[ScottMay]

The Roland Rating has changed. Since they're now charting more information in games, they've added a weighted (2:1) PER differential of player vs. his opponent(s), which they feel will reflect good individual defensive play.

More here:

http://www.82games.com/0506/rolandratings0506.htm

Any of our statheads (Mike, Johnston) care to weigh in?

 

[MikeDC]

I think the weak point is the "Opponent Counterpart" PER, because that's bound to be pretty "noisy" because I'm not sure whether it tracks who's actually being guarded or who is playing the equivalent role on the other team. Unless I misunderstand how they're done, I'd just use the player's PER, not NET PER.

For instance, on the Spurs, Manu is typically the SG and Bruce Bowen is the SF, but Bowen typically guards the best offensive wing player whether it's an SG or SF.

I also don't really like per-48 minute stats, which both PER and +/- are, because I think they miss out on some important but hard to quantify aspects of the game, like matchups, foul trouble, and pace of play.

Those criticisms taken, this is an improvement and I think it makes the ratings somewhat more useful. Just looking at them in common sense terms, they appear to make a bit more sense.

Something else I'd like to see is weighting the +/- time against opponents' winning percentage and possibly against the current game conditions. A high net +/- vs. a bad team or when the game is in "garbage time" may ought to count less than a high +/- vs. a good team or when the game is close.

 

[Rhyder]

Something else I'd like to see is weighting the +/- time against opponents' winning percentage and possibly against the current game conditions. A high net +/- vs. a bad team or when the game is in "garbage time" may ought to count less than a high +/- vs. a good team or when the game is close.

I think you can only use opponents winning percentage as sort of a standard deviation to the Roland rating. If the Bulls are any indication, they are currently 4-5. They are better than another 4-5 team that beat 4 bottom dweller teams and gotten blown out by 20 by playoff-calibur teams. i.e. we could easily have two or three more losses (or wins) simply because they are playing in close games. I know you are alluding to this with your "garbage time" argument, but using opponents winning % I don't think would correlate well with the Roland ratings.

As I've said many times in the past, the Roland ratings are really only useful when comparing the players playing on the same team; thus, I think a pretty good system is in place as long as you understand the assumptions. Starters on a deep team such as the Bulls or Memphis would have a lesser Roland rating than starters on lets say Phoenix or New Jersey, simply because the starters are significantly better than the talent on the bench.

 

[johnston797]

It's now a low tech approximation of the DanVal system.

I think there is merit in PER and +/- net. Per Mike's post, not sure I know enough about the Opponent Per but at a high level makes sense.

I like PER and I am glad this was done for no other reason than it's an easy place to look it up. Previously, I had to go a couple screens deep in 82games and do some easy math.

p.s. Somewhat off-topic, and it's very early, but Duhon and Gordon's great defense from last year per the +/- net is not holding up. I definetely can't decide how much to value that stat.

 

[MikeDC]

I think you can only use opponents winning percentage as sort of a standard deviation to the Roland rating. If the Bulls are any indication, they are currently 4-5. They are better than another 4-5 team that beat 4 bottom dweller teams and gotten blown out by 20 by playoff-calibur teams. i.e. we could easily have two or three more losses (or wins) simply because they are playing in close games. I know you are alluding to this with your "garbage time" argument, but using opponents winning % I don't think would correlate well with the Roland ratings.

As I've said many times in the past, the Roland ratings are really only useful when comparing the players playing on the same team; thus, I think a pretty good system is in place as long as you understand the assumptions. Starters on a deep team such as the Bulls or Memphis would have a lesser Roland rating than starters on lets say Phoenix or New Jersey, simply because the starters are significantly better than the talent on the bench.

Interestingly though, based on net +/-, look at those teams you mention. Examples:

Phoenix:
Nash -32.8
Marion -24

Bulls:
Gordon 9.8
Hinrich -4.7
Songaila 6.2

Memphis:
Jones 17
Gasol 9

New Jersey:
Kidd 22.6
RJ 12.1
Vince 19.5

Looking at all of that, it's damn hard to make heads or tails even within a team. Memphis seems to be a deep team where you'd think there's not much seperation between the starters and the backups. But there's a pretty big disparity. The Nets you'd think would have a big disparity between starters and backups and they do. The Suns have a big disparity but it goes in the opposite way you'd think - their backups rate out higher than the starters. Weird. The Bulls are sort of a mix - there's a lot of strange stuff going on there and it's pretty hard to figure out even within a team what's going on.

 

[johnston797]

Interestingly though, based on net +/-, look at those teams you mention. Examples:

Phoenix:
Nash -32.8
Marion -24

......... Weird. The Bulls are sort of a mix - there's a lot of strange stuff going on there and it's pretty hard to figure out even within a team what's going on.

Last year for the Suns looked as Rhyder suggested.

http://www.82games.com/0405PHO.HTM

Ok - so looking at Nash per 2005, I am starting to think that it makes no sense to look at this numbers unless at least 20 games have been played.

With last year's data, I will say that over a course of the year there were only 2 players that I though were definetly good with a negative +/- by more than 2 points: Hinrich and Caron Butler. So it's defintely correlating with something.

 

[SausageKingofChicago]

I also don't really like per-48 minute stats, which both PER and +/- are, because I think they miss out on some important but hard to quantify aspects of the game, like matchups, foul trouble, and pace of play.

