After Friday night, every single team finally played its 41st game in the 2009-10 season. And that means that I am able to evaluate every player on ever team at the season’s midpoint. As I noted last year, many stories could be told with this data. But it’s probably best to hold off on the stories and get right to the data. So without further introduction, here is the Wins Produced and WP48 [Wins Produced per 48 minutes] of every player at the midpoint of the 2009-10 campaign.
Table One: All Teams and Players at the Midpoint of the 2009-10 Season
Again, I am sure anyone looking at this data could tell a few stories. Here are some stories that come to mind when I look at this data.
- Clearly this is NOT the year of the Super Teams. Here is what I said last October: It does seem quite possible that at least one of the four best teams in the West will win at least 62 games. And so it’s possible at least four teams (Cleveland, Orlando, Boston, and at least one Western team) will win more than 75% of their games; making the 2009-10 the year of the Super Teams. At this point, not a single NBA team is on pace to win 62 games. Part of this is due to injury. Part of this is due to declines in player productivity. And so far, all of this says I was apparently wrong last October.
- Although the final record forecast was flawed, many of the teams people thought would be lead the league are indeed leading the way. The top teams in the NBA – in terms of Wins Produced and efficiency differential – are the LA Lakers, Cleveland Cavaliers, Boston Celtics, Atlanta Hawks, and San Antonio Spurs. Each of these teams has posted a Wins Produced mark in excess of 27.5, which means at the midpoint of the season each team is half way to 55 wins. The list of teams that is half way to 50 wins includes Denver, Orlando, Portland, and Utah.
- Shifting our focus to the players… the Most Productive Player in the league at the midpoint is LeBron James. King James has already produced 13.8 wins this season, and he is on pace to finish with 27.7. This means that LeBron has already produced more wins – and is on pace to finish with more Wins Produced – than the Indiana Pacers, Minnesota Timberwolves, and New Jersey Nets.
- LeBron is obviously the most productive small forward. Here are the most productive players at each of the other four positions: PG: Chris Paul (9.8 Wins Produced), SG: Andre Iguodala (8.0 Wins Produced), PF: Marcus Camby (11.8 Wins Produced), C: Dwight Howard (10.0 Wins Produced).
- The top rookie thus far is DeJuan Blair, who has produced 4.17 Wins Produced. Tyreke Evans, though, has produced 4.16 Wins Produced. Obviously Blair is more productive on a per-minute basis.
- Turning our focus to the least productive players in the game we see that Ryan Hollins (-1.9 Wins Produced) is leading the way. He is followed by Darius Songaila (-1.5 Wins Produced). The least productive rookies include Earl Clark (-1.4 Wins Produced in only 249 minutes) and Jonny Flynn (-1.3 Wins Produced).
Those are just a few of the stories that I noticed when looking at this data. As the complete rosters for the All-Star game and Rookie Challenge are announced, I will again return to this data.
Let me close on a completely different and far more important topic. The recent earthquake in Haiti has left more than 100,000 people dead. A donation to the Red Cross — or other worthwhile charity – would certainly help out those who survived. So please give if you are able.
– DJ
The WoW Journal Comments Policy
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
Lior
January 24, 2010
It’s important to note that the “simple projection” does not take into account injuries and suspensions (for good reason). In other words, anyone who wants team-level projections needs to do to more with the data.
For example, the Blazers are not really expecting Oden and Przybilla’s production for the second half of the season to match their production in the first half. Same with the Wizards and Arenas.
Lior
January 24, 2010
Notable: Per-minute Marcus Camby and Kevin Love are approximately as effective as Lebron (they have slightly higher WP48). In other words, it is easier to be an outlier as a center than as a SF.
Jason Kidd at age 37 is offering 90% of the productivity of Chris Paul at age 25.
docrostov
January 24, 2010
If you arrange the data by WP48, and remove all players that have played in either less than 10 games or less than 100 minutes, the top five for each position (as teams have designated to ESPN) are as follows (with their WP48 and ranking among all players according to WP48 in parenthesis).
