I’m a born bullshitter. But when it comes to the efficacy of the Marginal Win Score metric I use on this blog, I have no interest in making snake oil claims that do not test true.
Here’s what I’ve learned of it so far. MWS48 appears to be a very strong tool for predicting the wins a given group of players would be likely to produce based upon each player’s past performance. (See EndNote for a discussion of this point.)
It is not, however, real strong at predicting professional performance based upon college performance. I am going to take a longer, harder look at how it might translate, but from past attempts its only reliable in one sense.
The “sense” I refer to above is the following. If a prospect did not produce eye-popping numbers in college the odds are stacked against him producing eye-popping numbers in the pros. (this is effectively the very same point made by Berri and Schmidt in Stumbling on Wins. In fact, Marginal Win Score appears to do no better, and probably does worse, than Parent Win Score at projecting college prospects.)
John Wall did not produce eye-popping “win” stats in college. His Win Score per 40 minutes was 7.21. That would project to a Win Score in the NBA of something like 5.94. The average point guard in the NBA has a Win Score of around 6.80. Thus, unless Wall improves upon his efficient production against much better competition than the SEC provides, he is unlikely to produce huge wins.
By comparison, in one season at Memphis, Chicago PG Derrick Rose posted a WS40 of 8.51 (I estimate his MWS40 at +2.97, similar to my estimation of Wall at +2.29). Rose has thus far proven to be an average win producer in the NBA. He may get better — and “average” is pretty good — but we’re talking about number one picks, so the expectations are higher.
DeMarcus Cousins, on the other hand, produced pretty outstanding “win” numbers in college. His WS40 was 17.97. That projects to a WS48 in the NBA of 13.64. The average NBA center has a WS48 of 12.30. And with Cousins size he projects as a pretty solid defender, so I see most likely above average wins coming from him.
Cousins one season in college compares favorably to Kevin Love’s one season at UCLA. Love posted a WS40 of 18.69 in that season. (I estimate his MWS40 at +7.02, and I estimate Cousins MWS40 at +6.29). Love has been a tremendous win producer so far in the NBA, despite being surrounded by a lousy Timberwolves team.
But again, I thought the same about Michael Beasley. His college numbers projected him as a superstar. Hasn’t happened yet. Doesn’t look like it will.
And last season I got cute by trying to adjust Stephen Curry’s numbers by competition. He didn’t seem to perform well against the better competition, so I downgraded him. I was wrong again.
But I was right on Love, Ty Lawson, DaJuan Blair, Ryan Anderson, Al Horford, James Harden… all players whom the national media downgraded (Doug Gottlieb said Harden had “no particular skill”).
So its a mixed bag. But when someone doesn’t produce monster numbers in college, it certainly raises concerns. The burden of proof, it seems to me, shifts to those who would argue that he will be more productive against professional competition than he was against amateurs.