To many, Dampier had become the face of the new, starless Warriors: hardworking, unspectacular, overachieving -- "blue collar," as they say in the sports world. On Jan. 28, the day before coaches were to receive their All-Star ballots, and just hours before the Warriors' game against Minnesota, Musselman phoned some of his Western Conference colleagues and lobbied on Dampier's behalf for their All-Star votes. That evening, Dampier made a sort of closing argument: He scored 21 points, grabbed 19 rebounds, and blocked five shots in the Warriors' upset victory over a Timberwolves team that has become one of the NBA's best. "Hopefully," Dampier told reporters after the game, "this was a good impression on the coaches."
The coaches may have been impressed, but not Roland Beech, a skinny 35-year-old who spends his days staring at arcane basketball statistics in Excel spreadsheets. Beech, born to English parents, is a small, unassuming guy with a narrow face, a poof of light brown hair, and a hint of a British accent. He works on the second floor of a roomy house in unincorporated Aptos, near Santa Cruz. His office is bare -- a futon, a TV, a desk, a gurgling aquarium -- but it's likely there is as much quantified information about the National Basketball Association here, on a small computer humming beneath a kid's crayon drawing, as anywhere outside the league itself. This is the home of Beech's NBA Web site, 82games.com, which tracks the season in obsessive statistical detail, down to the kind of abstruse metrics that pro basketball has thus far overlooked, underused, or entirely ignored.
Right now, Beech's spreadsheets tell him that Dampier, touted all year as one of the NBA's most improved players, has actually been a liability. "Basically," Beech says one afternoon in January, clicking around on his computer and bringing up a page of Dampier's numbers, "the team is worse offensively when he's on the floor. It's pretty dramatic. They're about a tenth of a point per possession worse. Field goal percentage is down a bit. Rebounding is a wash.
"It's just strange. He's one of these guys -- and you find them sprinkled around -- who would seemingly be positive players, but for some reason aren't."
Just how Beech perceives this hidden impact could change the way basketball is watched, coached, and even played.
Beech's site went online last fall, an auspicious time to set out on a project of this sort. It was, after all, the year that geeks and jocks finally shook hands, and the sports world tiptoed into a kind of age of reason. In 2003, statheads saw their patron saint, Bill James, baseball's ultimate outsider, working as a senior adviser to the Boston Red Sox, conceiving studies on things like the effectiveness of tall left-handed pitchers at Fenway Park. And they found all their shibboleths between the covers of a best seller -- Moneyball, about the Oakland A's, a team built on the spreadsheets of Harvard grads.
The geeks have yet to find the same kind of foothold in basketball, a relatively young professional sport that hasn't generated a century's worth of statistics. Nor have they established the kind of large community that sustained baseball research for years on the margin of the sport; the Association for Professional Basketball Research wasn't even created until 1997. Twenty years ago, Bill James famously wrote that "baseball statistics, unlike the statistics in any other area, have acquired the powers of language," a verdict that still holds. Partly, that's due to the nature of the game. In baseball, the most important interaction, pitcher versus hitter, is one-on-one, with solid statistics to express the outcome. In basketball, though, a jump shot from 15 feet could be the result of a pick-and-roll, a kick-out from the post, a crisp pass from the perimeter -- a complex interaction among teammates, in other words, reduced to two points and an assist in the next day's box score. "The players are so tied together," Beech says, "that there are a lot of statistics not reflected in the personal stats."
82games.com sprang from that idea -- that a player's performance is intimately connected with the team context. When he began work on the site, Beech was drawn to two key measures of a basketball player's effectiveness: plus-minus and on-court/off-court. The first stat tracks the team's performance when a player is on the court. That's to say, if Erick Dampier were to enter the game with the Warriors trailing by four and go to the bench with the Warriors trailing by 10, the team would have ceded six points to its opponent and thus Dampier's plus-minus for that time would be -6. Similarly, on-court/off-court breaks down a team's statistics -- how it shoots, defends, rebounds, and so on -- with and without a certain player in the lineup. Both stats have their flaws, which Beech readily acknowledges (and would like to correct). A good point guard with a strong backup might look worse than an average point guard who's playing ahead of a developmental-league call-up from Fayetteville. But as rough indexes of a player's effectiveness, plus-minus and on-court/off-court are often illuminating.