POINT GUARD
1. Chris Paul (0.372; 4 overall)
2. Jason Kidd (0.314; 11 overall)
3. Steve Nash (0.296; 12 overall)
4. Rajon Rondo (0.276; 14 overall)
5. Deron Williams (0.216; 39 overall)
SHOOTING GUARD
1. Mike Miller (0.269; 19 overall)
2. Manu Ginobili (0.266; 21 overall)
3. Dwyane Wade (0.245; 25 overall)
4. Andre Iguodala (0.234; 29 overall)
5. Kyle Korver (0.220; 34 overall)
SMALL FORWARD
1. LeBron James (0.420; 3 overall)
2. Gerald Wallace (0.344; 7 overall)
3. Kevin Durant (0.240; 28 overall)
4. Jeff Pendergraph (0.221; 33 overall)
5. Paul Pierce (0.209; 39 overall)
POWER FORWARD
1. Marcus Camby (0.448; 1 overall)
2. Kevin Love (0.431; 2 overall)
3. Pau Gasol (0.359; 5 overall)
4. Tim Duncan (0.357; 6 overall)
5. Troy Murphy (0.315; 10 overall)
CENTER
1. Dwight Howard (0.341; 8 overall)
2. Greg Oden (0.316; 9 overall)
3. Joakim Noah (0.279; 13 overall)
4. Ben Wallace (0.274; 16 overall)
5. DeJuan Blair (0.273; 17 overall)*
* – In most games I’ve seen, Blair is listed as the Spurs center.
Jed
January 24, 2010
At first glance, there seem to be a lot more players with a WP48 above .300. I believe the stat is scaled each year so that the mean is .100, but it looks to me like the variance has gotten much larger. Any explanations for this if true?
Italian Stallion
January 24, 2010
Jason Kidd generally scores high in this system because he’s an excellent rebounder relative to the average PG (which gives him a lot of credit relative to his peers), but he is not penalized for being a poor scorer (as long he retains average or above average efficiency on his limited usage ).
His poor defense relative to the many other top PGs is also an issue.
There isn’t a snowball’s chance in hell that Jason Kidd is the second best PG in the NBA anymore. Even he would have no problem listing 1/2 dozen guys that blow him out these days.
Sam Cohen
January 24, 2010
I know the Sixers were planning to play Igoudala at shooting guard this season, but I thought that idea was scrapped pretty early on in the season and he’s played primarily small forward. Does anyone know if this is correct?
(Also- what’s going on with Kyle Korver… my recollection is that in the past he never had a very high WP48.)
Thanks for the numbers. Very interesting.
dberri
January 24, 2010
Sam,
I have Iguodala playing both positions, but slightly more minutes at shooting guard. Just checked 82games.com and they have a split also with the advantage at small forward.
If Iguodala is a small forward, than Dwyane Wade is the most productive shooting guard (7.7 Wins Produced).
Justin Wehr
January 24, 2010
Great stuff!
I charted the data and posted it on my blog: http://wehrintheworld.blogspot.com/2010/01/lebron-james-is-officially-best.html. Thought you might be interested.
dberri
January 24, 2010
Justin,
Very interesting graph. And yes, LeBron consistently does more than Kobe.
tywill33
January 24, 2010
DB,
Awesome work! You know what’s funny, I was just going to ask you if Ryan Hollins was the worst player in the NBA… and there you had it already!
Last night the guy fouled out in 8 minutes!!! I was flabbergasted by his incompetence. Can you say “Potsie” Weber??
Oh, and he managed to grab zero rebounds.
dberri
January 24, 2010
tywill33,
Remember, Potsie could actually sing. Although as a friend he was not very helpful.
benamery21
January 24, 2010
dberri: I must be doing this all wrong. I assume the minor differences I get in comparing numbers for most of these players relate to the team adjustment and the slight inaccuracy of adjustment from PAWSmin to WP48. I seem to get significantly larger differences with some players, however. Is the conversion from PAWSmin less accurate for players closer to average for their position? Are you still using the same position adjustments (.225, .215, …)? Would you consider posting a bit more of the data/calc for comparison purposes?
Tindall
January 24, 2010
Interesting that Ariza’s bottomed out. Wasn’t he around .300 wp48 last year and the years before? He’s now getting far more minutes – don’t players usually produce more when they get more minutes?
