I was thinking about the potential of a stat that's the equivalent of the MLB's wOBA, which I've felt for a while was the golden stat in that sport and the one that should be used instead of OPS which I see as no better than PER http://www.fangraphs.com/library/offense/woba/
Essentially I'd figure it'd go like this
- Use the massive resources of +/-, APM, etc. methods you guys have developed, except instead of concentrating on players, just concentrate on individual statistics
- The result would be to get values like "If a player has X Pts * Y TS% in a game, it's worth Z Ws" "X Rebs = Z Ws", "X Asts = Z Ws". Basically the answer to this formula that looks like this:
wOBA = (0.691×uBB + 0.722×HBP + 0.884×1B + 1.257×2B + 1.593×3B +
2.058×HR) / (AB + BB – IBB + SF + HBP)
If +/- research the community can come up with a stat that's similar for the NBA, I think that'd be pretty cool
Has anyone does this stat?
Re: Has anyone does this stat?
WP and EV have done stuff like that.
Re: Has anyone does this stat?
Dr. Positivity: check out Evan Zamir's work on Adjusted 4-Factor Plus-Minus, where he calculated each player's impacts on each of the 4 factors. I know it's not exactly what you're looking for, but does deal with the difference between team stats and player stats and using APM at that level.
(As for WP, it calculates the value of stats at the team level, but does not properly translate to the player level, instead assuming direct translation--i.e. a rebound is worth a possession at a team level, so a player who grabbed a rebound gets credit for adding a rebound to the team--in actuality, it's far less. http://www.countthebasket.com/blog/2008 ... -rebounds/)
As for individual stats--Advanced Statistical Plus/Minus (ASPM) (my stat) I think does almost exactly what you are saying--I regressed onto long-term (10 year) average RAPM using individual player advanced stats, constructed similarly to how you are indicating. However, I found very strong interaction between assists and scoring, to the point that only a composite term from those statistics are used. My derivation is here: http://godismyjudgeok.com/DStats/aspm-and-vorp/ . I think it's pretty close to what you wanted, and has gained good acceptance in the advanced stats community.
(As for WP, it calculates the value of stats at the team level, but does not properly translate to the player level, instead assuming direct translation--i.e. a rebound is worth a possession at a team level, so a player who grabbed a rebound gets credit for adding a rebound to the team--in actuality, it's far less. http://www.countthebasket.com/blog/2008 ... -rebounds/)
As for individual stats--Advanced Statistical Plus/Minus (ASPM) (my stat) I think does almost exactly what you are saying--I regressed onto long-term (10 year) average RAPM using individual player advanced stats, constructed similarly to how you are indicating. However, I found very strong interaction between assists and scoring, to the point that only a composite term from those statistics are used. My derivation is here: http://godismyjudgeok.com/DStats/aspm-and-vorp/ . I think it's pretty close to what you wanted, and has gained good acceptance in the advanced stats community.
Re: Has anyone does this stat?
Dr. P, are you just looking for a linear weights formula?
You won't get very satisfying results just using raw stats. For example, 10 rebounds in a game with 100 rebounds is not worth as much as 10 in a game with 70. And .475 shooting is pretty good in a game of .425 shooting.
You won't get very satisfying results just using raw stats. For example, 10 rebounds in a game with 100 rebounds is not worth as much as 10 in a game with 70. And .475 shooting is pretty good in a game of .425 shooting.