Latrinsorm
04-07-2014, 05:26 PM
± makes a lot of sense (hockey's been using it for centuries) but it carries a lot of teammate effects, and we'd rather have individual stats. The most basic approach that I've described before is comparing a player's ± with the the team's when he sits, or On-Off (http://forum.gsplayers.com/showthread.php?88781-NBA-On-Off-and-Serge-Ibaka), but that leaves odd results like Nick Collison and Mike Conley being elite players.
There are no NBA games today due to the minor league championship, and ESPN took this opportunity to unveil yet another ± stat: Real Plus Minus (http://espn.go.com/nba/statistics/rpm/_/sort/RPM). It's the same approach others (http://www.82games.com/barzilai2.htm) have taken before with somewhat less self-aggrandizing titles, a more sophisticated way of trying to control for variables besides the individual. If you look at raw ± (http://www.sportingcharts.com/nba/stats/individual-player-plus-minus-statistics/2013/) for this season you can see how it changes the order pretty significantly, but we still end up with Collison #6 and Conley #15.
When so many empiro-statistical methods tell us something that we know is wrong, it might be time to question what we know. Maybe Nick Collison like Chris Andersen has found a way to optimize his effort for the comparatively short time he's on the court, maybe Mike Conley really is good and Memphis' pace and location occludes his impact. Or maybe Kendrick Perkins and Jerryd Bayless just stink, and no amount of statistical subtlety can weed that effect out. Either way, one thing that's for sure is that RPM is able to suss out what my crude method did in one season instead of four, which is a big advantage.
.
We can't really compare RPM directly with raw PM because the first is given in terms of 100 possessions, but if we adjust raw PM to per 48 minutes we'll be pretty close. We can then take the top 25 guys in RPM and do real - raw, so whoever has the highest number is carrying his teammates(/system/etc.) and whoever has the lowest is benefiting from them:
2.46 Anderson Varejao
2.18 Mike Conley
1.32 Taj Gibson
1.11 Amir Johnson
0.78 LeBron James
0.16 Joakim Noah
-0.28 Dirk Nowitzki
-1.18 Chris Andersen
-1.36 Kevin Durant
-1.67 Kevin Love
-1.81 Channing Frye
-2.28 Dwight Howard
-2.29 LaMarcus Aldridge
-2.53 Tim Duncan
-2.71 Russell Westbrook
-3.30 Patrick Beverley
-3.43 Ricky Rubio
-4.02 Stephen Curry
-4.15 Chris Paul
-4.26 DeAndre Jordan
-4.92 Chris Bosh
-5.63 Blake Griffin
-6.06 Nick Collison
-6.30 Andre Iguodala
-9.87 Manu Ginobili
First of all, it's really interesting how dramatically the results are rated towards the negative. This makes sense though: to be a great player you need help, there's only so much an individual can do.
Second, LeBron is still carrying the Heat, which I doubt anyone would expect. The RPM analysis has only been done on the 2014 season and can only be done with play-by-play data (like all other ±), plus they haven't published their methodology so I can't do it myself, but it would be fascinating to see how his real-raw went by year.
Third, I told you guys about Chris Paul. Here's yet another advanced metric where the dude dominates every guard. Criminally underrated. (I told you about Duncan too, but he's pretty properly rated.)
Fourth, Durant!!?? Obviously this research is not going to impact the MVP vote in any way, and he's a lock to win it, but like LeBron's history it'll be fascinating to see if voting hews to this advanced metric the way it does to Win Shares.
There are no NBA games today due to the minor league championship, and ESPN took this opportunity to unveil yet another ± stat: Real Plus Minus (http://espn.go.com/nba/statistics/rpm/_/sort/RPM). It's the same approach others (http://www.82games.com/barzilai2.htm) have taken before with somewhat less self-aggrandizing titles, a more sophisticated way of trying to control for variables besides the individual. If you look at raw ± (http://www.sportingcharts.com/nba/stats/individual-player-plus-minus-statistics/2013/) for this season you can see how it changes the order pretty significantly, but we still end up with Collison #6 and Conley #15.
When so many empiro-statistical methods tell us something that we know is wrong, it might be time to question what we know. Maybe Nick Collison like Chris Andersen has found a way to optimize his effort for the comparatively short time he's on the court, maybe Mike Conley really is good and Memphis' pace and location occludes his impact. Or maybe Kendrick Perkins and Jerryd Bayless just stink, and no amount of statistical subtlety can weed that effect out. Either way, one thing that's for sure is that RPM is able to suss out what my crude method did in one season instead of four, which is a big advantage.
.
We can't really compare RPM directly with raw PM because the first is given in terms of 100 possessions, but if we adjust raw PM to per 48 minutes we'll be pretty close. We can then take the top 25 guys in RPM and do real - raw, so whoever has the highest number is carrying his teammates(/system/etc.) and whoever has the lowest is benefiting from them:
2.46 Anderson Varejao
2.18 Mike Conley
1.32 Taj Gibson
1.11 Amir Johnson
0.78 LeBron James
0.16 Joakim Noah
-0.28 Dirk Nowitzki
-1.18 Chris Andersen
-1.36 Kevin Durant
-1.67 Kevin Love
-1.81 Channing Frye
-2.28 Dwight Howard
-2.29 LaMarcus Aldridge
-2.53 Tim Duncan
-2.71 Russell Westbrook
-3.30 Patrick Beverley
-3.43 Ricky Rubio
-4.02 Stephen Curry
-4.15 Chris Paul
-4.26 DeAndre Jordan
-4.92 Chris Bosh
-5.63 Blake Griffin
-6.06 Nick Collison
-6.30 Andre Iguodala
-9.87 Manu Ginobili
First of all, it's really interesting how dramatically the results are rated towards the negative. This makes sense though: to be a great player you need help, there's only so much an individual can do.
Second, LeBron is still carrying the Heat, which I doubt anyone would expect. The RPM analysis has only been done on the 2014 season and can only be done with play-by-play data (like all other ±), plus they haven't published their methodology so I can't do it myself, but it would be fascinating to see how his real-raw went by year.
Third, I told you guys about Chris Paul. Here's yet another advanced metric where the dude dominates every guard. Criminally underrated. (I told you about Duncan too, but he's pretty properly rated.)
Fourth, Durant!!?? Obviously this research is not going to impact the MVP vote in any way, and he's a lock to win it, but like LeBron's history it'll be fascinating to see if voting hews to this advanced metric the way it does to Win Shares.