“Insanity: doing the same thing over and over again and expecting different results.”-Albert Einstein
I wrote in a private e-mail today:
“Yesterday’s piece was one of those pieces that just happen. I’d been working all weekend on some consulting work I need to hand in and when I went to write the Around the Wow piece the quote “oh the places you’ll go” popped into my head. Before I knew it, I was looking up Dr. Seuss books.
For some reason, my spontaneous/unplanned stuff can be really good (the last 4 pieces were all like that), whereas some of the more thought out stuff leaves me unsatisfied. Weird.”
This is fortuitous because two days ago I wrote a post on the 100 worst seasons/players since 1978. This inspired some familiar responses.
Now the studies referred have some technical problems ( If you run a regression you must account for other variables that could affect your result). This is why you’ll see me control for small samples (minutes), age, position and teams frequently. Prof. Berri’s covered this before in the comments here http://dberri.wordpress.com/2010/05/01/ted-leonsis-endorses-stumbling-on-win or very thouroghly in Stumbling on Wins just buy the book). I’ve covered this before (see The Basics , and here ).
I pulled every players rebounding numbers from 1979 on (for every player with > 800 minutes played, see controls).
I looked at every player’s:
- offensive rebounds per 48 minutes (orb48)
- defensive rebounds per 48 minutes (drb48)
- total rebounds per 48 minutes (trb48)
- And for Grins and giggles, looked at all players and a subset of player ages 25-30 (see more controls, this time for age).
For your amusement that table is here. (This also allows for peer review of my findings).
My results? Here:
Do rebounding numbers fluctuate wildly based on teammates and diminishing returns? Short answer is no. Year to year correlation for total rebounds per 48 minutes is at 91% regardless of the age grouping. In fact, 82% of the population in the sample had an absolute error in year to year rebounding of less than 1.5 rebounds. If I break out the Minitab as well and graph it up:
And that looks like the very definition of a normal distribution.
So Wins Produced consistent season to season? Check
Rebounds not varying radically? Check
Unnamed individuals pestered into writing additional material on this? Check and check (You’ll just have to wait to find out).
Shareef
11/23/2010
What Team would replicate the middle clogging players type that have inflated WP#? If your front line was C Marcus Camby, PF Ben Wallace and SF Gerald Wallace and your back court was PG Rajon Rondo and SG Dywane Wade. That team wouldn’t win sh*t but your silly stat. Of course Ben Wallace never gets his rebounds eaten by teammates, If you have Ben Wallace you don’t also sign Reggie Evans
“Spacing is offense and offense is spacing”– Chuck Daily
arturogalletti
11/23/2010
So you’re saying a team with Rondo, Wade, Gerald Wallace, Camby and Ben Wallace wouldn’t win? They might not win a three point contest but I bet they could win the East. Throw in some three point shooters off the bench and underrated defenders (think Posey/House/Barnes/Blake/West types) and you’ve got a championship team most likely.
ilikeflowers
11/23/2010
Why are you wasting your time at a site that specializes in silly stats? Go make use of your life. Also, given that no teams are constructed that way that team would likely be at the margins of the model’s usefulness (then again maybe not). Finally, only an idiot, a scientist, or a mad genius would build that team since you are maximizing diminishing returns on possessions! They’re still gonna clobber many teams, but they’ll likely fail against a more balanced strong rebounding team. Like Arturo said, add in a few average or better scorers and you have a championship contender – with said average or better scorers being the MVP’s of course.
Guy
11/23/2010
Arturo: I don’t understand the point of your post. Has anyone argued that player rebounds are not highly correlated from year to year? I don’t think they have. The argument is that players with high rebounding totals do not actually add as many rebounds to the team as WP estimates. And players with low rebounding totals do not subtract as many as predicted. That’s the argument. It’s fine to disagree with it, but you should address the real argument.
