Christian Wood (PHI) is better than Andrew Wiggins. I have proof.

[AbeVigodaLive]

http://espn.go.com/nba/statistics/rpm/_/page/8/sort/RPM" onclick="window.open(this.href);return false;

Furthermore... according to Wiggins' defensive +/- stats... he's the 407th best (out of 413) players.




[Note: Why is Minnesota using him as its crunchtime offensive guy and the guy checking the opposing team's best player? He's worse than D-League candidates like Johnny O'Bryant III!]

[Kingcarl24]

I believe Christian Wood is what Cleve had on his wedding night.

[AbeVigodaLive]

I believe Christian Wood is what Cleve had on his wedding night.

[Robceltsfan]

If you think that's funny, check out who #413 out of 413 is.

[AbeVigodaLive]

If you think that's funny, check out who #413 out of 413 is.

Damn. Some good ones on there.

J. Okafor is the worst. But Kobe and Derrick Rose aren't too far ahead of them. It's even funnier when you look at people ahead of them on the list. Consider...

Tyus Jones season stats:

14 minutes.
0 - 5 fg
1 - 2 ft
1 point
1 assist
1 foul
1 rebound
30 Offensive rating
102 defensive rating


He's better than Kobe Bryant, Derrick Rose, E. Mudiay, J. Okafor, Stanley Johnson, Corey Brewer, etc...

[Kingcarl24]

If you think that's funny, check out who #413 out of 413 is.

How about #409?

[Robceltsfan]

If you think that's funny, check out who #413 out of 413 is.

Damn. Some good ones on there.

J. Okafor is the worst. But Kobe and Derrick Rose aren't too far ahead of them. It's even funnier when you look at people ahead of them on the list. Consider...

Tyus Jones season stats:

14 minutes.
0 - 5 fg
1 - 2 ft
1 point
1 assist
1 foul
1 rebound

He's better than Kobe Bryant, Derrick Rose, E. Mudiay, J. Okafor, Stanley Johnson, Corey Brewer, etc...

I guess it's all about how much negative impact you're having on your team.

Kobe and Rose kill their teams for a longer period of time than Tyus Jones does.

[Havlicekstealsit]

I believe Christian Wood is what Cleve had on his wedding night.

Annnnd we're done here.

[gaskill15]

Real Plus-Minus? Meh

[AbeVigodaLive]

Real Plus-Minus? Meh

The stat geek wonks are trying to take over the NBA like they did MLB. It won't work. I HOPE it won't work. It doesn't translate in the same ways.

Roles matter. Context matters. Situations matter. The eye test matters.



[Note: And I'm considered a "stat" guy on here. But there are limits to what they can tell us on the basketball court when there are SO MANY moving parts (noise) at any given time.]

[FPL]

RPM (formerly xRAPM) is a nice metric, but this early in the season, so much of it is derived from the SPM prior (statistical plus minus, a metric like a BPM comprised of box score stats, with other info like height, age), that it probably doesn’t match the reality this season so far.

You’ll see pure RAPM (no box score data) soon, probably around midseason (the RPM creator, J.E., has already listed his pure RAPM some on his twitter). Both RPM and RAPM (as long as there is a prior; non-prior informed or NPI doesn’t have the same issue) are generally low on rookies, since without any prior league experience. In recent memory, RPM and RAPM were slow to react to improvements by Durant and Davis, and the same will be the case with Wiggins.

All in all it’s a powerful family of metrics. As long as you compare players in similar roles/with similar primacy on system to one another, they make for nice comparative tools (and in contests to predict wins, those from the RPM/RAPM family have performed the best, and have beaten Vegas/media consistently; this in addition to the fact that aside from the Nets, Knicks, and Lakers, all teams have bought into analytics in general).

[gaskill15]

Real Plus-Minus? Meh

The stat geek wonks are trying to take over the NBA like they did MLB. It won't work. I HOPE it won't work. It doesn't translate in the same ways.

Roles matter. Context matters. Situations matter. The eye test matters.



[Note: And I'm considered a "stat" guy on here. But there are limits to what they can tell us on the basketball court when there are SO MANY moving parts (noise) at any given time.]

Look at this description for their RPM:
"Player's estimated on-court impact on team performance, measured in net point differential per 100 offensive and defensive possessions. RPM takes into account teammates, opponents and additional factors"

"Additional factors". They don't even know what all goes into their "formula".

