So, friend of the blog, Andrez Alvarez asks:

*Your entire player ranking thing is too cool. I feel it was hidden in your article. IMO that could be it’s own post.*

Done and done

*.*Enjoy.

**Note: I updated this to make it more readable and sorted by final minute projections. Click to see full size.**

Arturo,

Awesome work! Thanks for listening to the suggestion.

Wait, what happened to chris paul’s minute projections?

I manually adjusted him back to 2008 & 2009 levels. I modeled all the minutes then checked them by hand using depth charts from ESPN and Sham Sports. Fun way to spend my evening.

this list is missing Iverson in the turkey league.

Arturo, I am dissapoint.

I see that Camby is expected to be the 2nd most productive player, which raises an important question I’ve long had about WP. A big part of Camby’s WP value comes from his rebounds — more than half of his value above an average player. Dr. Berri’s theory is that each rebound equals roughly one extra possession for his team (he says there are some diminishing returns in the NBA, but this is at most a small factor). But here’s my question: why is it that Camby’s teams do not actually appear to get Camby’s extra rebounds?

Let’s consider Camby’s last 6 seasons in which he played for a single team and was a dominant rebounder, from 03/04 to 08/09. His Reb48 was 16.8, so if we treat him as a PF then WP says he was adding 5.4 rebounds for his team. Over six seasons, this means Camby should have added 1,472 boards above average for his teams, almost 250 per season. And how many rebounds were these teams above average? Actually, they were below average at -229 (35 rebounds below average per season). So Camby’s teammates collectively were 1,700 rebounds below position average over these six seasons — an incredible 3.5 per game (equivalent to losing the production of one average NBA player).

Now, any one great rebounder might be saddled with below-average teammates just by chance. But if Dr. Berri is right, almost half of all the great rebounders should have teammates who are average or better rebounders, and so the team’s net rebound advantage should be at least as large as the individual rebounder’s. But as best I can tell, every great rebounder has had the same misfortune as Camby — they ALL have far below average teammates. So my question is: can anyone find some great rebounders (let’s say someone who is 3 Reb48 above position average) who actually seem to have added those rebounds to their team total? I don’t think this creature exists.

It’s also notable what happens when Camby joins a team. His first full season in Denver the team was -26 (below average), while the year before Denver was +85 — so team rebounding declined. Similarly, when he joined the Clippers team rebounds declined (from -141 to -174). So adding Camby’s 16.8 Reb48 coincides with a drop in rebounding.

I don’t see how we can square these numbers with the idea that players with high rebound totals are really adding some large numbers of possessions for their teams. Thoughts?

I haven’t checked your calculations, but even if they are true, you’re making a big leap based on one example. “But as best I can tell, every great rebounder has had the same misfortune as Camby ā they ALL have far below average teammates.” Have you looked at anyone other than Camby? As you say, it could just be that Camby “might be saddled with below-average teammates just by chance.”

I’m sure a great rebounder takes away some rebounds from his teammates, just as a great scorer takes away some scoring from his teammates. But from what I understand, multiple great rebounders should be able to co-exist and make their team better.

Let’s look at another example. For years, Kevin Garnett was a great rebounder on bad teams. Then he joined the Celtics, and the Celtics became a great rebounding team. How does that compare with your story about Camby?

That would be true if it were only one example. But it’s not — it’s true of all high-volume rebounders: Rodman, Wallace, Garnett, etc. Feel free to check any and all such players. I cannot find a single high-volume rebounder — defined as 3 rebounds/48 above position average — whose teams came close to realizing all those rebounds. So consider it a friendly challenge: find a few.

BTW, Garnett did NOT put up huge Reb48 numbers in Boston — it was quite a drop from his Minn performance. However, Garnett did put up impressive Reb48 in Minn — for example, WP says he added 262 rebounds for Minn in 2006-07. But the team was 51 rebounds below average that year, meaning his teammantes were more than 300 rebounds below average, almost 4 per game! And what happened when Garnett took all his rebounds to Boston? Minnesota became a better rebounding team!