You say Clifford Robinson is basically the worst player in NBA history, having cost his teams a huge number of wins. This is largely because WP believes he cost his teams 3,700 rebounds they would have gotten if Robinson were replaced by an average player. But unfortunately for WP, Robinson’s teams were actually very good rebounding teams on average, more than +1 rebounds per game. That means Robinson’s 4 teammates at any given time were contributing about 5 rebounds above average — I think that’s better than any actual NBA team in the last 30 years. Let me repeat that: according to WOW/WP, Clifford Robinson’s teammates ( just 4 guys at a time), were much, much better rebounders than any actual NBA team. So the worst big man ever just happened to play alongside the greatest rebounding teammates that any player has ever had — on 4 different teams over 15 years. Quite a coincidence, don’t you think?
But the coincidences are everywhere. Dennis Rodman played with about the WORST rebounding teammates any player ever had, players who produced hundreds of rebounds less than expected, again covering many seasons and many different teams. And there are dozens of other examples. Here’s a challenge for you: find me some exceptions. Find me some 15 Reb48 centers or 14 Reb48 forwards who didn’t have below average teammates. I don’t think you can do it.
And this shows you why player’s y-t-y correlation isn’t the issue. Rodman had about the same number of rebounds each season — he was consistent. Robinson too was consistent — always a low Reb48. What changed was the rebounding of their teammates — every time Rodman joined a team, the other players’ rebounds went down (and reverse for Robinson). But since it’s spread around 10 other guys, it doesn’t impact the correlation a lot. (Reb48 also correlates well because the variance is so huge.) But Jefferson’s teammates always ended up with very impressive rebound totals.
But let’s move beyond rebounds. I bet there are large diminishing returns for WP overall. Why don’t you check it out and let us know. Just tell us the correlation between WP at Center and the WP delivered by a team’s 4 other positions. Repeat for each of the 5 positions. Here’s my wager: you will find a large negative correlation — the more WP a team gets from any one position, the less it gets from the other 4. But if WP works the way it’s supposed to, largely separating a player’s contribution from that of his teammates, the correlation will be very low. You have the data, so let’s just settle this debate and move on….
EvanZ
11/23/2010
“So the worst big man ever just happened to play alongside the greatest rebounding teammates that any player has ever had — on 4 different teams over 15 years. Quite a coincidence, don’t you think?”
Guy, did Robinson have rebounds taken away from him everywhere he went? Was he always surrounded by 4 amazing rebounders?
Also, I’ve read criticisms of guys like Rodman, Wallace, Love, Lee, etc that accuse them of taking away rebounds from other players, as if they do this intentionally. But I rarely see this criticism, for example, of Tim Duncan. I looked up the rebounding rates for David Robinson before and after Duncan arrived, and there doesn’t appear to be any significant change in Robinson’s rebounding rate. Isn’t that what you would have expected? I guess what I’m working my way towards here is the following question: Do you think there are cases where specific players *cause* diminishing returns, while others, simply rebound well in their area and don’t cause this effect? I think that is a legitimate possibility, that would change the argument from whether there are diminishing returns, in general, to whether there are certain players that we can measure the effect more strongly/rigorously.
Thoughts from you, Arturo, DJ, anyone else are welcome, of course.
Guy
11/23/2010
Evan: I don’t totally understand the question about Robinson. I don’t think he was really surrounded by “amazing rebounders” on every team — that’s what WOW/WP says happened. I assume Robinson was assigned a certain ROLE on his teams, and grabbing rebounds was not a high priority (for a player at his positions) — that was expected of other players. So when we say players “take rebounds from teammates” it’s not usually literally grabbing the ball from their hands — we mean that some players take a large share of easy rebounds that several players could get, and which the team will usually get no matter which player does it. But yes, I agree with your main point: I’m sure the ratio of actual team rebounds gained to individual rebounds acquired varies by player (and we know it’s very different for OReb than DReb). However, it’s almost certainly the guys with the highest Reb48 who have the worst ratios (and many — other than Rodman — tend to have a larger share of DReb). Measuring this at the individual player level is the “holy grail.”
