During September, the NBA blogosphere briefly erupted in an intense debate about the value of John Hollinger's all-in-one PER metric in specific, and per-minute statistics in general. If that sounds a bit remarkable--and it ought to--keep in mind that September is an awfully quiet month in the NBA. The value of per-minute numbers is generally considered one of the primary tenets of APBRmetrics, but proving their value to critics who believe they overrate reserves has proven difficult. That didn't stop Tom Ziller from publishing a great study at BallHype.com that, to me, made a strong case for per-minute numbers.
One of the arguments statistical analysts have made in favor of per-minute numbers is the ability to identify players stuck on the bench as breakout candidates when they get more minutes. A quintessential example is Milwaukee Bucks guard Michael Redd, who backed up Ray Allen early in his career before becoming an All-Star the year after Allen was dealt to Seattle. Andrei Kirilenko and Zach Randolph are other examples of players touted by analysts well before they became mainstream stars.
How well does good per-minute performance portend future success? To test that, I turned to the "Diamond Rating," invented a few years ago by a former colleague of mine, Kevin Broom of RealGM.com. Looking to identify so-called "diamonds in the rough," Broom devised a simple general formula that works with any per-minute rating (R in the following equation):
R*40 - R*Min/G + (R - leagueR)*40
The fewer minutes a player plays per game, and the better he does in those minutes, the better his Diamond Rating. Unfortunately, the Diamond Rating doesn't always identify breakout candidates. Dikembe Mutombo, for example, typically posts a very good Diamond Rating. At his age, which unofficially might threaten Julio Franco's, Mutombo is believed to be best suited for a limited role (you might argue otherwise, based on his impressive stint in place of an injured Yao Ming last season). Calling him a breakout candidate would be silly. There are also some players who are simply so good they show up on the list despite playing starters' minutes. Tim Duncan is a fixture, both because of his MVP-caliber play and because Gregg Popovich likes to keep him fresh for the postseason.
As a result, it makes sense to eliminate players from consideration who are over 27, have more than five years experience, play more than 30 minutes per game or played less than 250 minutes total (too small a sample size to be reliable). Using those rules and the per-minute incarnation of my WARP rating, I went back through the top five breakout candidates as identified by Diamond Rating from 2001-02 through 2005-06 and evaluated how well those lists actually projected future performance.
Player Tm Dia Yr0 Yr1 Yr2 Yr3
Chris Andersen DEN 18.4 10.9 15.4 14.5 21.3
Jamaal Magloire CHA 16.7 18.9 29.8 33.9 30.6
Etan Thomas WAS 16.1 13.1 13.5 24.1 20.8
Rafer Alston MIL 15.5 12.0 20.9 31.5 34.0
Andrei Kirilenko UTA 14.2 26.2 27.7 37.1 32.9
(Yr0 is minutes per game in the season in question; Yr1 is MPG the following year and so on.)
Hits: Magloire's position on this list is explained by fluky 55.1% shooting from the field (he hasn't been over 50.4% since) and the presence of veteran starter Elden Campbell. Magloire supplanted Campbell to start all 82 games in 2002-03. Though he's never been as good on a per-minute basis as in 2001-02, Magloire was an All-Star in 2004 before declining mobility relegated him once again to being a role player. Alston, too, quickly made good on his promise, becoming a starter in 2003-04.
As mentioned above, Kirilenko (along with Redd, who finished sixth this season in Diamond Rating) is one of the poster boys for the use of per-minute statistics. Kirilenko was already playing regular minutes, but he joined Magloire as a 2004 All-Star after John Stockton and Karl Malone departed, giving Kirilenko a starring role in Utah.
Mixed results: One of my favorite NBA players before being disqualified from the league in January 2006 for violating the league's anti-drug policy (he's eligible to apply for reinstatement next January), Andersen was always close to or on this list. Andersen's prodigious per-minute shot-blocking showed him worthy of more minutes even when he saw limited action as an undrafted rookie. The same is true of Thomas, though neither player has become more than a role player.
