It bugs me when people say statistics lie, because this simply isn’t possible. With a few exceptions, statistics can lie no more than your calendar can. All statistics are, after all, merely records of facts. What statistics can do is mislead, particularly if you start with the expectation that they are entirely reflective of the players that generate them and not outside context.
Let’s talk about a particularly timely example: DeMarcus Cousins. The Sacramento big man, who was suspended over the weekend by Paul Westphal during Westphal’s final days as head coach, is shooting 38.6 percent from the field this season. Turning to advanced numbers does little to help Cousins, whose True Shooting Percentage (47.8 percent) is far below league average. It’s easy to look at those numbers and criticize Cousins for taking so many shot attempts. (He’s using a team-high 27.8 percent of the Kings’ possessions.)
When you actually watch Cousins’ shots, however, they’re not so bad. If we could edit a tape that showed only the shot itself and not the outcome, I suspect you’d agree that his shot selection has improved from last season. Cousins still has a tendency to shoot too many long two-pointers when he faces up, but he’s been relentless in attacking the offensive glass, creating a number of attempts around the basket. In fact, his percentage of shots at the rim (per Hoopdata.com) is up from 33.0 percent of his tries as a rookie to 53.8 percent this season.
The problem is Cousins simply isn’t finishing. His accuracy at the rim, from the same source, is down from making 62.4 percent of his close-in shots to 42.1 percent this season. Watching Cousins’ shots, a couple of factors are at play here. For one, Cousins does tend to get too cute around the basket, whether because he’s fearful of getting his shot blocked or just prefers a little extra degree of difficulty. Given that, Cousins has still had an inordinate number of shots bounce in and out so far. I’d guess this has happened at least a dozen times on his 70 shot attempts.
Over the course of the season, such luck will even out, but 70 shots isn’t nearly enough of a sample to make it reliable. Right now, his shooting percentage at the rim is mostly attributable to noise. It’s one thing to understand that intuitively, and another to see it in action when the ball bounces out.
Sometimes, the most important understanding of statistics is to know when they aren’t telling us anything meaningful. It’s tempting to find patterns amidst the chaos. The studies cited in Thinking, Fast and Slow indicate that it’s basically human nature to attribute to much of performance to an individual and too little to outside factors. That’s why it’s necessary to keep in mind the importance of sample size this early in the season, and sometimes even mentioning it isn’t enough.
The latter issue is why I felt the great Henry Abbott’s post on TrueHoop on Monday comparing True Shooting Percentages and shot attempts during the first week of the season was unfair. Henry was sure to mention multiple times that it’s early, but still singled out players (including Cousins) as hurting their team by shooting so much. Especially among readers who aren’t as familiar with the variability of statistics, that’s dangerous, and I think it paints statistical analysis in a bad light. Unfortunately, when those readers consider numbers that don’t match up with what they see on the court, they’re likely to determine that the problem is with the statistics and not how they’re being used.
Their conclusion? Statistics lie.
(Stay tuned for tomorrow’s post on DeAndre Jordan as part of De- week on Basketball Prospectus Unfiltered. Not really.)