My explanation for his drop in productivity would be that he’s handling the ball a lot more and is responsible for creating his own shot this year. As such, his assists are up… but his turnovers, shooting efficiency, and rebounds (due to energy expended on the offensive end) are down.
Houston, on the whole, is performing as you expected though – there’s chatter about Carl Landry’s getting 6th Man of the Year. I haven’t heard a lot about Aaron Brooks, but I could be mistaken… so perhaps the media and fans are less braindead than you expected?
I don’t always agree with you, and I think this model has a long, long way to go, but it’s a start in the right direction. I can’t wait for the next book.
evan
January 24, 2010
justin, your post was def worth reading.
Lior
January 24, 2010
Italian Stallion: The calculation of WP48 penalizes Kidd for being a poor scorer (i.e. for scoring less points than other point guards) — this is since WP48 is (ignoring team adjustments) basically the difference between the player’s production and the production of an average player at the same position. Subtracting the production of an average point guard from Kidd’s production includes subtracting the scoring production of that average point guard.
f19ure
January 25, 2010
@Lior
Paul is 24.
One thing worth noting, is that CP’s been hobbling on a healing ankle for much of the season. I’ve watched a good many Hornets’ games on League Pass, and Paul is slow.
He’s still a step above most defenders, but his trademark quickness is gone. He’s far more reliant on jumpers (now) than he was at the start of the season (where he regularly took the ball to cup).
There’s no doubt the ankle injury is affecting him.
Leon
January 25, 2010
Ryan Hollins is awful haha.
Nice to Gerald Wallace right up there (2nd in wins produced). Great fun to watch, and always very productive at the same time.
thecornerthree
January 25, 2010
Ariza just goes to show how effiency is related to usage/role, IMO
Italian Stallion
January 25, 2010
Lior,
Thanks for that clarification.
I may have to look at the calculations a little closer when the second book comes out.
It just seems to me that scoring might be underrated on this system relative to other contributions despite the positional adjustment. If so, that’s what accounts for guys like Kidd and Wallace etc… rating highly while some popular scorers don’t.
Here’s is “roughly” the way I understand it.
Please correct me if I am wrong!
PER basically says that you get 2 points for every field goal minus 1 point for every “miss”. So you if hit 50% of your shots you get credit for each point scored.
The problem with this formula is that if you have a FG% in the mid 30% range, you are still getting some credit for scoring despite the fact that your efficiency is terrible. IMHO, that’s a very legitimate complaint. IMO, PER tends to overrate inefficient scorers.
This system is roughly 2 points for every FG minus 1 point for every “shot attempt”. So you get no credit unless you hit more than 50% of your shots. That seems reasonable relative to the typical efficiency of 1 point per possession.
The potential problem is that if you are 1-2 that’s basically the same as 15-30 even though one guy scored 30 points and the other guy 2 points.
I know this gets back into the usage/efficiency and diminishing returns debate that has already been discussed to death, but something doesn’t sit right with me.
If you are going to require a player to have above average efficiency in his scoring (greater than 1 point per possession) before giving him credit for scoring, then perhaps players shouldn’t get any credit for rebounds, assists, blocks etc… unless they are above average also.
Yet a below average rebounder, assist man etc… still gets credit for each rebound, assist etc….
The way I see it, you should get some credit for every point you score at average or above average efficiency + extra credit for higher than average efficiency (and vice versa).
Scoring is a skill that not everyone has in equal quantities even if there is some flexibility in usage.
The positional adjustments may capture some of this, but I think what I am describing accounts for the larger number of low usage scorers that do very well here relative to other systems and the perceptions of most coaches, players, scouts, fans etc…
To be clear, that doesn’t mean everyone is right and this system is wrong. It could easily be the other way around.
But what you are saying is that the market for players is wildly inefficient. That’s such a bold statement it needs to be examined.
Few markets remain wildly inefficient for decades. That’s especially true when the people involved are experts that have been experimenting with this stuff in real life for decades trying to maximize a team’s output.
You could sort of suspect that even if they don’t understand everything, they would stumble onto many of the correct conclusions by trial an error after many decades.
I think inefficiencies can be found, but rarely at extreme levels for decades. Something doesn’t sit right with me, but I will examine some of the thinking in the next book.
