Editors’ note: I’d like to welcome aboard Justin Willard to the Analytics Game team. He’s been writing very interesting, must-read pieces on his own site, ascreamingcomesacrossthecourt.blogspot.com. He will now be contributing full-time to the site, so enjoy. — Evan Sidery
There’s a divide between NBA fans and the analytics contingent. This divide is being breached more and more every day, but there’s still a lot of work to be done in explaining NBA statistics to a wider audience.
For instance, NBA stats, in advanced terms, are framed per possession or per minute, but the general basketball public finds that approach troubling or just plain mysterious. However, it’s built on really simple ideas that have relevance branching out into so many different areas of influence that even casual fans should understand it.
One of the most important practical applications of advanced numbers in basketball is the ability to compare dissimilar players. Manu Ginobili is one of the best illustrations of this concept — how do you value a high-level player who doesn’t play the minutes a typical star does? Even in his heyday, Manu never topped 31.1 minutes per game. If you compare him to other All-Star-caliber guards who play 38 minutes, how do you weigh Manu’s performance fairly?
This is where certain advanced stats come in, and it is not some super advanced topic that requires a math or science-based degree.
A common basketball stat that estimates value is Win Shares, popularized by basketball-reference.com.
Team wins are credited to players via box score stats. It can be used to compare a player like Ginobili to a heavy minutes starter who isn’t as efficient.
For example, Joe Johnson averaged 21.7 points per game in 2008, beating over Ginobili’s 16.5 in 2007; and Joe averaged more assists while equaling him in rebounds. If Johnson averaged more points by a significant amount, then how did Manu have more Win Shares? The answer, basically, is that Ginobili was so much better on a per possession minute basis that it gives his team more value. He got more points out of his shot attempts, and on a per minute basis he was outscoring Johnson.
A critic would counter that Johnson is adding more value because he was actually on the court and averaged more points per game, but if Johnson doesn’t play that doesn’t mean his 21.7 points disappear completely; his possessions are divvied up among other else, and it’s the difference between what Johnson produces and what a Johnson-less team would produce that we’re interested in. That is the crucial point in understanding value stats. Replacing Manu’s production, meanwhile, would be more difficult because of how efficient he is with his shots and how well-rounded he is as a player. Ginobili is even more valuable to a championship contending team because of he good he plays on a per minute basis. Good teams likely have a few players to carry the scoring load; if you want to make a difference you have to provide at a higher level.
By the way, another metric would have similar results. With Hollinger’s PER, Manu had 14 estimated wins added in 2007, while Johnson had 11.9 in 2008. And Manu crushes Johnson in advanced plus/minus stats.
While it’s fairly easy to see how Manu is better than Joe Johnson, it’s not exactly tough competition. A better match-up would be Ginobili and Vince Carter in 2007, arguably his best season post-Toronto. Carter cracked 25 points per game; Ginobili never made it past 20. Yet despite Carter playing 50% more total minutes, they’re nearly even in Win Shares. How does this happen? Again, don’t think of Carter adding 25 points to his team; he’s using up opportunities that can be (partially) replaced by teammates. Manu’s stats were less replaceable because he was ridiculously efficient and filled up the stat-sheet in numerous ways.
This is related to another term that’s used frequently in sports analytics: replacement level.
The replacement level is the level at which you can find a cheap player at any possible time. It’s used a lot in baseball where positions are less malleable and if you’re missing a shortstop, you have to find someone who can play that position. In the NBA, rotations are shorter and positions are fluid so absences or injuries are mostly covered by your same core players, but you still need guys on, say, veteran’s minimum contracts to fill in any voids.
Why do we set the baseline so low and not at something like an average player? A Fangraphs post (about baseball, of course) explains one important reason why:
If the baseline is set to average, then the results of the metric will actually favor inferior talents whose lack of skills convince their manager to keep them chained to the bench in lieu of better players who actually take the field on a semi-regular basis.
Since replacement level is set low, almost every player will have a Win Share score ranging from 0 to higher values. If you see other similar value metrics in basketball, it will most likely be using a very low baseline, often based on a calculation of replacement level.
Setting the baseline that low is useful for a few reasons, but it’s why basketball-reference is listing an unexceptional player with a long career, like Derek Fisher, with a Win Share total near legends with short careers, like David Thompson or Bill Walton. Fisher has 62.3 Win Shares, but if we compared him to an average player it’s -6.1 Win Shares (read this as he’s 6 wins below what an average player would provide). Yao Ming, who was clearly better than Fisher at his best, only has 65.9 Win Shares, but compared to the average player he’s providing 33 more wins.
This is where Manu Ginobili has even more of an advantage over players like Joe Johnson and Vince Carter. If you’re on the same team as players like Duncan and Parker, you have to have compelling reasons to take the ball out of their hands.
His per possession stats are mesmerizing — there’s a reason San Antonio plays so well with him on the court with his ability to take step-back three’s, throw bullets to open players, or drive to the basket with his patented Euro-step, picking up a high rate of free throws.
Manu doesn’t play heavy minutes, but he’s so good when he plays he’s still more valuable than most starters. Production is mostly replaceable; Ginobili is not.
