I’ll be honest, there is no perfect way to rank basketball players. But some methods really give ludicrous results. Based on the current season how would you rank Nick Collison as an NBA player? If you saw the Dallas-OKC series you saw how well Collison defended Dirk and how well OKC played with him in (OKC was +9 points in 135 minutes Nick was in and -29 points in 109 minutes he was out). Yet Nick’s PER rating ranks 268th in the league!. I just do not think a box score metric can capture the full measure of a player’s ability. Many things (mostly involving defense) happen on the court that the box score does not track (taking charges, setting screens, denying Dirk the ball, etc.). The following numbers show how great Collison is.
| Collison | Durant | Harden | Westbrook | Perkins | Maynor | Sefolosha | Mohammed | Ibaka | |
| Collison and other in | 10.58488 | 13.16189 | 10.74301 | 10.94008 | 21.39435 | 8.795835 | 9.21063 | 13.85873 | 10.93732 |
| Collison out and other in | dnp | 4.199121 | 0.259533 | 2.976136 | 2.581464 | -5.45218 | 2.248966 | -3.88659 | 2.110555 |
| Collison in and other out | dnp | 5.903702 | 9.968239 | 10.19988 | 9.623786 | 12.35439 | 11.02362 | 9.924921 | 10.31993 |
| Collison out and other out | 1.361913 | -16.6958 | 2.120075 | -8.00712 | 0.937869 | 2.633226 | -0.38008 | 1.859774 | -0.00036 |
These numbers show how well Thunder played per 48 minutes (adjusted for strength of opponent) in different situations. For example, looking at the Durant column we see with Nick and KD in Thunder played 13 ppg better than average and with Nick out the great KD in OKC was only 4 points better than average . With Collsion and KD out OKC was a horrible 17 ppg worse than average and with Collison in and KD out OKC was +6 ppg. You can see that putting Nick in invariably made the Thunder better. Overall with Nick in Thunder were +10.6 ppg and with Nick out only +1.4. While Adjusted +/- may be fairly noisy over the course of a season, these numbers show that during his time on the court Nick is an elite player.
By the way, our Adjusted +/- for Nick Collison this year has him 2 ppg worse than average on offense but 8 ppg better than average on defense, and he ranks 28th in Adjusted +/-
By the way Steve Nash ranked 22nd in PER but if you look at the following data you will see why he ranked #1 in Adjusted +/-. With Nash in Suns were good +7 ppg; with him out there were horrible -15 ppg. Brooks and Dragic (the backup point guards) are decent players, so do not blame them. Steve Nash is the ultimate example of a player who makes his teammates better.
| Dudley | Frye | Nash | Hill | Gortat | Warrick | Brooks | Dragic | Childress | Richardson | Carter | |
| Nash and other in | 7.43 | 8.27 | 7.24 | 7.88 | 8.56 | 2.81 | 54.33 | 7.61 | 6.98 | 8.31 | 5.45 |
| Nash out and other in | -12.13 | -16.54 | dnp | -20.16 | -11.25 | -14.82 | -7.70 | -19.60 | -15.11 | -38.88 | -14.83 |
| Nash in and other out | 7.11 | 3.97 | dnp | 4.50 | 6.42 | 8.34 | 7.00 | 7.22 | 7.26 | 6.85 | 8.51 |
| Nash out and other out | -24.16 | -13.92 | -15.04 | -13.25 | -18.16 | -15.38 | -18.27 | -10.31 | -14.98 | -12.84 | -15.10 |
Thanks Wayne. All of us Thunderheads who see all the games know what a warrior Nick is. If you only look at points and rebounds, or are searching for fantasy players you will look right past Colly, but his value is obscured by his stat-lite game and the flash of others on the team.
Great analysis.
Comment by Joe — May 30, 2011 @ 9:36 pm
Interesting post. I think it would be fairer to box score metrics in general if you pointed out that Win Shares for instance has a far more positive view of Collison, rating him the third best Thunder player in the playoffs (after Durant and Harden). Not sure what the numbers were for the Dallas series but probably similar. And other box score metrics that value scoring differently arrive at dramatically different values for Nash as well.
I think those guys represent extremes but you make a fair point, the date is very suggestive in their cases that the box score is missing something very substantial.
Comment by Metsox — May 31, 2011 @ 12:03 pm
how does win shares rank Collison during regular season?
Comment by wwinston — May 31, 2011 @ 12:19 pm
Win Shares has Nick at .113 WS per 48 minutes for the regular season or slightly above average overall. http://www.basketball-reference.com/teams/OKC/2011.html
Win Shares assumes that shot defense is a team product and everyone gets equal marks on that and that might undervalue Nick’s defensive contribution. Collison appears to be above average on that according to the defensive points per shot data here http://stats-for-the-nba.appspot.com/pps. But Nick has seems to also have a positive impact on defense rebounding and maybe more.
Your metric seems to rate him higher than RAPM. Regularization has its effect, proper or not. Do you have any commentary of the impact of regularization or the method used?
Comment by Crow — May 31, 2011 @ 3:20 pm
Thanks. I just am not interested in regularization because I think we pick up defensive ability as well as offensive ability. I think that is the main contribution made by Adjusted +/- that other metrics miss. Of course, the price we pay may be more noise, but when you look at the pair data I shared it is usually ver consistent with a player’s Adjusted +/- data. We clearly see for Collison no matter who he played with he made OKC a lot better.
Comment by wwinston — May 31, 2011 @ 3:35 pm
I didn’t read your post carefully enough. I thought the data was for the Dallas series. Win Shares has Collison as an above average player this year (in contrast to PER which rates him significantly below averagae). But yes, Win Shares doesn’t like him nearly as much.
Comment by Metsox — May 31, 2011 @ 5:17 pm
[...] Nick Collison and Steve Nash: An Appreciation: Wayne Winston highlights the underappreciated +/- performance of Collison & Nash. [...]
Pingback by Nick Collison and Steve Nash: An Appreciation » Stathead » Blog Archive — May 31, 2011 @ 6:04 pm
Collison’s .113 WS per 48 minutes for the regular season was slightly above the average overall “performance level”, but among all players his rating was good enough to be about top 25% percentile on WS/48. Average or above average performance level is rarer than might be expected. Top performers elevate that average performance level considerably.
Comment by Crow — May 31, 2011 @ 7:41 pm
PER is clearly the metric most influenced by usage or the lack of it in Nick’s case. If you want to compare leading scorers, PER may have some use. If you want to compare players across position or roles for overall team impact then other metrics including Adjusted +/- may be more useful.
Comment by Crow — May 31, 2011 @ 9:55 pm