interesting study in NCAA basketball that indicated that on a team level defensive statistics were more noisy than offense, with three-point defense in terms of percentage made as one of the noisiest, with three-point attempts per forty minutes much more predictive.

All of this is not to say that tracking contested shots is useless, open shots are really easier to make and you can’t deter three-point shots without contesting them. League wide we can make some generalizations about contesting threes and efficiency in the NBA thanks to SportVU. This data comes from the NBA SportVU data via Krishna Narsu, using this and last year’s data:

Very Tight defense on a three0point shot, defined as a defender within 0 to 2 feet is pretty rare, only 1% of either catch and shoot threes or off the dribble, and not included are the number of three point shots deterred. With ‘Tight’ defense defined as the defender within 2 to 4 feet, the cross tab shows a 3.6% drop on catch and shoot threes, indicating it is worth the effort even if we can’t tell for an individual player whether they have had an effect.