Skip to main content
Original Issue

Measure of Success

Those stat guys are at it again, and now the Moneyball math of baseball has come to the NBA. Armed with dazzling equations, NBA front offices are finding entirely new ways to quantify a player's talent and judge his real value

If Dean Oliver and his peers are right, then you are wrong. Wrong if you think Michael Redd is a very good player, wrong if you think Jason Collins is a bad one and wrong if you believe Shane Battier is just another Dukie with a so-so NBA career. ¶ Oliver is a Cal Tech grad with an engineering Ph.D. who works as a paid consultant to the Seattle SuperSonics. He is also part of a small but growing movement, comprising both league insiders and outsiders, that sees its sport through a statistical prism similar to that of the young, laptop-toting generation of baseball executives made famous in Moneyball, Michael Lewis's best-selling book about Billy Beane and the Oakland A's. The teams at the forefront of the movement have hired math whizzes such as Oliver, 36, or former Rhodes scholar candidates such as Sam Presti, 29, the Spurs' assistant general manager, or Stanford MBAs such as Sam Hinkie, 27, a special assistant to Rockets G.M. Carroll Dawson. Joining forces with a burgeoning cult of independent statheads and academics, these new insiders have the same goals as their more celebrated baseball brethren: to identify, through complex statistical analysis, trends, talent and value that no one else sees. By looking deeper than traditional measures of success like ppg, rpg and FG%, they are challenging conventional NBA wisdom and changing, if at first incrementally, how players are evaluated and teams are scouted.

Take, for example, the case of Collins, the fifth-year center for the Nets. To the casual fan Collins is rather unimpressive. He rarely scores, doesn't block many shots for a center and has an embarrassing habit of laying in balls that, at 7 feet tall, he should be dunking. He is the type of player who could go his entire career and never make a SportsCenter highlight, an anonymity reinforced by his career stats (5.6 points, 4.9 rebounds, 0.6 blocks). But what if one were to dig deeper and measure other aspects of his game? The number of charges taken. The positioning on rebounds. The efficiency of picks set. The fouls not committed.

Perhaps then one would come to the same conclusion as Oliver's compatriot Dan Rosenbaum, a 35-year-old UNC Greensboro economics professor, occasional correspondent of Mavericks owner Mark Cuban and originator of a respected player-rating system. According to Rosenbaum's calculations, Collins is not a stiff at all but one of the NBA's premier defensive centers: the fourth-most effective in the league over the last three seasons, behind only Ben Wallace, Dikembe Mutombo and Theo Ratliff. The methodology is complex (box) but at its core his system measures how New Jersey performs when Collins is on the floor versus when he's off it. Think of it as basketball's version of hockey's plus-minus ratio with a few esoteric twists. The upshot: Over the last three seasons the Nets have been remarkably more effective at the defensive end with Collins in the lineup; they foul less, allow fewer free throws, rebound better and allow fewer points. "He's very consistent and consistently very good," says Rosenbaum, "meaning he's either the luckiest center alive and teams just fall apart when he's on the court, or he's doing something."

On the other hand, Rosenbaum argues that Redd is, statistically, a defensive disaster, his worst-rated two guard in the league by a wide margin. Not even the Bucks guard's scoring ability (23.0 points per game in 2004-05) can counterbalance his defensive flaws. Over the course of any given 100 possessions, the Bucks are 4.5 points worse on defense with Redd in the game--and only 2.5 points better on offense. As for Battier, by Rosenbaum's calculations he was the best defensive small forward in the league last season. Memphis was 6.3 points better (per 48 minutes) than its opponent with Battier on the floor and 4.8 points worse with him on the bench.

This approach is far from an exact science, a point that even the statheads emphatically make. For one, unlike baseball, in which individual performance can be easily isolated, the success of a basketball player is influenced by nine others. Still, coaches such as the Rockets' Jeff Van Gundy and the Spurs' Gregg Popovich and front office executives such as the Sonics' Wally Walker are keeping an open mind about their sport's new math. Says Walker, "In the bigger picture it is helpful. It does allow us to do apples-to-apples comparisons of players and combinations. Data points you can add to the old-fashioned [measures]."