"I wanted to understand a bit more about the game myself, to figure out why some players are effective and not effective," says Beech, who uses several computer programs to mine statistics from lengthy, play-by-play game logs. It's one of the strengths of his site that he poses, then goes about answering with empirical evidence, simple, almost naive-sounding questions about the sport. From time to time, Beech will post a report, scattered with exclamation points, under a title like "The value of an offensive rebound" or "The value of a blocked shot." He has a very basic curiosity about the game, and when he talks basketball, he seems to generate ideas for studies midconversation, even midsentence, a kind of mental give-and-go.
Beech has been doing this sort of thing all his life. Born in New York state and raised in England and the Bay Area -- his father was a mathematician and later a database architect at Oracle -- Beech was always good with numbers, reading box scores at age 8 and playing stratomatic basketball, a stats-based board game using real teams and players, soon after. He'd play with his friends, who would pick the "showtime" Los Angeles Lakers; he'd take the Warriors. (Even today, he names Joe Barry Carroll -- Golden State's No. 1 draft pick in 1980 and perhaps the all-time most reviled Warrior -- as one of his favorite players. "Most people think he was a huge underachiever," Beech says, "but I liked his game. He was a good scorer, and he rebounded OK. I kind of like the guys who have bad reputations, the guys who are ragged on a little bit. JB was essentially a decent player, but everybody expected so much more from him.") Beech didn't play sports much, but he'd go to the occasional Giants or Warriors game with his family. He remembers reading, and enjoying, one or two of Bill James' Baseball Abstracts in high school, but says he was never entirely taken with James. "I've never been interested in historical analysis," Beech says. "I think it's kind of futile to compare players from different eras, which is a driving force with a lot of baseball research, and certainly basketball, too."
Beech studied English at Berkeley, but he soon turned his knack for numbers into a career. He'd spend much of his time at Golden Gate Fields, sometimes going to the track twice a week. "That's the big numbers game," Beech says. "I was interested in understanding what the numbers were about." He landed a research job with a portfolio manager who wanted to know if horse racing is a beatable game, and later he took a job with a company providing downloadable statistics for the horses, as well as for football, basketball, and baseball.
In 1997, Beech launched an NFL site, TwoMinuteWarning.com, approaching football in much the same way he now approaches basketball: as a team sport with mostly inadequate measures of actual individual contribution to the enterprise. (The research here is more oriented toward gamblers and fantasy football players; last season, Beech charged $299 for a subscription. "Pretty good income there," he says. "It's a very healthy business. Football's a betting sport. I get everybody from the guys with $20 parlays to guys betting five to six figures a week.") He decided to start 82games because, simply, he "saw it could be done."
"That's probably the excitement of basketball," Beech says. "There's so much research that can be done. Whereas something like baseball, it's a much more defined thing; so much has been done there already. But basketball is pretty wide open. You watch a basketball game, you'll see a lot of things and think, 'Well, gee, I've never seen statistics on that.'"
Warriors center Erick Dampier, left off the NBA All-Star team this year, offers a good case study in the Beech method. By most conventional measures he is a valuable component of the Warriors' lineup; by some measures, he should have been an All-Star; by Beech's, though, he is trade bait. Sometimes, Dampier's stat lines miss the story. In the Warriors' recent 97-90 upset of Minnesota, Beech's numbers show, Dampier didn't have anywhere near the kind of impact his personal statistics in the box score -- 21 points and 19 rebounds -- would indicate. In fact, it's possible he even hurt the team in his 41 minutes on the floor, meaning that the game was won in the seven minutes he was toweling off on the bench: His plus-minus for the game was -4 -- Golden State outscored the Timberwolves by 11 when Dampier was on the bench. "He definitely has some kind of negative influence on the other players," says Beech, who wasn't surprised when Dampier's name recently came up in a trade rumor.
Beech says he's found some resistance within the league to this sort of measurement, but thinks his numbers hold up. "There's a sense that it's a little unfair, perhaps, to lump guys in with their teammates -- that you don't want to judge [Orlando star] Tracy McGrady by the rest of his teammates right now, because he doesn't have any good teammates," he says. "But if you actually go to his numbers, yeah, he is a big positive influence on his teammates. The numbers seem to work, to my mind. If they explore them a little bit, they'll find that."
The NBA -- which for years didn't even track things like minutes played, blocked shots, offensive and defensive rebounds, steals, turnovers -- generally has been slow to embrace the work of people like Beech. It's a league still run on gut feeling, and the "brains" feel left out. "I don't think they know we exist," says John Hollinger, who recently published his second Pro Basketball Prospectus, a book of statistical analysis. "They think this stuff is half a notch above witchcraft." Adds a slightly more put-upon colleague: "It's like anybody else who knows something. You know Galileo? Did this guy know what he was doing? And what happened to Galileo? They killed him." Beech has made some inroads into the NBA culture: He speaks with assistants and scouts, not to mention Dallas' maverick owner, Mark Cuban, who complimented Beech's work. Beech also supplies statistics for Hall of Famer Rick Barry's KNBR radio show.