Arturo: I have made a lot of factual claims here and in other threads. That “great rebounders” always have below-average teammates, while “bad rebounders” have good ones. That winning teams don’t actually get many more rebounds than losing teams. That differences in Reb48 don’t explain much of the difference in WP at the team level. And so on. I hope you will respond to some of those, rather than only responding to a claim that no one has made (that individual players’ rebound totals fluctuate wildly from year to year).
The irony here is that I’m the one taking the “Berri position” on rebounds, not WP/WOW. When fans look at a player scoring 25 points and say “that’s a great player,” WP correctly says “Hold on — we don’t know if he’s a great player until we see how much he improved TEAM scoring. There are big diminishing returns on shooting, and we need to account for that.” And that is all 100% correct. I’m just saying the same exact thing about rebounds, while WOW/WP instead says “Yay! Rebounds!” and ends the analysis there.
ilikeflowers
11/23/2010
So after all this, your point is that teams can mitigate bad rebounding? Of course. And they can mitigate everything else too through effort. We’re talking about comparative value. You are crystal clear on the rebounding issue. You posit it as fact that rebounding is overvalued in wp48. Of course it must be greatly overvalued since reducing rebound values to 0.7 and other coefficients has little impact on wp48 – this was done what 5 years ago? So, where are the identical (hopefully competent) studies on the diminishing returns for all the other stats? Rebounding obviously can only be overvalued when compared to other stats. And just to add to the useless anecdotes, if your team could put forth the effort to minimize its deficiencies in just one area, which one would you chose? Scoring, assists, or rebounds? I’d choose the one with the most bang for the buck and hope that the others didn’t suffer too much. Of course this is why one has to look at everything.
Also, given the apparent superior size and respect that your circle enjoys, where is your player level predictive model? We need to add it to the smackdown. Talk the talk and walk the walk.
Perhaps Dre can drop in your proposed coefficients to see if they improve or decrease the model’s accuracy. Not that that will satisfy the rebounding crowd if it comes out negative.
EvanZ
11/23/2010
Shooting efficiency can’t really have diminishing returns. In fact, it’s likely the opposite. The more good shooters you put on a team, the better for everyone’s percentages, because the defense can’t focus on any one particular man.
Now, usage is another story. I mean, imagine what would happen if AI started playing with Melo. Fortunately,of course, you don’t have to imagine it. Look at the dramatic drop in AI’s usage when he moved to Denver. It dropped almost immediately by 10 %-points!
I’ve looked for similar case studies in rebounding (like the Duncan/Robinson example), but haven’t come across a really dramatic one yet that illustrates the idea as well as the Iverson/Melo example does for usage.
ilikeflowers
11/23/2010
I’m not really talking about shooting efficiency just the basic per minute stats, but efficiency certainly could have diminishing returns. A player who shoots better/worse as they shoot more/less or play more/less minutes would be one example (tough to demonstrate causality though – how do you control for good coaches minimizing a player’s minutes when they aren’t playing well?). This also touches a bit upon the hot/cold shooter issue that IIRC you were looking at recently.
Really, my point is that just like individual stats, team stats (and the relationships between team stats and individual stats) may be dependent upon one another and therefore one cannot cherry pick stats for their regressions and then reasonably claim without qualification that there is a problem with some model. All the possible relationships must be addressed. Is it really that hard? All Guy has to do is include the rest of the stats in his regression (in the same manner) and run em again. If that shows the same dramatic overvaluation (I think dberri has even gone down to 0.5 on the boards) then he’s got something, otherwise not so much.
EvanZ
11/23/2010
“A player who shoots better/worse as they shoot more/less or play more/less minutes would be one example (tough to demonstrate causality though – how do you control for good coaches minimizing a player’s minutes when they aren’t playing well?)”
You’re misunderstanding the way I’m talking about shooting efficiency. The point I was trying to make is that if you put 5 60% TS shooters on the floor at the same time, there’s no reason to believe that there would be diminishing returns. In other words, they would all still be expected to be (at least) 60% shooters.