And fatigue

And other teams figuring out a way to play you and negate you the longer that your out there and are a factor.. defenses adjust and per 48's cannot account for that in any plausible way IMO


Rick Pittino disagreed however

And Travis Knight and Co thank the good lord that he did

 

[Rhyder]

Interestingly though, based on net +/-, look at those teams you mention. Examples:

Phoenix:
Nash -32.8
Marion -24

Bulls:
Gordon 9.8
Hinrich -4.7
Songaila 6.2

Memphis:
Jones 17
Gasol 9

New Jersey:
Kidd 22.6
RJ 12.1
Vince 19.5

Looking at all of that, it's damn hard to make heads or tails even within a team. Memphis seems to be a deep team where you'd think there's not much seperation between the starters and the backups. But there's a pretty big disparity. The Nets you'd think would have a big disparity between starters and backups and they do. The Suns have a big disparity but it goes in the opposite way you'd think - their backups rate out higher than the starters. Weird. The Bulls are sort of a mix - there's a lot of strange stuff going on there and it's pretty hard to figure out even within a team what's going on.

I didn't look at the numbers before suggesting the teams I mentioned. However, it being only 10 games into the season, it is a relatively small sample size.

Interesting data you pointed to, definately.

 

[MikeDC]

Last year for the Suns looked as Rhyder suggested.

http://www.82games.com/0405PHO.HTM

Ok - so looking at Nash per 2005, I am starting to think that it makes no sense to look at this numbers unless at least 20 games have been played.

With last year's data, I will say that over a course of the year there were only 2 players that I though were definetly good with a negative +/- by more than 2 points: Hinrich and Caron Butler. So it's defintely correlating with something.

I think you guys are right that over a season or so, the data will aggregate out and the right correlation will be seen between net +/- and quality.

But... and I mean this in the nicest possible way because I'm interested in using such data myself... I can't use it to do anything. While there's a strong correlation, there are so many other things going on at the individual level that I don't see how a single result can tell you anything of value.

I know it's not meant as a "stand-alone" statistic, but it's used that way, and it has no meaning without some context and narrative. And that makes it an annecdote, not a statistic.

 

[johnston797]

But... and I mean this in the nicest possible way because I'm interested in using such data myself... I can't use it to do anything. While there's a strong correlation, there are so many other things going on at the individual level that I don't see how a single result can tell you anything of value.

I was a big skeptic. Then less so in part b/c John Hollinger adopted the +/- net for his defense and I think he is a stud. But after seeing this year's data, I am leaning back to the skeptic side, too.

 

[MikeDC]

I was a big skeptic. Then less so in part b/c John Hollinger adopted the +/- net for his defense and I think he is a stud. But after seeing this year's data, I am leaning back to the skeptic side, too.

Take a look at last year's again. I'm not sure I'd buy your argument that Butler and Hinrich were the only guys you were pretty sure where mischaracterized. The following are guys besides Kirk and Butler who had a net +/- less than -2:

Curry, Marquis Daniels, Yao, Wright, Wells, Posey, RJeff, Crawford, Nazr, Dwight Howard, Leandro Barbosa, SAR, Mobley, Brent Barry, Devin Brown, Reggie Evans, Nazr, Jalen Rose, Jarvis Hayes

1- Hinrich, Crawford
2- RJefferson, M. Daniels, Barry, Wells
3- Butler, Posey
4- Dwight Howard, SAR, Reggie Evans
5- Yao, Curry, Wright

That'd be a pretty interesting team to watch.

Now I'm not saying there's "no" correlation at all. But it does seem to me that if the correlation is so lose that I can go through and pick out 20+ guys that rate out poorly and field a good team with them, then it's hard to put a lot of confidence in it.

I do think it picks up something... but that something is so prone to being outweight buy other stuff that it makes it a dangerous tool for evaluating individual players.

 

[johnston797]

Take a look at last year's again. I'm not sure I'd buy your argument that Butler and Hinrich were the only guys you were pretty sure where mischaracterized. The following are guys besides Kirk and Butler who had a net +/- less than -2:

OK - I forgot that I didn't look at the whole league when I was trying to figure this out a month ago. Too many clicks. Drives me nuts about this stuff. If it's so critical, shouldn't it be easier to review. Just like PPG.

Curry, Marquis Daniels, Yao, Wright, Wells, Posey, RJeff, Crawford, Nazr, Dwight Howard, Leandro Barbosa, SAR, Mobley, Brent Barry, Devin Brown, Reggie Evans, Nazr, Jalen Rose, Jarvis Hayes

I wouldn't definetely say that Curry, Marquis Daniels, Wright, Wells, Posey, Crawford, Nazr, Leandro Barbosa, SAR, Mobley, Brent Barry, Devin Brown, Reggie Evans, Nazr, Jalen Rose, Jarvis Hayes provided huge value to their teams. As compared to the alternatives.

RJeff was hurt so his data is incomplete.

I would add Yao and Dwight Howard to the list of Kirk and Butler. Not sure about SAR.

 

[bullsville]

I can tell you that with the new game charting method Roland is using, it tells exactly who is guarding the offensive player that ends the possesion (by shot or bad pass), which should make his defensive stats much more accurate.

 

[johnston797]

I can tell you that with the new game charting method Roland is using, it tells exactly who is guarding the offensive player that ends the possesion (by shot or bad pass), which should make his defensive stats much more accurate.

I don't think they have released the new defensive stats that you are refering, too. It's definetely not part of the New Roland Ranking links in this thread.

 

[bullsville]

I don't think they have released the new defensive stats that you are refering, too. It's definetely not part of the New Roland Ranking links in this thread.

No, you're 100% correct there. The info on who is actually doing the defending (as well as any help defenders or double-teams) is currently being charted, but he isn't ready to start compiling and 'interpreting' all the info yet.