The reigning NBA MVP may not have been as valuable as he looks

By H.H.

KEVIN DURANT got what he wanted. In his very first season after he abandoned the Oklahoma City Thunder to join the Golden State Warriors and form an unprecedented super-team, his new club romped through the playoffs, clinching its second title in three years on June 12th. The post-season was far crueller to Russell Westbrook (pictured), the superstar teammate Mr Durant left behind in Oklahoma. After the Thunder slumped to a good-not-great won-lost record of 47-35 in Mr Durant’s absence, they were eliminated in the first round of the playoffs.

Nonetheless, two weeks after the 2016-17 campaign came to an end, it was Mr Westbrook that had his day in the sun. On June 26th North America’s National Basketball Association (NBA) named him its Most Valuable Player (MVP). No matter how modest his club’s accomplishments, the MVP is an individual award. And on the surface, it was hard to argue with the selection: Mr Westbrook had achieved a statistical benchmark accomplished only once before in the league’s history, by averaging a triple-double for an entire season.

The triple-double—attaining double-digit totals in three separate statistical categories in a single game—is basketball’s standard benchmark of virtuoso versatility. In its standard form of points, rebounds and assists (versions including blocks or steals are extremely rare), it requires an athlete both to create at least ten successful scoring opportunities for his teammates—a skill generally associated with point guards, the position occupied by the sport’s shortest players—and to secure possession of the ball following a missed shot at least ten times, a play that usually requires the height and strength of power forwards or centres. (The third plank, scoring at least ten points, is far easier to obtain.) Although triple-doubles are observed with some regularity—since the 1983-84 season, they have occurred once every 26.5 regular-season games, or roughly twice a week—they are still rare enough for fans to consider witnessing one a highlight, and for the player who accomplishes the feat to be the unquestioned star of the game. During the past 34 years, teams with a player who racks up a triple-double have won three-quarters of the time.

A handful of players in NBA history have specialised in triple-doubles. They have generally been rare and coveted “point forwards” such as Grant Hill or LeBron James, who have both the size to “crash the boards” for rebounds and the dribbling and passing ability to serve as their team’s primary playmaker. Earvin “Magic” Johnson, a six-foot, nine-inch (2.06-metre) tall point guard, became the personification of the feat in the 1980s, and still ranks second all-time with 138 triple-doubles; Jason Kidd, a more classic pass-first point guard, was an unusually fine rebounder for a 6’4” player, and amassed 107 between 1994 and 2013. But the unquestioned king of the triple-double is Oscar “The Big O” Robertson, a 6’5” point guard from the 1960s. For Mr Robertson, triple-doubles were the norm, not the exception: he accumulated 181 of them over his career, the most in history by a large margin. And in his magical 1961-62 campaign, he posted per-game averages of 30.8 points, 12.5 rebounds and 11.4 assists over the course of an entire season.

Enter Mr Westbrook. During the last year he played alongside Mr Durant, who has a strong claim to be the finest scorer in the NBA, he achieved marks of 23.5 points, 10.4 assists and 7.8 rebounds per game—a superlative season, but still a daunting 2.2 rebounds per contest short of matching Mr Robertson’s record. Moreover, following Mr Durant’s departure, the Thunder were left without a serious scoring threat save Mr Westbrook himself. As a result, Mr Westbrook had to shoulder an unprecedented share of Oklahoma City’s offensive load: 42% of his team’s possessions ended with the ball in his hands, exceeding both the previous record of 39% set by Kobe Bryant in 2005-06 and Michael Jordan’s high-water mark of 38% in 1986-87.

While sure to inflate Mr Westbrook’s scoring, that increased responsibility could easily have reduced his output of triple-doubles. Because assists are only granted when a teammate successfully makes a basket, a player who shoots more frequently and has weaker offensive talent around him is less likely to accumulate them. Similarly, the more a player shoots, the fewer opportunities he has to box out opposing defenders to corral offensive rebounds.