Hollinger estimates that basketball research is at least 20 years behind baseball research, which is understandable. The two have evolved in entirely different ways. Baseball statheads found the sport's holy grail -- on-base percentage, the determining factor in how many runs a team will score -- and much of the research spun off from there. It's a measure of how far baseball research has come that it has its own common noun: sabermetrics, derived from the acronym for the Society for American Baseball Research. Now even Bill James, supreme objectivist, is apparently confident enough in his stature as a baseball thinker that he has added a subjective element to his new player ratings.
But basketball has no holy grail. In his recent book, Basketball on Paper, Dean Oliver writes: "[T]he ideal player rating statistic is just not possible." Unless, he goes on, "you can simulate a player with all possible combinations of coaches, and all possible combinations of strategies against all possible combinations of opposing teams ...."
For a brief moment, in the mid-'90s, the sport actually moved forward in the field of statistical analysis, with a data-mining program called Advanced Scout, developed by Inderpal Bhandari at IBM, then an NBA sponsor. "I wanted to develop a data-mining program that exemplified the use of data mining as a mechanism to enable ordinary people, average decision-makers, to deal with this tremendous glut of data," says Bhandari, now CEO of a New York-based data-mining company called Virtual Gold. "Basketball coaches fit that mold very well."
Working with Tom Sterner, then an Orlando Magic assistant coach and now the Warriors' top assistant, Bhandari created a program that could dig through reams of statistics and unearth patterns that stray from the norm. Soon much of the league was using, and trusting, Advanced Scout. Both Sterner and Bhandari tell the story of the Miami-Orlando series in the 1997 playoffs. The Magic trailed 2-0 in the five-game series, having lost the two games by a combined 52 points. Desperate, the coaches turned to Advanced Scout, which found that Orlando actually outscored the Heat when backup point guard Darrell Armstrong was on the floor. Armstrong was given a more central role -- 21 points and eight assists in Game 3; 12 points and nine rebounds in Game 4 -- and the Magic fell just short of winning the series.
But IBM pulled the plug on its NBA sponsorship deal a few years ago; now teams have to pay Virtual Gold $10,000 for the program, according to Sterner. Not many do. (Bhandari acknowledges the drop-off, but declines to say how many teams currently use Advanced Scout.) "The hardest thing in the world is trying to convince somebody to spend 10 grand on a stat package," Sterner says. "It's hard for people to conceive that that's an important element. And to the people who utilize this information and see the value in it, it's mind-boggling to them to think we'll spend $40 million on a player, but we won't pay 10 grand to decide whether we got the right guy."
Several games into the season, with the team playing surprisingly well, the Warriors staff began sniffing around for a free copy of Advanced Scout. Golden State's roster had been overhauled in the off-season -- the Warriors lost Gilbert Arenas, a point guard thought to be a budding star, to free agency and traded away their leading scorer, Antawn Jamison, in a nine-player deal that brought them a moody, explosive point guard in Nick Van Exel -- and again the franchise found itself in what has become its standard predicament, stuck somewhere between a renovation and a playoff run. When the Warriors went 8-7 to open the season, their best start in 10 years, there was even some cautious optimism. Still, the new-look Warriors were just getting used to each other. Around Thanksgiving, the coaches, struggling to get a feel for the Warriors' rotation, wanted to see how their different lineups were doing. Which groups were scoring well? Which were defending? Which were rebounding? All things Advanced Scout could provide.
The Warriors never found a free copy. ("I would spend the 10 grand," Sterner says, "but I'm not the person making the decision." In the NBA, it should be noted, a general manager can find $10,000 between the cushions of his couch.) Instead, someone tipped off head coach Eric Musselman to Beech's site, which is now bookmarked in Sterner's browser. "Muss said, 'Check this out,'" the Warriors assistant recalls. "We all went on, started looking, and said, 'Aw, shit, this is nice.' It's the closest thing to Advanced Scout that's out there." What the coaches quickly noticed was the impact of backup forward Brian Cardinal, the last player invited to Golden State's training camp this season. Cardinal's plus-minus was off the chart, which wasn't entirely a surprise; anyone watching the Warriors during the first month of the season could sense Cardinal was doing something right, stat lines aside.