Today franchises--and, for that matter, anyone with a computer--have access to countless complex statistics that are disseminated through the Internet, most notably by the game-charters at, a website that provides a staggering amount of data, sliced and diced in hundreds of different ways. This season Roland Beech, a 36-year-old suburban dad who runs the site out of his Northern California home, will have more than 100 volunteers charting games and tracking everything from contested shots to off-the-ball player movement. Not surprisingly, among the most avid visitors to the site are NBA front-office personnel, one of whom asked in a recent e-mail, "Can you add rebound of own shot percent to the rebounding stats?"

The growing appetite NBA front offices have for this outsider-generated data has, in turn, created a market for hiring these statheads on staff. They're employed largely as advisers, not decision-makers, but it's not far-fetched to think that they'll be pulling the strings in the near future. Among the most promising from this group is Celtics senior vice president for operations Daryl Morey, 31, who graduated from MIT's Sloan School of Management and considers Bill James, the patron saint of quantitative analysis in sports, to be his role model. While Morey by no means ignores points per game, rebounds per game and other statistics popularly held up as benchmarks of success, he also recognizes that those numbers can inflate (or deflate) a player's value. Instead he is constantly looking for other, more obscure indicators of success such as turnover ratios, eFG% (a weighted field goal percentage that takes into account the added value of three-pointers) and productivity per possession. Yet all of these apparent abstractions have a clear bottom line. "It's the same principle," says Morey of the comparisons with Moneyball. "Generate wins for less dollars."

That has led Morey and the Celtics to such players as Dan Dickau, whom the Celtics acquired in a sign-and-trade this summer from the Hornets for a second-round draft pick. During his first two years in the league, the 6-foot point guard was renowned more for his moppish hair than his skills. After being traded from the Mavericks to the Hornets last season, he was, for the first time in his young career, given a chance to play significant minutes, and he averaged 13.2 points and 5.2 assists. But those statistics told only part of the story. What attracted the Celtics to Dickau were some less-heralded numbers. His ratio of 4.7 assists last season for every bad pass is on par with the 4.8 average of Steve Nash, widely considered to be the game's premier pure point guard. One can reasonably surmise that playing with better players, Dickau would have had a higher ratio. This is not to suggest that Dickau is a Nash-caliber player, only that, at the price of $7.5 million over three years, Dickau might have been undervalued by the market.

The new math is not just for evaluating individual player value. It's also a useful tool in scouting team tendencies. During the postseason Oliver--who is best known for his book, Basketball on Paper, which is full of sprawling equations and includes chapters addressing such vexing questions as "The Significance of Derrick Coleman's Insignificance"--focuses on Seattle's opponents. Using a program he created called Roboscout, which draws on box scores, shot chart data from and play-by-play information, he seeks tendencies that a more traditional scout might not notice.

Last spring, for example, as the Sonics prepared to face the Spurs in the second round of the playoffs, Oliver turned up evidence that while San Antonio was a dominant defensive team, particularly in the paint, it was not bulletproof. "When you go at the midrange, there was a big hole," he explains. "Compared to the rest of the league, the Spurs are 30-35 percent less vulnerable than the rest of the league from three-point land but 30 percent more vulnerable from midrange." So, partly on Oliver's advice, the Sonics pulled up for 15- to 18-foot jumper after jumper. In the end Seattle increased its midrange shooting more than any other Spurs opponent and surprised many people by taking a superior San Antonio team to six games. "If you have a good midrange game against us, you have a better chance," confirms Spurs assistant Mike Budenholzer. "And with the Sonics, since we wanted to keep them off the three-point line, that left us weaker in the midrange game."

as one can imagine, not all basketball people buy into the concept that some geek with a computer can tell them how to play the game. Still, one doesn't find the generational divide or the hostility between traditionalist and stathead that's so pervasive in baseball. This is, in part, because the NBA numbers spit out by the computers of Oliver, Hinkie and Morey often reinforce the beliefs of old-schoolers rather than refute them. In fact, the number crunchers have found some unlikely allies within basketball's old school. Del Harris is not young (he's 68), mathematically inclined ("I can't even remember my phone number") or high-tech (rather than a tablet PC or laptop, he carries around thick blue binders of stats, marked "offense" and "defense"). Regardless, the Mavericks assistant has long been one of the coaches most open to statistical analysis, dating to his days as coach of the Rockets, Bucks and Lakers. As a result, he has cred with both crowds.