That is not necessarily the expectation for rebounding percentage, although it is not as clear cut as usage. Does that make sense? I’m bringing two statistics: one that obviously has diminishing returns (usage) and another that does not (shooting efficiency). Rebounding is more like the former, but the extent to which it is significant is (clearly) debatable.
ilikeflowers
11/23/2010
No, I understood you. My point is that it is possible that there are players whose efficiency (in any way that you wish to define it) can vary (predictable) with the number of shots taken and/or minutes played. I would expect the same results from that team that you do, however I can think of circumstances in which some of those players would experience a reduction in efficiency. If they all shot the same kind of shot best and some of them experience a change in their shot type selection for example.
EvanZ
11/23/2010
If your model for wins showed that usage and shooting efficiency were correlated, then I agree. Of course, the model would need to have data to support that assumption.
nerdnumbers
11/23/2010
Ilikeflowers,
An annoying Dave does is refer to old posts/books when I send him questions some times. In Stumbling on Wins (free on Kindle this week), Dave does in fact alter the rebounding coeffs and while it will change the value, it does not alter the players ranking. IMO a lot of the rebounding/team parts of the WP are people that don’t get the stats but don’t like the results. As such they point at the math they don’t like or understand and say “the problem is there!”
ilikeflowers
11/23/2010
Yeah, I imagine that he would find it more annoying than anyone. What gets me isn’t raising rebounding (or any other) questions about the model, it’s the stating of opinions or concerns as facts. I have usage questions, and I don’t think that enough positions are being used, and I have concerns about the adequacy of the defensive stats, marginal returns, over under performance of teams relative to individuals, etc… but I don’t run around bluntly stating that the model is definitively wrong about this or that possibility because I ran a regression on a few stats.
Guy
11/23/2010
Ilikeflowers/Nerdnumbers: Are you refering to Dr. Berri’s post where he says that changing the rebound coefficient to .7 produces a .99 correlation for Win Score? Don’t you think it’s an odd argument to say both “it makes no difference what coefficient I use for rebounds” and also “my coefficient is correct, and it’s important to keep using that coefficient?” There are about 6 reasons that study doesn’t resolve the issue. But in any case, let’s have Arturo run the numbers using .7 for OReb and .3 for DReb, and see what the correlation is for WP48 among players with at least 1500 minutes. I think you will see a signficant change.
nerdnumbers
11/23/2010
Guy if I earn $1 dollar for every $2 dollars you earn and that is stable over time then our earnings will have a 1.00 correlation. In short I can look at how much I’ve earned and know with certainty how much you’ve earned. However if someone comes up to me and says “How much has Guy earned?” and I give them the amount I’ve earned I will be wrong. The goal is to get an accurate win estimate, not move the coefficients around to make the experimenter happy.
EvanZ
11/23/2010
There are a few issues with Robinson that may simply make him an outlier, and not a victim of diminishing returns:
1) He may have actually been a bad rebounder.
2) He may have “learned” to be a bad rebounder by playing on those Portland teams that actually did have a ton of good rebounders (Williams, Kersey, Duckworth, and even Drexler).
3) For purposes of WP, are we really classifying him as a PF/C? Wasn’t he more of a SF/tweener? If we simply change his position to SF, the WP will change dramatically. He probably still won’t be considered good by any means (not with a 50% TS), but we may not come out as the “worst rebounder of all time”.
EvanZ
11/23/2010
*he* may not come out as the worst rebounder.
(We probably would be.)
Guy
11/23/2010
Evan: this could all be true. My point isn’t that Robinson wasn’t a weak rebounder and didn’t cost his teams any rebounds at all — I assume he was and did. But what can’t possibly be true is that Robinson cost his teams 3,700 rebounds, or anything close to that. He’s just an unusually powerful example of how WP mishandles rebounds. And the goal here is not to just identify players as “good” or “bad”, but to quantify their contributions — right?
EvanZ
11/23/2010
It could simply be a case where we are miscategorizing a player.
(Can’t someone go back and look up old box scores and find what position he was actually listed at?)
Bill Gish
11/23/2010
Arturo, I think that you are being entirely too dismissive of Shareef’s point.
