Quick Thoughts on RPM
When ESPN debuted RPM last week, I have to admit my first reaction was classic hipster: I've been down with APM models since before they were cool (though, as I'll get into later, I'm not a huge fan of any "one number" metric). That bit of churlishness aside, I want to join the echoing chorus of those lauding ESPN for taking the plunge into the world of non-box score evaluation and especially into giving defense its due.
However, as the Kevin's, Pelton on Zach Lowe's podcast and Ferrigan over at Hickory High have discussed in the last few days, the stat, while more accurate than most, has many of the same problems of any "one number" stat. Namely, the removal of context. To quote Ian Levy yesterday:
The context and the noise, which these models work so hard to control for, are exactly the things I’m interested in. I don’t just want to know which player is better. I want to know why and in what ways. I want to know what that implies about both the player and team, his teammates and opponents, and basketball as a whole. As constructed and presented, I typically find precious little of that information in plus-minus statistics.
I just wanted to call quick attention to two specific areas where RPM (and really, any APM-based stat) may lose some valuable information. Both of them are related to the idea that the problem of using on/off data is there are some things that affect a game when a player leaves the court.
The Perk Effect
This is my semi-affectionate term for guys getting a bump (or in some cases, such as Joakim Noah being "hurt" in APM models for having competent backups like Omer Asik and Taj Gibson over the years, a penalty) for being better than a player who plays a similar role on a team yet the players rarely share the floor. At least part of Nick Collison's high rating in RPM is a result of him being simply not Kendrick Perkins. Similarly, front line players like Kevin Love and Stephen Curry are perhaps overrated slightly by virtue of having sub-replacement backups.
The Memory (and the Bonus) Remains
The other, more subtle point is this. Fouls. With a tip of the hat to James Brocato of Shut Up And Jam who has talked about this a little on twitter, it is worth noting that many of the "defensive bigs" who score highly in RPM (Collison, Amir Johnson, etc.) are players who foul at a high rate. Which means the opposition ends up in the bonus a fair amount when they play. Of course, given standard NBA practice of sitting players in "foul trouble" (no matter how few minutes they are expected to play), the players most responsible for the bonus (and thus extra free throws on non-shooting fouls) aren't in the game when these free throws occur, so the lessened defensive efficiency resulting from these fouls is actually assigned to our foul-happy hero's replacement. As with the Perk Effect, the difference in efficiency between players in a similar role can inflate the apparent contribution of the "better" performing player. Thus by committing to many fouls, a player can "hurt" his playing time rival in an APM model and therefore give himself a perverse bonus for bad defense or over aggressive rebounding.
On the flip side, players who are good at drawing fouls might be undervalued, because their teammates can take advantage of the extra time spent in the bonus, but also because of how the fouls they draw affect the opponents' play. Putting fouls on starting players brings the backups into a game, and forcing the opposition to play their worse players more is obviously helpful in terms of winning the game. Even if these fouls don't cause foul trouble, they can still reduce the aggression with which opponents play out of concern for picking up the next foul and thus being forced to sit on the bench during a normal rotation spot, losing minutes and thus counting stats and ultimate money.
None of this is to say RPM is a "bad stat" but it is not as unbiased or proponents would have us believe, and these effects along with those mentioned in the links above are some of the important caveats to keep in mind.
On the flip side, players who are good at drawing fouls might be undervalued, because their teammates can take advantage of the extra time spent in the bonus, but also because of how the fouls they draw affect the opponents' play. Putting fouls on starting players brings the backups into a game, and forcing the opposition to play their worse players more is obviously helpful in terms of winning the game. Even if these fouls don't cause foul trouble, they can still reduce the aggression with which opponents play out of concern for picking up the next foul and thus being forced to sit on the bench during a normal rotation spot, losing minutes and thus counting stats and ultimate money.
None of this is to say RPM is a "bad stat" but it is not as unbiased or proponents would have us believe, and these effects along with those mentioned in the links above are some of the important caveats to keep in mind.
RSS Feed