Mr Westbrook was undeterred. As expected, his scoring output in 2016-17 soared to 31.6 points per game, making him the 31st player in history to exceed the 30-a-night barrier. But his assists also held steady despite Mr Durant’s departure, at 10.4 per contest, and his rebounding vastly increased as well to 10.7. Only one player 6’3” (Mr Westbrook’s comparatively modest height) or under, Lafayette “Fat” Lever, had ever averaged even eight rebounds per game; Mr Lever′s highest mark was 9.3.

Triple-double trouble

After matching an achievement most analysts assumed would never be equalled, Mr Westbrook’s MVP award was a foregone conclusion. But determining whether he was in fact the most valuable player in the league—and whether triple-doubles are even a strong proxy for producing wins and avoiding losses—requires further study.

The most sophisticated statistical techniques currently available for evaluating NBA performance are not kind to Mr Westbrook. According to Regularised Adjusted Plus-Minus (RAPM)—a measure developed by Jeremias Engelmann of ESPN, which compares a team’s scoring margin when a player is on the court versus when he sits on the bench, and adjusts for the quality of everyone else playing at the same time—Mr Westbrook’s 2016-17 was in fact rather humdrum. It found that the Thunder scored 2.7 more points per 100 possessions when Mr Westbrook was on the court than they would have with an average player in his stead, and allowed 0.3 fewer. The combined impact of 3.0 placed him just 26th in the league.

It’s hard to reconcile such a modest mark with either Mr Westbrook’s apparent domination of virtually every contest he played in, or with the Thunder’s performance as a team. Oklahoma City gave him almost no supporting cast: Steven Adams is a fine defensive centre, and Enes Kanter a solid scoring one, but that’s about all that can be said for the rest of the Thunder’s roster. If Mr Westbrook were really not even a top-25 player, the Thunder would probably have had a losing record rather than being a low-level playoff club.

At the same time, if Mr Westbrook’s season were in fact the best campaign in the history of the NBA by a huge margin—as his surface statistics indicate—then Oklahoma City should have been a title contender even without giving him much help. After all, Mr James led the 2006-07 Cleveland Cavaliers to the Finals, and the 2008-09 version of the team to the NBA’s best record, without the help of a second star. This principle is not ironclad: the Chicago Bulls somehow managed to post a losing record in Mr Jordan′s highest-scoring season. Nonetheless, in a sport as superstar-driven as basketball, there is good reason to impute the bulk of a great team’s achievements to its strongest contributor—and, conversely, the blame for a middling club′s lack of accomplishments to its best player.

So why might Mr Westbrook’s barrage of triple-doubles be less valuable than it appears? Two immediate explanations stand out. The most obvious is that half of basketball is defence, which triple-doubles do not measure. If Mr Westbrook were an utter sieve allowing opposing scorers to run rampant—like, say, Kyrie Irving, the current Robin to Mr James’s Batman on Cleveland—that would account for much of the gap. However, his steals rate—which generally serves as a strong proxy for a guard’s defensive prowess—was above-average. And RAPM found him to be a slightly above-average defender as well.

A second reason is that triple-doubles aren’t what they used to be. As the NBA has adopted a more up-tempo game, and developed offensive strategies that allow versatile players to make better use of their range of skills, triple-doubles have become far more common. In 2016-17, one was seen every 10.5 games, making them 2.5 times as frequent as their long-run average. If not for Mr Westbrook’s all-time record 42 triple-doubles, James Harden of the Houston Rockets, who amassed 22, would have set the single-season high excluding Mr Robertson. A rising tide lifts all boats, and if everyone is compiling more triple-doubles, today’s league leader in the category may not add more wins to his club’s ledger than the statistical leaders of yesteryear did.

