I mentioned Friday that Chris Duhon probably wasn't good enough to play in the NBA. I decided to show statistically why that's the case.
On his website, John Hollinger has created a formula (he calls it PER) rating NBA players based on a myriad of variables and standardizes it so that 15.00 is the league average. Anyway, after looking at the site I decided to use his top 50 point guards for the 2002-2003 season to try and predict how Chris Duhon might do if he were to play three years in the NBA.
To do the analysis, I took the 50 top point guards, got their game stats from their last three years of college and their first three years in the NBA. It should be noted that the sample size of 50 was reduced to 27 because of several factors. First, players in the league fewer than three years were dropped from the analysis (It can be argued that any data--even if it's just a rookie year--is better than nothing. Probably, but this analysis is a first pass, just to see what the data may look like). Also, finding college stats for players not drafted proved difficult, and as a result they were also dropped.
So what are good predictors of performance? Maybe the first question to ask is, "Are college statistics a good predictor of professional statistics?" Well, looking at points scored, rebounds, assist-to-turnover ratio and field goal percentage seems like a good place to start. [Before we go any further, I should note that I converted all statistics to output per-minute. For example, points per minute (PPM) were used to compare individuals across leagues (college and professional) based on their points scored per minute. The thinking here is that many professionals do not get a lot of playing time early in their career and scoring per minute seems like a better way to compare performance.] The correlation between PPM in college and PPM as a pro was 0.59, which indicates a positive relationship--as one increases, so does the other. Rebounds per minute (RPM) on the other hand had a correlation of 0.18, which indicates a pretty weak relationship between the number of rebounds one has in college and the number of rebounds one has as a pro (this may especially be true for point guards who may spend less time inside as a pro than they did as a collegian). Assists per minute (APM) had a correlation of 0.67, field goal percentage (FG) had a correlation of 0.66 and assist to turnover ratio per minute (ATRM) had a correlation of 0.68, all of which indicate a pretty strong, positive relationship between amateur and professional numbers.
After looking at the correlations, I decided to run four regressions--one each for PPM, APM, ATRM and FG. To estimate PPM as a pro, I simply regressed PPM(pro) on PPM(college). I follow the same procedure for APM, ATRM and FG.
Results:
As an NBA player, Chris Duhon's three year averages are predicted to be:
(Rankings are based on 27 point guards currently in the NBA)
*Points per game: 7.1 (which ranks 26 out of 28--ahead of only Charlie Ward and Eric Snow. Larry Hughes was 1st)
*Assists per game: 5.0 (which ranks 15 out of 28--Mike Bibby is 1st and Rafer Alston is 28th)
*Assists to turnover ratio: 2.7 (which ranks 3 out of 28--only behind Moochie Norris and Mark Jackson)
*Field goal percentage: 42% (which ranks 11 out 28--Scottie Pippen is 1st and Rafer Alston is 28th)
If we average Duhon's rankings in these four categories he's 14th -- which puts him right in the middle of the pack. I should note that all of these numbers are estimated with some error. For example, I'd be much more comfortable stating the following: I'm 95 percent sure that the actual number of assists per game Duhon will have after three years in the NBA is between 2.0 and 3.3. Of course that really offers no insight into how he might fair because 3.3 would rank him 1st and 2.0 would rank him 24th.
In any event, if we just concern ourselves with the point estimates above, I'd have to say that Duhon would do much better than expected. The statistic that really sticks out is his assists to turnover ratio. You could easily overlook his low points per game total and field goal percentage if he were able to maintain an assist to turnover ratio close to 2.7. In light of this analysis, maybe Duhon would make an above average pro. Just kidding. I still think he's undersized, a poor shooter and can't beat people off the dribble.
So, where does that leave us? In looking at the analysis, I think it's important to remember a couple of things. First, I only looked at 27 point guards currently playing in the NBA (the fact that of the original list of the 50 top point guards, nine went undrafted says something about the process of selecting players and probably should be looked into). Looking at say, all the point guards to play in the NBA over the last 10 years would probably give more precise estimates of a player's future performance. Second, are college statistics a good predictor of professional statistics? This is a question I raised earlier and one that still needs to be addressed. It's not clear that points, assists, rebounds, assist-to-turnover ratio and field goal percentage are the best predictors of future performance. That said, I think it's a good place to start. For example, the models did a very good job of predicting how Mike Bibby, Sam Cassell and Tyronn Lue would perform as pros. However it did a poor job of predicting how Rafer Alston, Stephon Marbury and Charlie Ward would perform (which in part might explain why Alston was last in assists per game and field goal percentage in the estimates above). To go one step further, if we project how Damon Bailey would do after three years in the league based on his numbers at Indiana, he would average 8.3 points, 4.4 assists, an assist to turnover ratio of 2.3 and shoot 41% from the field. For the record, Bailey didn't score a point, get a rebound or register an assist in the NBA.
So the bottom line is: a) making predictions is tricky (see Portland and Sam Bowie), b) the more information you have, the better, and, c) there are certain to be other variables that help predict performance. After going through all this, I'm sure Chris Duhon is destined for a pedestrian career in the NDBL and in three years will be sitting next to Chris Collins and Johnny Dawkins on Coach K's bench.