The way I would put it is that, if it is as self evident as I think it is that defensive rebounding is sharply subject to diminishing returns, then, of course, NBA GMs and coaches don’t start a lineup with Ben Wallace AND Reggie Evans.
Therefore, all your amazing stats do is reflect the fact that a good rebounder on one team will likely be a good rebounder on a new team given that the new team presumably acquired our hypothetical good rebounder because they needed one, and he isn’t redundant.
I know from listening to you and your friends podcast that you guys sometimes resort to the evidence of your eyeballs watching games. It seems to me that if one watches basketball at any level we’ve all seen players deferring to teammates to the point of almost literally letting them take the ball out of their grip and thereby inflating one players rebound numbers at the expense of the other and providing visual evidence of diminishing returns.
It also seems to me that the better counterargument might be that this argument over diminishing returns is trivial. There are arguably diminishing returns on every stat, which fact may be enshrined in the old observation that:
There’s only one basketball.
Fred Bush
11/23/2010
Guy, wins produced is ultimately about *wins*. If you want to argue that rebounds are overrated, you need to show that teams that bring in terrible rebounding big men who can shoot like Cliff Robertson get more *wins* with him than without him, and that teams that bring in incredible one-dimensional rebounders like Rodman do not get more wins than without them. That’s what’s gonna persuade us, and I don’t think you’re going to find that data (because, among other things, if you outrebound the other team by 3 in a game, you will win 68% of the time).
Fred Bush
11/23/2010
*Robinson
Arturo, you def. need an edit button
Guy
11/23/2010
Fred: We don’t disagree about the importance of rebounding at the team level at all. A +3 rebounding team will win a ton of games. The problem is, there generally aren’t any such teams. Last year, the best teams in the NBA in WP were about +.8 rebounds/game on average, which is worth about 2 wins. So far this year, we see the same pattern. So the obvious question is: why can’t we find +3 teams?* If there are really +4 and +5 players, then there should be a lot of +3, +4, and +5 teams. But they don’t exist. Why? Because of diminishing returns, because these players don’t really add that many rebounds at the team level. The math is very straightforward here.
* Historically, there may be some +4 or +5 teams. I haven’t investigated that. But in today’s game, the range of rebounding among teams tends to be very narrow.
Gabe
11/23/2010
@Guy:
“A +3 rebounding team will win a ton of games. The problem is, there generally aren’t any such teams.”
http://sports.yahoo.com/nba/stats/byteam?&cat1=Total&cat2=diff&conference=NBA&year=season_2009&sort=233
EvanZ
11/23/2010
The Warriors differential was -10 last season…is it a coincidence that the team had not a single legitimate big man and was playing “small ball” essentially the entire year? And where did those 10 rebounds go?
bduran
11/23/2010
Guy,
There are several +3 rebounding teams almost every year. Adding good rebounders certainly shot the Spurs rebounding differential up. In ’08-’09 they were +.6. Next year they were +3.3. This year they are +3.6. I would bet getting Blair and McDyess have something to do with that even though the Spurs already have a great rebounder in TD.
Guy
11/23/2010
Gabe: great question. You have to adjust for opportunities, since good teams have relatively more dreb opportunities (they shoot well, and make opponents miss). Take Cleveland, which you show as +4. On offense they were .251, compared to .263 for the league (-.012); on defense they were .772 compared to league average of .737 (+.035). Assuming 42 chances on each end, that’s a net gain for Cleveland of +0.97 — basically, +1 rebound per game. You can see reb% here: http://www.basketball-reference.com/leagues/NBA_2010.html.
bduran
11/23/2010
Back to my Spurs example then. They had a 50.4% rebounding rate in 2008-2009. They lost one solid rebounder (Kurt Thomas) and one not great rebounder (Oberto) from the front court. They gained McDyess and Blair who are very good rebounders. They replaced all the minutes lost and took some of Bonner’s (not a good rebounder). The end result was a jump from just above to average to one of the best rebounding teams in the NBA with a rebounding rate of 52%. So it would seem that,at least in this one case, adding good rebounders did in fact increase team rebounding.