[FPL]

Look at this description for their RPM:
"Player's estimated on-court impact on team performance, measured in net point differential per 100 offensive and defensive possessions. RPM takes into account teammates, opponents and additional factors"

"Additional factors". They don't even know what all goes into their "formula".

RPM uses ridge regression - a method similar to OLS, but in the case of ill-posed problems (since there are more 10-man lineup combinations than players in the league...something like 35k including the season+playoffs, vs 400 something players in the league). So teammates and opponents, all players in the league, are treated as variables.

The additional factors are the SPM prior (box score + age + height), and some adjustments made (playoff possessions are weighted higher, possessions in blowouts are weighted less, when guys go to the line instead of using the actual FT made/missed when accounting for margin using player FT% to eliminate luck, adjustments for back-to-backs/4 games in 5 nights situations, etc.) that J.E. (stat maker) has determined to be factors that improve predictive ability through separate studies (now whether it improves explanatory power is another debate entirely, which is why I utilize pure play-by-play RAPM when evaluating players...which also has the advantage of being something that can be reproduced).

[AbeVigodaLive]

RPM (formerly xRAPM) is a nice metric, but this early in the season, so much of it is derived from the SPM prior (statistical plus minus, a metric like a BPM comprised of box score stats, with other info like height, age), that it probably doesn’t match the reality this season so far.

You’ll see pure RAPM (no box score data) soon, probably around midseason (the RPM creator, J.E., has already listed his pure RAPM some on his twitter). Both RPM and RAPM (as long as there is a prior; non-prior informed or NPI doesn’t have the same issue) are generally low on rookies, since without any prior league experience. In recent memory, RPM and RAPM were slow to react to improvements by Durant and Davis, and the same will be the case with Wiggins.

All in all it’s a powerful family of metrics. As long as you compare players in similar roles/with similar primacy on system to one another, they make for nice comparative tools (and in contests to predict wins, those from the RPM/RAPM family have performed the best, and have beaten Vegas/media consistently; this in addition to the fact that aside from the Nets, Knicks, and Lakers, all teams have bought into analytics in general).

Is this the same metric we've discussed here before where a guy like Towns was considered worse than a guy like Reggie Evans?

If it offers incomplete data... why even share it publicly? Other than to get us talking about it on forums like this?



[Note: Again, I dig stats. A lot. I think EVERY team should be using analytics in one way or another. To not do so... is silly and archaic. But if NBA discussion "evolves" into a bunch of computer-generated metrics and acronyms as the end-all/be-all for judging players... I'm out. I just don't think basketball can be placed into such neat little compartmentalized boxes like that.]

[FPL]

Is this the same metric we've discussed here before where a guy like Towns was considered worse than a guy like Reggie Evans?

If it offers incomplete data... why even share it publicly? Other than to get us talking about it on forums like this?



[Note: Again, I dig stats. A lot. I think EVERY team should be using analytics in one way or another. To not do so... is silly and archaic. But if NBA discussion "evolves" into a bunch of computer-generated metrics and acronyms as the end-all/be-all for judging players... I'm out. I just don't think basketball can be placed into such neat little compartmentalized boxes like that.]

No clue, didn't see the Reggie Evans/Towns comparison. Again though, the rookie prior makes guys early on look a lot worse (Towns would look much better in pure RAPM for a single year).

Well you can't replace all the qualitative scouting obviously...I think the two have to go hand-in-hand, and one should inform the other (the eye test is tremendous, but also on its own doesn't have much value as a cop out in debates, either in the fandom or internally among organizations...now if we're speaking of detailed scouting reports, with tendencies, skillset breakdowns and the like, THAT is truly valuable). But I think what analytics are replacing are a lot of the naive box score stats and metrics which probably did more harm than good for years and years (like the Hollingers of the world...).

I do hope that we keep getting more and more quality data, to be mined and scrubbed over time. There are some massive datasets, and there are so many hidden trends, some of which are confirming long-held beliefs, and some of which are correcting false conventional wisdom.

[AbeVigodaLive]

Is this the same metric we've discussed here before where a guy like Towns was considered worse than a guy like Reggie Evans?

If it offers incomplete data... why even share it publicly? Other than to get us talking about it on forums like this?



[Note: Again, I dig stats. A lot. I think EVERY team should be using analytics in one way or another. To not do so... is silly and archaic. But if NBA discussion "evolves" into a bunch of computer-generated metrics and acronyms as the end-all/be-all for judging players... I'm out. I just don't think basketball can be placed into such neat little compartmentalized boxes like that.]