For years the Mavericks have worked with Jeff Sagarin (of Sagarin football ratings fame) and Indiana University professor Wayne Winston. The duo, who created a system called WINVAL in 2000--a precursor to Rosenbaum's adjusted plus-minus formula--sends regular updates throughout the season to Cuban, Harris and coach Avery Johnson. "Some of the conclusions," says Harris, who parses the data, "make you laugh, like when they take data from a few games and tell us one of our best defenders is actually our worst." Still, there is plenty of promising data to consider.

Last year, after Game 5 of Dallas's second-round playoff series against Phoenix, Winston sent an e-mail that broke down how different Mavs combinations fared against various Phoenix lineups. The correspondence highlighted a recurrent postseason theme. As Winston wrote, in scenario after scenario, "Daniels Stack horrible," "Daniels and Stack a disaster," "Stack and Daniels a killer." In each situation the team fared poorly--a minus-13 point differential here, a minus-15 point differential there--when Marquis Daniels and Jerry Stackhouse play together. Harris discussed the findings with Johnson, who took them into account in substitution patterns (he didn't even play Daniels in Game 6), even if they weren't easily explained. "It didn't make sense to us why," says Harris. "Both are good players, and both do well with other combinations. But together, it didn't work out."

As for the players themselves, most have no idea that they've been reduced to living, dribbling equations. Sonics forward Nick Collison, for example, is unfamiliar with the new math, even though Oliver works for his team. "I've heard about what he does, seen him at practice," says Collison, "but I'm not sure how it works." When he was informed that according to Oliver, he is one of the NBA's more effective reserves (opposing teams shot about 3% worse when the Sonics sub was in the game), Collison brightens up. "Good," he says. "Then he's a genius."


Offensive Rating= (Pts/Pos) √ó 100

Defensive Rating= (OppPts/OppPos) √ó 100


The best way to compare teams is by points per possession, rather than per game. While the top equation measures total team possessions (a stat the NBA does not keep), the bottom two calculate offensive and defensive efficiency, which is calculated as points scored and allowed per 100 possessions. Last year Ginobili's Spurs led the league with an astounding +9.05 net points per 100 possessions. (The Heat was second at +7.52.) Similar equations are used to measure individual player efficiency, in which shot attempts are used as a variable rather than possessions. Last season Ginobili ranked fourth in the league in this category, behind only Amaré Stoudemire, Damon Jones and Yao Ming.



[c1=1 if player 1 is playing at home =-1 if player 1 is playing away =0 if player 1 is not playing

ck=1 if player K is playing at home =-1 if player K is playing away =0 if player K is not playing

e=i.i.d. error term]


This equation is but one segment of a more complex formula that Rosenbaum uses to measure a player's effectiveness, or more specifically, his plus-minus (the difference between points scored for and against his team while he's on the floor and when he's off). The rating is adjusted for factors such as age, garbage time or, in this case, home court advantage, which on average skews a player's stats higher. Under the larger formula, Collins is rated as the NBA's fourth-most-effective defensive center over the last three seasons.


For extensive Roland ratings, plus the top players in other unconventional categories, go to


One of the simplest measures of a player's worth is his value to his team when he's on the court versus when he's off it. At this is called the Roland Rating, which measures a team's net points, per 48 minutes, with each player on the floor. Though Roland admits it isn't perfect--substitution patterns and the quality of a player's teammates must be taken into account--it's a good place to start.

THE BEST (2004-05)

1. Tim Duncan +16.6

2. Jason Kidd +16.0

3. Manu Ginobili +15.5

4. Dirk Nowitzki +15.3

5. Steve Nash +15.0


1. Alan Henderson -17.6

2. Leandro Barbosa -12.6

3. Rodney Buford -10.8

4. Andres Nocioni -10.6

5. Devin Harris -10.2

*Minimum 1,000 minutes









Dirk Nowitzki