EvanZ
11/23/2010
One way that we could go about looking at diminishing returns is to do something similar to what gave rise to the “Millsap doctrine”. Look at teams that lost a starter at some point during the season (maybe after the trade deadline so that no additional moves could be made), and find the difference in rebounding rates of the replacement and his teammates before and after the injury. Has anyone done this study or heard of something like this being done?
ilikeflowers
11/23/2010
Just make sure the analysis is for all the stats (together), not just rebounding. If a team loses a great rebounder and they minimize the damage by having a player or two chip in more on the boards, are other things suffering (assists, steals, shots, efficiency, etc…)? Conversely, if a team loses a great scorer, assists man et al.
EvanZ
11/23/2010
” If a team loses a great rebounder and they minimize the damage by having a player or two chip in more on the boards”
If teams could do this, why wouldn’t they do it all the time?
ilikeflowers
11/23/2010
Isn’t that the entire point of Guy’s rebounding argument? At the team level rebounds are easily replaced by other players and are therefore overvalued at the individual level? If the simple conclusions of the rebounding regressions are reality then isn’t this what is happening?
EvanZ
11/23/2010
exactly – so you and I *wouldn’t* expect other players to get more rebounds than they already were (before the replacement), but Guy would. Right?
ilikeflowers
11/23/2010
My opinion is that a team can mitigate the loss of a great rebounder or scorer or assister to some extent without replacing them with another great rebounder. I also think that this cannot be done without consequences elsewhere. If you’re fighting harder for a rebound than usual then you may be less able to help your team in other ways than usual. Of course this is where a great coach might help with a magic scheme or lineup that maximizes rebounds without hurting anything else. Ultimately the question is to what extent individual loss can be minimized by the team. I think that wow’s position with rebounding is that it can be mitigated, but in general not very significantly so. Until I see something to convince me otherwise I’m fine with that answer.
arturogalletti
11/23/2010
Boy, you’ve all been very busy.
Interestingly so have I. Monster Post coming tonight.
Westy
11/23/2010
It was noted, “…since reducing rebound values to 0.7 and other coefficients has little impact on wp48 – this was done what 5 years ago?”
What does “little impact” mean here? A high correlation remains? That may be true and yet there may be significant changes in players’ rankings. Some may not call that ‘little’ impact.
Further, even if DREB are consistent, how is that an argument for full accreditation to an individual player? Team average shooting rates are also fairly consistent, and thus the overall opportunities for DREB are fairly consistent. But, as Guy points out, if there are more misses, there are more rebound opportunities. And having a high number of DREB is explicitly linked to good defense. So couldn’t an argument be made that since the number of misses caused by a player on D remains consistent, he’s consistently contributing to others’ ability to garner DREB, and so he should share the credit?
At the team level, we know that valuing the rebound at about 1 regesses to team wins well. But if each player who grabs the ball gets 0.6 credit for that, and each defender (who argubly helped cause the miss) gets 0.1 credit, the team credit remains the same. And I think we’d see obvious shifts in the rank of players accordingly. And I also think this remains consistent with individual rebounds being consistent.
In Robinson’s case, isn’t it possible he contributed to his teammate’s rebounding totals by being a good defender and causing more misses? Were his teams consistently above league average in FG% defense?
EvanZ
11/23/2010
I think there are two separate issues that are being conflated (repeatedly), so let me state them, and if anyone disagrees, feel free to followup:
1) Is DREB% (OREB%) a “real” (i.e. intrinsic) quality that a player possesses – and thus, will carry from team to team.
2) The weight of a rebound in terms of actual wins.
Whether (1) is true or not, doesn’t really imply one way or the other the value of the weights in (2) or the importance of rebounding towards winning.
EvanZ
11/23/2010
Westy, I meant to write a reply to the post, not your specific comment, so it’s not directed toward you specifically.
Westy
11/23/2010
No worries, Evan. And I think you point out a worthwhile observation. I would say that, yes, DREB% could be intrinsic, and yet, if a rebound is not ‘weighted’ correctly, that not be reflected in terms of wins correctly.