Player Tm Dia Yr0 Yr1 Yr2 Yr3
Etan Thomas WAS 17.3 13.5 24.1 20.8 15.8
Zach Randolph POR 15.1 16.9 37.9 34.8 34.4
Jeff Foster IND 14.6 10.6 23.9 26.1 25.1
Andrei Kirilenko UTA 13.9 27.7 37.1 32.9 37.7
Jerome Moiso NOH 13.8 12.6 11.9 6.0 -
Hits: I move Thomas to the hit category because his "breakout" was much more immediate in 2003-04, when he nearly doubled his minutes per game and averaged 8.9 points, 6.7 rebounds and 1.6 blocks, all career highs. Randolph more than doubled his minutes per game while slightly improving his per-minute scoring and rebounding to win Most Improved Player honors. Randolph has gone from being underrated in the eyes of statistical analysts to overrated, given his typically indifferent defense.
Mixed results: Foster is a little bit of an oddball; though he was young and it was still relatively early in his career, he had played 20+ minutes per game before seeing his role shrink during Brad Miller's lone full season in Indiana. Miller left for Sacramento as a free agent at season's end, and Foster's minutes bounced back as a result.
Misses: A lottery pick by Boston in 2001, the athletic Moiso played just 211 minutes in his first two seasons before enjoying a solid campaign off the bench in 2002-03 in New Orleans. Moiso shot 52.0% and posted solid per-minute rebound and block rates. He played decently, though less often, the following year in Toronto before heading overseas in 2005.
Player Tm Dia Yr0 Yr1 Yr2 Yr3
Mike Sweetney NYK 19.2 11.8 19.6 18.5 8.0
Chris Andersen DEN 18.2 14.5 21.3 17.8 -
Dan Gadzuric MIL 17.6 16.8 22.0 12.0 15.6
David West NOK 17.1 13.1 18.4 34.1 36.5
Stromile Swift MEM 16.7 19.8 21.3 20.4 19.1
Hits: West's chance to break out was postponed by a bone bruise in his knee that limited him to 30 games in 2004-05. He replaced the traded Magloire in the starting lineup in 2005-06, and averaged 17.1 points and 7.4 rebounds, finishing second in Most Improved Player voting.
Mixed results: I'm tempted to put Sweetney in the "miss" category after his dismal 2006-07 season in Chicago, but he did successfully ramp up his minutes in his second season. Still, those of us chanting "Free Mike Sweetney" a couple of years ago--and I was as vocal as anyone--have since been quieted. Andersen played a career-high 21.3 minutes per game in 2004-05, and that would probably have been a high-water mark even without his suspension.
Misses: Gadzuric had his best year in 2004-05, averaging 7.3 points, 8.3 rebounds and 1.3 blocks per game. The Bucks showed little faith in him, however, trading for Magloire--who is getting a stunning amount of attention here--after drafting Andrew Bogut with the top overall pick. His role has decreased the last two seasons. Swift, meanwhile, is a good example of the limitations of per-minute stats. A long-time favorite of analysts, Swift's playing time has always been limited by his frequent mistakes. His plus-minus numbers are not as strong as his individual statistics, and several coaches have been unwilling to play Swift much more than 20 minutes per game.
Player Tm Dia Yr0 Yr1 Yr2
Anderson Varejao CLE 18.7 16.0 15.8 23.9
Andris Biedrins GSW 18.0 12.8 14.7 29.0
Dan Gadzuric MIL 17.4 22.0 12.0 15.6
Zarko Cabarkapa GSW 15.7 11.9 8.3 -
Al Jefferson BOS 15.3 14.8 18.0 33.6
Hits: It took a year for Biedrins, just 18 when he came to the NBA, to make his presence felt. In his third season last year, he emerged as the perfect big man for Don Nelson's run-and-gun style, averaging 9.5 points, 9.3 rebounds and 1.7 blocks per game, and shooting 59.9% from the field. Jefferson, too, made 2006-07 his breakout season, averaging 16.0 points, 11.0 rebounds and 1.5 blocks per game. This summer, he was the centerpiece of Minnesota's return for Kevin Garnett.