Guy,

My short answer to this is that the correlation to winning for all those guys is very high. Rodman,Wallace, Garnett & Camby (and we could add Barkley, Malone and Howard) have very definetely moved the needle strongly to the positive in terms of wins. Add any of these guys to a team and they improve and it something that keeps repeating time and again. So the exact impact to the team bottom line stats isn’t as important as the impact to the win total. Above average rebounders correlate to winning.

BTW, this also holds true for very low-rebound players. They do not appear to lower team rebound totals by anything close to what WP predicts. Bargnani is a good example. Last year WP says he cost Toronto 175 rebounds, compared to an average forward, but Toronto was only 37 rebounds below average. The year before he was -167, but the team was just -60. Just as all “great rebounders” have below-average teammates, all “bad rebounders” seem to have above-average teammates. This can’t be a coincidence. Pretty consistently, each marginal rebound at the player level — up or down — appears to result in somewhere between .2 and .3 rebounds at the team level. So it looks as though the high-reb48 players are mostly (not entirely) taking rebounds from their teammates rather than the opponent, while low-reb48 are mostly surrendering rebounds to teammates rather than the opposition.

i really like your train of thought here, Guy.

i wonder, when looking at a low rebounder like that italian canadian, if his negative affect is shown in activities around rebounding, like fast break chances and transition defense. if wings and guards are having to use more energy and focus to rebound, is this limiting their chance to break down the court or to get back on D? this also opens up the idea of tiered value on the court; a team must secure the ball to take next steps…i’m babbling.

great comments Guy and great work Arturo.

Arturo, what is the evidence that high-rebound players correlate to winning? That certainly hasn’t been true for David Lee. Or the evidence that adding high-rebound players improves a team? I don’t know of any study that shows this. Have you done one? The correlation between reb% and winning is actually pretty weak at the team level, so I’d be surprised if there is much correlation at all with having a high-rebound big man.

And even if there is a correlation, how strong is it? WP says clearly that every player rebound adds one rebound to the team, and 33 rebounds equals one win. Is there any evidence showing that this 1:1 relationship is true?

I’m a little surprised you are so unconcerned about the failure of these high-rebound players to actually improve team rebounds. This seems like an issue of fundamental importance: if these players aren’t creating possessions as WP claims, how do we know whether — and how much — they are contributing to winning? I think this is one of the critical issues for you and the WOW network to address. As I’m sure you know, concerns about how WP treats rebounds (in addition to the usage/efficiency tradeoff ) is what prevents some people from accepting the WP method. If you can show that rebounds are properly valued — or alternatively, modify WP to correctly value rebounds — that would be an incredibly important contribution.

Guy,

The correlation for WP (and WS and +/- as well) to wins is well documented. Prof. Berri is publishing a paper on this soon I will spend some time on the issue once it comes out. I’m reminded though of a great quote from Oliver Heaviside:

Why should I refuse a good dinner simply because I don’t understand the digestive processes involved?. It’s a little disingenious to point to total rebounding stats and ignore the effect of the rest of the team around the player. The math is a little more complex. I’m feel comfortable with saying that great rebounders help your team.The David Lee comment is somewhat unfair considering he was on a horrible,horrible team. Let’s see how this season plays out.

But Arturo, WP is not based on the idea that high-rebound players help teams win because of some ancillary, unmeasured gains. It’s based on the finding that each extra rebound at the team level adds one point (1/33 of a win) — which is true. Then an assumption is made that each rebound captured by a player results in one extra rebound for the team. This second step is simply an assumption, based on no additional research I’m aware of. If you don’t think that this assumption is true, then that creates a very serious problem for WP, as I’m sure you understand.

The correlation of WP to wins at the team level is of course neither here nor there. Team WP translates into point differential, so of course the correlation is very high. A metric that simply allocated the point differential among players based on minutes played would also have a .95 R^2 — but that wouldn’t make it a good metric, right? The question is correctly apportioning credit among a team’s players, and in doing that WP assumes one player rebound = one team rebound.

David Lee is just one example, of course. My point is only that high-rebound players are found on both winning and losing teams. It’s not self-evident they are a big factor in winning.