There may also be a more subtle explanation: as Werner Heisenberg would remind us, it may be impossible to know a player’s “true” proclivity for triple-doubles, because the act of counting them may influence how often they occur. NBA players are well-aware that amassing a triple-double is one of the best ways to ensure they lead the nightly sports highlights, increasing their marketability and potentially their compensation. As a result, a player who is one or two rebounds or assists short of the feat as the clock winds down, and sees “23-9-12” when he gazes up at the box-score screens suspended above halfcourt, may alter his behaviour in order to surpass the round-number threshold.

This “stat-padding” phenomenon has already been observed in single-season win totals for pitchers in Major League Baseball. Winning games is difficult, and winning more games is harder than winning fewer ones. As a result, in a typical year, there are more pitchers with five wins than with ten, and more with ten than with 15. In fact, the trend that higher-win seasons are rarer than lower-win ones holds true for every integer between 0 and 30, with just one exception: there have been more 20-win seasons, which just so happens to be the standard benchmark of an outstanding campaign, than 19-win ones. This anomaly strongly suggests that pitchers sitting on 19 wins give more effort, that their managers let them stay in games longer, or both, in order to maximise their odds of getting to 20.

The evidence, your honour

A similar method can be deployed for triple-doubles in basketball. Since 1983-84, there have been just under 350,000 player-games in which a player has scored at least ten points. In these contests, the distributions of rebounds and assists follow a smooth distribution. Between four and 13 assists, the ratio of the number of games observed at each step to the one before declines gently from 73% to 69%. Similarly, the ratio also falls gradually between seven and 16 rebounds, from 85% to 70%.

Games in which a player reached double digits in both points and rebounds are significantly less common (57,106 player-games), and rarer still are double-doubles in points and assists (15,939). In these contests as well, the distributions in the third category generally behave as expected. Following peaks at three rebounds and one assist, within the range of commonly observed totals, each higher single-game total is less common than the one before.

However, just as with pitcher wins in baseball, there is one striking exception. Among players with double-digit points and assists, there were more games with exactly ten rebounds (514) than nine (429, see chart). That strongly suggests that players on the brink of a triple-double sag off the players they are guarding on defence, so that they can be better-positioned to grab a rebound in the event of a miss, and that their teammates probably clear out of the way to ensure that any “discretionary” rebounds wind up in the hands of the desired player.

The other permutation—examining the frequencies of assists when a player has double-digit points and rebounds—does not yield quite the same smoking gun. As one would expect, ten-dime games are less common than nine-assist ones, just as is true at every other point on the right side of the distribution. However, the difference between the two figures is so small that it suggests rampant stat-padding here as well. Among players already in double figures in the other two categories, every step from three assists to nine is between 59% and 66% as common as the one before it. However, ten-assist games are a whopping 97% as frequent as nine-assist ones. It is hard to account for this discrepancy without suspecting that players who are one assist short of a triple-double give up good shot opportunities, in order to pass the ball to a teammate and hope that he scores.

These anomalies have a surprisingly large effect on the frequency of triple-doubles: if the distributions were smooth rather than bulging at ten, the feat would be around one-eighth less common. And the more times a player approaches triple-double territory, the more often he is likely to modify his play to his stat sheet’s benefit and his team’s detriment. Of the 81 regular-season games Mr Westbrook played this year, there were 19 in which he wound up with either exactly nine or ten rebounds or assists and at least nine of the other category—creating an incentive to play for the box score rather than the win.

On June 30th the Thunder acquiredPaul George, a star small forward. Though not quite Mr Durant’s equal, Mr George will go a long way towards reinstating the status quo from before last season, when Oklahoma City was a formidable two-headed dragon rather than an overstretched one-man band. It will be no surprise if, with Mr George at his side, Mr Westbrook finds himself facing off against Mr Durant in next year’s Western Conference finals. If the addition of Mr George reduces Mr Westbrook’s triple-double output and reduces the perceived reward from padding his statistics, that might be an underappreciated reason for the reconstituted team’s expected success.