But for the coaches, Beech -- who hasn't actually met any of the Warriors staff -- had put a number on what, to that point, had been merely a gut feeling. "It really stood out," Sterner says one afternoon in his office, which overlooks the Warriors' practice gym. "Big time. It was so glaring it was an easy decision to make: Yeah, we've got to play this guy."
There is an egalitarian element to the stathead approach: The fat kids get to play, so to speak; the game is stripped of all distracting aesthetics, and the players are judged solely on their numbers. On the court, Cardinal doesn't look anything like an effective NBA forward. He's pale, balding, slow, decidedly earthbound, often clumsy. As a little-used reserve in Detroit, he spent so much time sliding across the floor for loose balls -- doing the dirty work and keeping the floors dust-free -- that a teammate took to calling him "The Custodian." But this year he has proved he belongs in the league; his numbers say so. "Brian Cardinal's season has been huge," Beech says. "You look at his stats -- yeah, he's shooting well and all that, but why would he make such a difference? I think it's because he's filling those kinds of areas you need -- setting good screens, hustling, energizing his own teammates, things like that."
At the very least, Beech's numbers say much more about Cardinal's play than his per game averages of 9.5 points and 4.5 rebounds. At the start of February, for instance, Cardinal ranked fifth in the league with a "Roland Rating" of +14.2 -- meaning the team was about seven points better than its opponents over 48 minutes when Cardinal was on the floor, and about seven points worse per 48 minutes when he was off the court -- which constitutes a remarkable impact for a backup forward on a so-so team. With Cardinal, the Warriors are better scorers, shooters, defenders, and rebounders: View PDF data (147kb)
"It frustrates me when they don't play Cardinal," Beech says. "He should be playing more. They should give him a chance in the starting lineup to see what happens. Maybe he's only good in his reserve role off the bench. Maybe that's his limitation. But why not toss him in there and start him?" Observation, hypothesis, experiment -- it's Beech's scientific method.
That probably won't happen, Sterner says. Musselman is known for keeping his lineups intact, and in the current climate for NBA coaches -- 14 of 15 Eastern Conference teams have changed coaches since the end of last season -- few would venture any kind of bold experiment. "It would be a big change," Sterner says. "I don't know if you do that right now. I think each coach has a philosophy that he believes in, and there's no right or wrong."
Unless you're losing a lot, it's suggested.
"Unless you're losing a lot," he says. That day, the Warriors were seven games under .500.
The Warriors play the Utah Jazz one Monday evening in Oakland, and it's possible that the most interesting thing to transpire over the course of the game is the conversation in Section 111, Row 28. There, and continuing into the parking lot afterward, Beech and Basketball on Paper author Dean Oliver touch on: why Beech thinks the time will come when he can disregard "the notion of the assist"; the most effective way to razz a free-throw shooter (according to one study at Duke University); the proper trajectory of a jump shot (according to another study); and the fact that Beech's favorite player is Joe Barry Carroll. It's a retro night at the Arena, in fact, with the teams and some of the fans in throwback jerseys, and Beech jokes that he should've worn his Joe Barry Carroll shirt. Once or twice, we get an arched eyebrow from a guy in the row ahead of us.
The Warriors wind up winning, 101-85, a rare blowout, and Dampier plays one of his best games yet, scoring 18 points, grabbing a career-high 24 rebounds, and finishing with a plus-minus of +20. Still, Beech isn't impressed with these Warriors. As he sees it, the team on the floor tonight was built old and creaky, in the hope that an experienced group would capture a low playoff seed this season and thus "breed good will in the fan base" -- essentially mortgaging the younger players' development on a long shot.
If Beech ran the Warriors, this is what he'd do: He'd start by trading the veterans, some if not all, even the productive ones -- point guard Van Exel (plus-minus, as of Feb. 6: -8.3), forward Clifford Robinson (-0.1), backup Calbert Cheaney (+7.5), and, especially, Erick Dampier (-6.2), whose value around the league is likely as high as it'll ever be and probably more than a little inflated. On Beech's team, second-year forward Mike Dunleavy and rookie Mickael Pietrus would both play 36 minutes a game to "find out what the Warriors actually have in these two guys." Beech would play backup point guard Speedy Claxton 36 minutes, too. He would play Cardinal "big minutes." He would hand the reins to acrobatic shooting guard Jason Richardson and see how he responded. He would find out just what he had to work with. He would, to put it another way, probably be fired after a week.
But never mind that. Beech grew up a Warriors fan, and it's the sad lot of a Warriors fan to watch team after team scrap their rosters and rebuild before anyone determined what was there in the first place. So this would be his big experiment, the stathead revolution brought to bear on a dire situation. It's simple, really: make a good team by making a team that actually knows itself.