So I’m puzzled by your saying “I feel comfortable” with a conclusion for which there appears to be no evidence at all. That doesn’t seem at all consistent with the rigorously evidence-based approach I thought has really been one of the hallmarks of your blog.

Guy,

To clarify, I have seen very good evidence to support this phenomena. Some of it has not been published so I feel uncomfortable getting into it until that happens. I will spend some time on this in the future as it’s a valid question. The key assumption is that players contribute to wins by increasing productivity for the team above average levels and your question (can a player generate numbers individually and decrease those numbers for the team) is a valid concern. I have seen evidence to the contrary and I promise we’ll revisit this in the future.

Oh and the one to one rebound correlation is not really correct. It’s one rebound above what the average player would generate for the same position (Same for points,blocks,steals etc.). WP makes a linear assumption which holds true in the majority of cases but the edges will always reflect diminishing returns.

“WP makes a linear assumption which holds true in the majority of cases but the edges will always reflect diminishing returns.”

OK, sounds like we agree that WP assumes a mainly linear relationship. If a player contributes 3 rebounds more than the average player at his position, the team should then have approximately 3 additional rebounds. Is that correct? (I understand that a team with 5 Rodmans cannot grab 130% of all rebounds. We’re talking about the normal range of teams here.) What I’m saying is that this is not even close to being true. When a PF posts an reb48 of 14.4, his team never ends up being +3 rebounds per 48 minutes. Same at other positions. I don’t expect you to share private research, but I just don’t see how the linear relationship can possibly be true in general, if it isn’t true for a single player. (That’s like the old joke: we lose money on each sale, but make it up in volume.)

Never is a very strong word, Guy. Very limited sample, but look at GSW for the preseason, with 3 rebounders at higher rates. Biedrins is good – but it would be very hard to argue that Lee and Adrien haven’t added significantly to the W’s rebounding.

Guy, I think Arturo has done a pretty good job of explaining himself – you might have found one section of the model that doesn’t “add up” when looked at in isolation, but this section is an important part of the model that predicts accurate results. There may be some minor adjustment that can be made from what you have pointed out, but at the end of the day, the model is the model, and it is extremely accurate.

sorry substitute the word “predicts” with the word “produces” in the middle of the above comment.

Raspu10: I’suggesting that no player ever improves his team’s rebounding. Of course that must be true. The question is whether WP is correct in saying each extra rebound above average adds a rebound for the team. Let’s take Lee: WP says that Lee adds about 4 rebounds every 48 minutes he plays. If you want to wager that GSW will improve by 4 rebounds for every 48 minutes Lee plays, I’ll gladly take that bet.

Jimbo: I don’t think you’re following my point (or I haven’t been clear). This isn’t a minor or technical issue, it’s a crucial assumption of WP, and Dr. Berri has argued repeatedly that it is true. If it’s not true, the way WP allocates wins among players needs to be overhauled, perhaps significantly. And the difference between what WP expects and what we observe isn’t small: it looks to me like the team rebound totals implicitly projected by WP are wildly innaccurate. But please, prove me wrong. Won’t anyone try to find some great rebounders who have added the predicted number of boards to their team total?

And I’m not sure what you mean when you say the model is “accurate.” We know that it adds up at the team level to match point differential. But we already know how many wins each team had — we don’t need WP to “predict” that for us. What we want to know is if it divides up credit correctly among players. And the only way to test that is to predict future performance, especially for teams that have added and subtracted players.

Oops: first line is “I’m not suggesting…”

[…] Galletti projects every single player in the NBA for 2010-11 and provides win predictions for every team. He also […]

Hey Arturo,

Thanks for the chart its awesome. Anyway you can release it as a html table or spreadsheet or something so we can sort it ourselves. I would love to sort based on most WP.

Thanks,

Brandon

No problem.

go here

[…] a friendly competition; Who could predict the regular season best? Arturo showed off a little by modeling the entire NBA! Yesterday I reviewed the Eastern Conference, today we’ll go over how the […]

[…] a friendly competition; Who could predict the regular season best? Arturo showed off a little by modeling the entire NBA! Below we go over how the Wages of Wins Network analysts (we need a short nick name, […]