My hometown NBA team, the Golden State Warriors, won more games this season than it had in the past two decades, a period lamented by Bay Area sports fans for the Warriors' immense suckiness (aside from the We Believe team, of course). And after this past 51-win season, the Warriors fired their head coach, Mark Jackson. Aside from supposed clashing with the front office, one area of speculation regarding Jackson's dismissal was the stagnation of a once exciting offense - during the 2013-2014 season, the Warriors' offense was decidedly mediocre, ranking 12th in efficiency. Jackson ran isolation-heavy offensive sets for either Stephen Curry or David Lee, and the Warriors passed less than any other team per game during the regular season.
Maybe the issue is this: the Warriors really only had three offensive players among their starting five: Stephen Curry, David Lee, and Klay Thompson. The other two starters, Andre Iguodala (who made the All-Defensive First Team) and Andrew Bogut, both played almost completely defensive roles. That's not to say Iguodala and Bogut weren't integral to the Warriors' success; despite having two minus defenders, Curry and Lee, the Warriors managed to have the third most efficient defense last season (their best defensive rating in 50 YEARS). Essentially, the Warriors starting five consisted of two players serviceable on only the offensive end of th floor, Curry and Lee, two serviceable only on the defensive end, Iguodala and Bogut, and one true two-way player who was about as valuable on offense as on defense, Klay Thompson.
But is this the right way to build a lineup? How does the amount of offensive-oriented players affect a team's offense and defense? How does having extreme specialization between offensive-oriented and defensive-oriented players, as the Warriors do, affect a team's success? To look into this, I used ESPN's brand-new statistic, Real Plus-Minus (RPM), which estimates an individual player's contributions on offense and defense, to determine the offensive or defensive orientation of each NBA team's most-used five-man unit during the course of the 2013-2014 season. If a player had a higher Offensive RPM than Defensive RPM, I considered him an offensive-oriented player; the reverse was also true.
Maybe the issue is this: the Warriors really only had three offensive players among their starting five: Stephen Curry, David Lee, and Klay Thompson. The other two starters, Andre Iguodala (who made the All-Defensive First Team) and Andrew Bogut, both played almost completely defensive roles. That's not to say Iguodala and Bogut weren't integral to the Warriors' success; despite having two minus defenders, Curry and Lee, the Warriors managed to have the third most efficient defense last season (their best defensive rating in 50 YEARS). Essentially, the Warriors starting five consisted of two players serviceable on only the offensive end of th floor, Curry and Lee, two serviceable only on the defensive end, Iguodala and Bogut, and one true two-way player who was about as valuable on offense as on defense, Klay Thompson.
But is this the right way to build a lineup? How does the amount of offensive-oriented players affect a team's offense and defense? How does having extreme specialization between offensive-oriented and defensive-oriented players, as the Warriors do, affect a team's success? To look into this, I used ESPN's brand-new statistic, Real Plus-Minus (RPM), which estimates an individual player's contributions on offense and defense, to determine the offensive or defensive orientation of each NBA team's most-used five-man unit during the course of the 2013-2014 season. If a player had a higher Offensive RPM than Defensive RPM, I considered him an offensive-oriented player; the reverse was also true.
How Many David Lees? How Many Andre Iguodalas?
So, how many David Lees versus Andre Iguodalas do you want on your team? Not surprisingly, the more offensively-oriented players a team has, the better its offense is and the worse its defense is. The only exception I found was that teams' defense was actually better with four offensively-oriented players than with three; the most likely reason for this, however, is a small sample size. Only three teams, the Clippers, Nets and Rockets, had four offensively-oriented players as defined by RPM as their most-played five-man units. The vast majority of five-man units surveyed had two or three offensively-oriented players, as would be expected. The average number of offensively-oriented players on the five-man units surveyed was 2.4, signalling essentially an equal distribution of offensively- and defensively-oriented players league-wide. Only one team, Tom Thibodeau's Chicago Bulls death machine had no offensively-oriented players on its most-used five-man unit, a fact which became increasingly apparent when the Bulls could not score on the Raptors during the opening round of the playoffs.
