The 4 Fallacies of NBA Analytics

Credit: Ezra Shaw / Getty Images

In recent years, basketball fans and writers have shed heavy pixels and ink on the ascendancy of geeks over jocks in pro sports. Heavy number crunching – complete with new stats and endless logarithms – have taken over the games we love.

And yet, according to one of the thinkers who sparked the statistical revolution, most NBA teams still don't know squat about how to win with numbers. David Berri co-authored his now-classic Wages of Wins: Taking Measure of the Many Myths in Modern Sport nearly a decade ago. His work has focused on understanding which statistics most impact the desired outcome: winning.

"Wins in the NBA are determined by possession factors," Berri says. In other words, getting possessions by rebounding and keeping them by not turning the ball over. "And then your ability to turn the possession into points: your shooting efficiency." Ergo, putting the ball through the hoop without missing many shots.

Reasonable enough. But decision-makers aren't always reasonable. Here are four mistakes you can count on NBA general managers making every year:

Discarding the Numbers
NBA team managers are vaguely aware of the importance of shooting efficiency. But, "Scoring still dominates the free agent decision," Berri says. That's why current MVP Stephen Curry lost out to born-to-chuck guard Tyreke Evans for Rookie of the Year in 2010. It's also why the Lakers pay Kobe Bryant $23.5 million to score 23 points a game, while missing two out of every three shots he makes – a far below average ratio. And since managers don't care about most of the numbers beyond scoring, it has become a common practice for NBA teams to hire bloggers with no formal statistical training as their number crunchers – and promptly ignored their advice.

Measuring the Wrong Things
Basketball generates a crazy amount of data. There are 80 or 90 possessions a game, per team, with numerous possible outcomes for each. The SportVU system records 72,000 images per game of every dribble, bounce pass, miss, and make. "In order for that SportVU stuff to matter," Berri says, "you have to establish what you collected actually helps me explain outcomes. Somebody was telling [Detroit Pistons head coach] Stan Van Gundy that they were tracking how far somebody runs during a game and he asked, 'How can that help me?'" Berri laughs hysterically. With so much (mostly useless) data, management falls back on the good ol' eye test. When you watch a game, you remember Carmelo scoring buckets. Less memorable are his 12 misses or 5 turnovers.

Analyzing the Wrong Things
With so many variables, teams simplify by jamming multiple stats into a single super stat. The most famous is John Hollinger's Player Efficiency Rating, or PER. "You can increase your PER just by taking more shots," Berri says, even when they're bricks. "Kobe Bryant is an above average player this year according to PER, but he's really awful. He can't shoot." The problem is the various stats in PER haven't been correlated with actual wins. "What is the impact of each of the statistics on outcomes, the value of a rebound, a steal, an assist?" That is the formulaic work of a trained statistician, a genuine geek.

Paying for Reputation
NBA teams pay premiums for players who were drafted high, even if their professional performance is terrible. "The Knicks really illustrate how much people get it wrong," Berri says. They give huge contracts to guys who went high in the NBA Draft, who score a lot of points on a lot of shots, missing by the dozen. But they are immune to coaching; they will always find someone else to give them a contract. On the flipside, you have successful teams like the San Antonio Spurs, and more recently the Atlanta Hawks. They hire unwanted castaways from other teams who take the shots their coaches want, or rebound or pass well.

RELATED: Stephen Curry's Guide to Three-Point Shooting

The Golden State Warriors are the exception, having drafted four important players (Curry, Thompson, Green, and Barnes) on this year's Finals squad. But they are the ultimate team players, ranking first in the playoffs in assisters per game. GM Bob Meyers has assembled a team that, based on its ability to outscore opponents, is, according to sports stat guru Nate Silver, the third best team in NBA history. The secret is simple: They are good at maintaining and efficiently using possessions – every Warrior getting playoff minutes rebounds, passes, and shoots well.

If your favorite team is committing any or all of the above logical fallacies, take heart. "It shouldn't affect the enjoyment of the game," Berri says. "It irritates stats people, but that's 0.1 percent of the audience." And if you're one of those 0.1 percent, well, look at it as Berri does. The NBA makes a great laboratory, a place to study decision-making at its best, its worst, its mean, and its standard deviation.