No clue, didn't see the Reggie Evans/Towns comparison. Again though, the rookie prior makes guys early on look a lot worse (Towns would look much better in pure RAPM for a single year).

Well you can't replace all the qualitative scouting obviously...I think the two have to go hand-in-hand, and one should inform the other (the eye test is tremendous, but also on its own doesn't have much value as a cop out in debates, either in the fandom or internally among organizations...now if we're speaking of detailed scouting reports, with tendencies, skillset breakdowns and the like, THAT is truly valuable). But I think what analytics are replacing are a lot of the naive box score stats and metrics which probably did more harm than good for years and years (like the Hollingers of the world...).

I do hope that we keep getting more and more quality data, to be mined and scrubbed over time. There are some massive datasets, and there are so many hidden trends, some of which are confirming long-held beliefs, and some of which are correcting false conventional wisdom.

Sure. All makes sense.

A metric with Andrew Wiggins as the 7th worst defender out of 413 players in the NBA? Much less so...

[FPL]

I think the biggest challenge facing analytics in the NBA today, is that it can be difficult to communicate results to those who need them...a lot of owners have made their names on Wall Street and the like, and more and more quantitative GMs and other front office executives are flooding in over time. Some coaches are pretty receptive to suggestions (Carlisle had a guy on the bench monitoring lineup data during games during the championship run), while others, not so much (*cough* Mike Woodson *cough*). If you have the data, but nobody is listening, what is the use?

Sure. All makes sense.

A metric with Andrew Wiggins as the 7th worst defender out of 413 players in the NBA? Much less so...

As noted in the quoted post and one or two above, he's a victim of the rookie prior (instead of assigning them a SPM value as a starting point for the regression, they get something like -3). Improves predictive value, but not necessarily explanatory value. He'll look much better in single-year NPI RAPM (which contains no box score data, or information from previous years, including the rookie prior). Same thing happened with Durant, Davis, or even going back further, LeBron, among others. Guys outpace their priors within ~3 years typically.

[AbeVigodaLive]

I think the biggest challenge facing analytics in the NBA today, is that it can be difficult to communicate results to those who need them...a lot of owners have made their names on Wall Street and the like, and more and more quantitative GMs and other front office executives are flooding in over time. Some coaches are pretty receptive to suggestions (Carlisle had a guy on the bench monitoring lineup data during games during the championship run), while others, not so much (*cough* Mike Woodson *cough*).

Sure. All makes sense.

A metric with Andrew Wiggins as the 7th worst defender out of 413 players in the NBA? Much less so...

As noted in the quoted post and one or two above, he's a victim of the rookie prior (instead of assigning them a SPM value as a starting point for the regression, they get something like -3). Improves predictive value, but not necessarily explanatory value. He'll look much better in single-year NPI RAPM (which contains no box score data, or information from previous years, including the rookie prior). Same thing happened with Durant, Davis, or even going back further, LeBron, among others. Guys outpace their priors within ~3 years typically.

Yeah, I'm not seeing how it's a viable stat/metric/ranking than for anybody within their first couple years in the league...

[FPL]

Yeah, I'm not seeing how it's a viable stat/metric/ranking than for anybody within their first couple years in the league...

RPM isn't, I'd agree with that. Pure play-by-play NPI RAPM on the other hand, doesn't suffer from the same shortcomings, since everyone starts at 0.

RE:Mauer

Nah, Deng looked quite good defensively for those Bulls teams for awhile though by impact metrics. Some role player types do pretty well, though that's part of comparing guys playing similar roles (while single year does reduce the rookie problem, collinearity can be an issue if you have guys playing the majority of their minutes together).

In terms of this year, it'll tell you that Leonard and Green are top 10 guys in terms of impact (and were last year as well), though I guess Jerry West is saying the same thing, so it's not *that* crazy. There are also some fun conclusions to draw about KG, but I've gotten into that before.

Anyhow gotta run, always fun to talk ball, have a good one.

[AbeVigodaLive]

Yeah, I'm not seeing how it's a viable stat/metric/ranking than for anybody within their first couple years in the league...

RPM isn't, I'd agree with that. Pure play-by-play NPI RAPM on the other hand, doesn't suffer from the same shortcomings, since everyone starts at 0.

Am I going to be told/scolded in a couple years that Kirk Hinrich is an All NBA player like I've been scolded for years that Joe Mauer is an advanced metric god?