In terms of differentials, it appeared that the most offensively-oriented five-man units, those with four offensively-oriented players, outscored their opponents by .09 points per possession, but there doesn't appear to be any trends regarding point differentials dependent on the number of offensively-oriented players in a five-man unit, aside from, well, more offensively-oriented players means better offense and worse defense.
In terms of differentials, it appeared that the most offensively-oriented five-man units, those with four offensively-oriented players, outscored their opponents by .09 points per possession, but there doesn't appear to be any trends regarding point differentials dependent on the number of offensively-oriented players in a five-man unit, aside from, well, more offensively-oriented players means better offense and worse defense.
What about player specialization?
Now let's take a look at specialization among five-man units and how the amount of "two-way" players in a five-man unit affects point differential. In my analysis, a player who I deemed a "two-way" player had less than a one-point differential between his Offensive and Defensive RPMs. While the definition is slightly arbitrary, it highlights players who offer roughly the same amount of value on the offensive and defensive ends of the floor. The most specialized teams, or those with the fewest "two-way" players, were the Raptors, Spurs, Trail Blazers, Knicks, Warriors, Mavericks and Lakers, each with zero specialized players; the inclusion of the Toronto Drakes, who seem to play a very team-oriented brand of basketball and with solid all-around players like Kyle Lowry and DeMar Derozan, surprised me. The least specialized teams, those with four "two-way" players, were the 76ers and the Bobcats; the presence of the 'Cats wasn't wholly surprising, but I just have no idea of what to make of last year's 76ers, especially because of the never-ending shifts in their rotations. We'll just leave them be.
Anyway, it appears that a stronger relationship exists here between specialization and point differential; the teams with zero specialized players performed the best; as we saw earlier, six of the eight teams with the fewest specialized players were playoff teams. The data for teams with three and four "two-way" players should be disregarded, however; as we saw earlier, the two teams that had four specialized players were the 'Cats and the 76ers, and one was, well, the Sixers, which entirely skews the data. The only team with three "two-way" players was the Cavs, who actually had a pretty good differential with their most-used five-man unit (dat Offensive Efficiency!).
Anyway, it appears that a stronger relationship exists here between specialization and point differential; the teams with zero specialized players performed the best; as we saw earlier, six of the eight teams with the fewest specialized players were playoff teams. The data for teams with three and four "two-way" players should be disregarded, however; as we saw earlier, the two teams that had four specialized players were the 'Cats and the 76ers, and one was, well, the Sixers, which entirely skews the data. The only team with three "two-way" players was the Cavs, who actually had a pretty good differential with their most-used five-man unit (dat Offensive Efficiency!).
Other measures of player specialization
A less arbitrary way to measure specialization among a five-man unit's players is by simply taking the standard deviation among a five-man unit of the difference between a player's Offensive and Defensive RPMs. As we can see in the graph to the right, there appears to be a slight correlation between specialization and higher point differentials (the higher the standard deviation value, the more specialized a player is). The correlation might have been stronger had it not been for that -.23 point differential outlier at the bottom of the graph (shoutout to the Magic for having by far the worst most-played five-man unit in the league).
So what?
So it looks like the Warriors set up their roster the right way - highly specialized between offense and defense, with two offensively-oriented players and two defensively-oriented players. Given the above conclusions, five-man units with higher degrees of specialization between players are more successful, and when building a rotation, the presence of both offensive and defensive specialists, such as Martell Webster or Tony Allen, is vital. Of course, there are many imperfections with the above analysis. First of all is the question of the validity of RPM. Some nuggets I found digging through RPM include:
- David Lee being a plus defender while being a minus offensive player
- Dirk Nowitzki and Vince Carter being better defenders than The Matrix
- Blake Griffin as a "two-way" player and a better defender than CP3
- The aforementioned David Lee having a higher defensive RPM than Anthony Davis
- The venerable Boogie Cousins having a negative offensive rating (leading all Centers in scoring) and a positive defensive rating (he's definitely not as bad on defense as his reputation, but he's clearly an offensive player)