The value of 1 additional rebound to a team, is I think somewhat more established. But who should individually get credit for having garnered that rebound? I don’t think the case is solid that it should only be the player who grabbed it. And thus, evaluation of the # of potential rebounds available versus those grabbed (DREB%) is valuable. And if the fluctuation in rebounds is just due to the fluctuation in those available, it would follow to me that those helping cause that fluctuation in availability should also be getting credit for the rebound.
So basically, if your DREB% is a constant 20.0%, your rebounds will still vary depending on the level of defense your 4 teammates offer.
Do we know that DREB% is more consistent than total DREB volumes?
Carlos_XL
11/23/2010
I just want to add that Guy makes some very good points, and most importantly, offers a clear and concise means of testing his argument. As your Mythbusters cartoon shows, thats the very definition of science, and I just want to commend him for adding to the discussion in a meaningful way, since so many other posters have been dismissive.
some dude
11/23/2010
yeah, i don’t buy into the claim about expected shifting numbers.
1. Reb48 is not rebound %
2. Why would we expect wild fluctuations? This assumption has two possible problems.
a. turnover. And by this I mean, team player changes. How often do teams change starting quality big men, especially up until the mid 90s? If teams aren’t changing their core groups much over time, then how could there be wild fluctuations?
b. With any turnover that does happen, how do we know who is being replaced? In fact, I would argue that the overwhelming majority of most players are about the same in terms of rebounding for their position. That means that the overwhelming majority of players being replaced are being replaced by someone of the same rebounding ilk.
c. There are only so few rebounds that can be “fluctuated” so that these results could be in favor of the claim. How many rebounds can an above average rebounder actually change on average? 1? 2? The differences are small to begin.
3. If I understand Guy’s argument correctly, then if you did the same thing with team rebounding numbers you’d find very little fluctuations in the numbers which would mean team rebounding is largely unaffected over time regardless of who they employ. If that’s true, and if the majority of all players are about the same in rebounding terms for their position, then we would expect the opposite – only small fluctuations- in individual rebounding numbers.
It is only at the ends of the curves that we would find the results you’re looking for. When you add an elite rebounder to a team, the other teammates getting worse and visa-versa. You couldn’t possibly see it by regressing over all players.
We have to see what happens when we add a great rebounder to an above average team and the reverse most specifically. This will give us what we need, IMO. Putting a great rebounder on a horrible rebounding team will obviously improve rebounding without hurting other rebounding numbers (since they can’t get hurt any more to begin with), so that doesn’t help much, either.
some dude
11/23/2010
To give an anecdotal example for Westy’s claim:
Last year Lamar Odom and Pau Gasol were the team’s two best rebounders on the Lakers. Despite this, the best defensive rebounding lineups had a Bynum-Gasol or Bynum-Odom tandem, not Gasol-Odom tandem. In fact, Bynum lineups as an aggregate were better defensive lineups than either Gasol or Odom lineups.
Curious, no?
Having watched the Lakers, I can tell you why. Bynum’s defensive impact. While he rebounded less than Gasol/Lamar, he altered more shots for easier rebounds and his boxing out allowed either Pau or Lamar (or others) to get easier rebounds while the Gasol-Odom lineups didn’t fare as well in those departments.
Not that the percentage was a huge change, but it was still there. So I do agree the credit for a rebound isn’t given out accurately.
EvanZ
11/23/2010
Team DREB% is highly correlated with DEFF and OPP eFG%. I’ve been wondering whether that correlation could be due to the fact that, in general, better rebounders are also better defenders, or as some dude suggests, better defensive players create more easily “reboundable” shots for themselves and their teammates. I’m thinking some of both.
bduran
11/23/2010
EvanZ,
I’ve wondered about this myself.
some dude
11/23/2010
most likely it’s as you said Evan, some from column a and some from column b.
http://dream-analysis.org
03/20/2012
I got what you intend,saved to bookmarks, very nice web site.