Mixed results: Varejao still has time to put himself into the "hit" category after playing a career-high 23.9 minutes per game last season. He might have seen a larger role in 2005-06 were it not for a shoulder injury that kept him out until mid-January. Now a restricted free agent, Varejao has yet to agree on a new contract with the Cavaliers. A change of scenery might give him the chance to break out that will be difficult in Cleveland with Drew Gooden and Zydrunas Ilgauskas as the starting frontcourt.
Misses: Even while playing the most minutes of his career, Gadzuric still landed on this list. As discussed previously, his role has diminished the last two seasons. Phoenix picked Cabarkapa 17th overall in the 2003 NBA Draft, but a solid season in limited minutes off the bench with the Warriors has proven to be the high point of his NBA career thus far; he missed all of 2006-07 with a back injury that eventually required surgery, and he did not appear in a training camp this fall.
Player Tm Dia Yr0 Yr1
Chuck Hayes HOU 21.7 13.4 22.1
Darko Milicic ORL 15.7 13.9 23.9
Shavlik Randolph PHI 15.5 8.5 13.8
Anderson Varejao CLE 14.8 15.8 23.9
Hits: After catching on with Houston as an undrafted rookie, Hayes was incredible as a rebounder and defender off the bench. Last year, Hayes split time at power forward with Juwan Howard (averaging 28.1 minutes per game in the Rockets' first-round series with Utah) and continued to be very effective, shooting 57.3% from the field. The addition of Luis Scola will limit Hayes' minutes going forward, but that shouldn't be considered a knock on his performance.
Mixed results: Milicic goes in this category because he had already played 20 minutes a game after being traded to Orlando midway through the 2005-06 season, meaning he didn't dramatically change his role last year. He steps into a starting spot in Memphis after getting a lucrative deal as a free agent. Now we'll find out what kind of player Milicic can be. It's much too early to tell with Randolph, who was playing well--though not that many additional minutes--early in 2006-07 before a gruesome ankle injury ended his season in late November. Back on the court, Randolph figures to play a larger role this season.
Misses: This assessment might be a little harsh for Arroyo, but he actually played slightly fewer minutes than he did after coming to Orlando with Milicic midway through 2005-06. The rest of that season, Arroyo played the best basketball of his career. He's talented enough to have started 71 games for Utah a few years back, but that half-season with the Magic is the only other time Arroyo has really put it all together.
Of the 25 players I looked at, I consider 11 hits, six misses and eight as having mixed results. With players like Kirilenko, Randolph and West, analysts using per-40-minutes stats were ahead of the curve. For the most part, the misses were not players as loudly touted by APBRmetricians, though Swift and Gadzuric had considerable support in this crowd.
Overall, the group of players averaged 15.3 minutes per game during the season that landed them on this list. That increased to 20.6 mpg the following season, then jumped up again to 24.6 mpg the year after that. In general, a high Diamond Rating is a good sign a young player will be moving into a larger role in the near future.
I looked at a few other markers to try to see how the Diamond Rating could be augmented for better prediction in the future. It should be no surprise that the younger and more inexperienced the player, the stronger an indicator Diamond Rating proved to be. Six of the 10 players 22 or younger were hits, with no misses. Meanwhile, there were more misses (three) than hits (two) amongst the seven players 25 or older. A similar trend is evident with experience in the NBA.
Other factors don't seem to provide much help. Minutes per game played by the player in the season that landed them on this list did not correlate strongly with eventual success or failure. I also looked at draft position to see if that conveyed some information about whether the player's performance was likely to be a fluke (inspired by someone who just missed this list, 2003-04 undrafted rookie Marquis Daniels). In an odd result, every group (lottery picks, second-round picks, undrafted players) had as many misses as hits except non-lottery first-round picks, who featured six hits and just one miss.
Check back Thursday for a list of the top-rated players by the Diamond method in 2006-07, and a look at which of them are likely to break out.
Kevin Pelton is an author of Basketball Prospectus.
You can contact Kevin by clicking here or click here to see Kevin's other articles.