Ken
January 25, 2010
IS,
Just remember though. Kidd rates high not just because of his rebounding, but also because he does all of the other little things. He gets steals, he still gets a ton of assists, his turnovers are down, he doesn’t commit many fouls…
Lior
January 25, 2010
Italian Stallion: Your problem is with the definition of “production”. Berri’s point is that in order for production to predict winning, it should be measured relative to a baseline. In the case of shooting, taking a shot has an opportunity cost, and PER does not take that into account.
Let’s make the simplifying assumption that the average NBA player shoots at 50%. Then a player shooting 15/30 in a game has not done anything particularly productive — had his team-mates taken the shots instead they would probably also have made 50% of their shots, and the team overall would have scored the same number of points. Also, any other player given the same number of shots would have done the same.
In other words: a team composed of unknown player X and 4 average players is guaranteed to be able to get baskets at a 50% accuracy by having the average players shoot, so player X shooting at 50% doesn’t bring anything new to the table — no matter how many shots that player takes. Any extra shot taken by player X takes away an equally good shot that would otherwise have been taken by a team-mate (that’s the opportunity cost), giving net benefit zero. Also, if we replaced player X with player Y, then player Y presumably would also shoot 50%, matching player X.
You are challenging this assumption — you are saying that a so called “low-usage” player shooting at 50% does not mean that the player will shoot 50% if tasked with shooting more. This might be the case, but the effect so small as to be negligible. Imagine we reduced the player from 30 to 18 shots and divided 12 shots among 4 other team-mates. Do you think that in 3 more attempts per game their percentage will suddenly drop off the cliff?
So yes, WP would over-estimate the number of wins of a team composed of Ben Wallace, Dennis Rodman, Bruce Bowen, Doug Christie and Jason Kidd — but no NBA team has ever fielded a such a line-up. Against the usual baseline merely shooting 50% is just doing your job, no matter how much you shoot.
Aside: Note that there is a fundamental difference between the way PER and WP have been developed. PER is based on a-priori assumptions on what helps teams win. WP is not constructed based on such assumptions — it is constructed based on finding which box-score entries actually determine wins. It doesn’t quite precisely say how the determine the wins. For example, if “scoring” ability existed, but was correlated with athleticism and hence with rebounding, then WP48 would credit this ability via its valuation of rebounds. Corollary to this is that proponents of the “usage” theory should show that incorporating it into WP improves the prediction of wins.
Footnote: The “average” shooting percentage in the NBA is not 50% exactly — but the difference is so small that using the correct percentage wouldn’t change much. It’s certainly within the overall error of the method.
Italian Stallion
January 25, 2010
LIOR,
I happen to agree with the assumption that usage is more flexible than most analysts think (especially fans). I don’t think it is as flexible as the assumptions in this model. I believe it matters at the margins and more or less in some circumstances that can be analyzed individually.
However, even if we assume I am wrong, you opened the door to why I think “efficient scoring” is undervalued relative to other contributions.
Forgive me for talking about the Knicks, but that’s the team I know best (poor me). :(
David Lee gets 11.4 rebounds per 36 minutes.
Let’s assume we instituted a “David Lee can’t rebound” rule.
Do you think the Knicks rebounds would drop by 11.4 in the time David Lee was on the court?
I think the chances of that are close to zero.
Gallo, Chandler, Harrington. Jeffries etc…. would all do a little better because they are all often in a position to get the same rebound. They defer to Lee because he is more aggressive and better at it. If Lee wasn’t allowed to get any boards, they would all automatically get some and also be instructed to get more aggressive about it.
Would they get back all 11.4?
I’d say there’s almost no chance of that, but they’d get back some.
So what is Lee really bringing to the table with his rebounds?
IMO he’s only bringing those rebounds that the other players wouldn’t have gotten, but he’s getting credit for every one.
Why should he get credit for every rebound even though some of them are fungible much like scoring?
I think if we are going to assume that scoring is almost totally fungible (which I don’t accept at extremes), then we should only credit rebounds, assists, blocks, steals etc…. when someone is bringing something above average to the table that could not be replaced by the other players on the court with him.
That would tend to upgrade the value of efficient scoring (over 50% FG%) relative to those other attributes and IMHO produce results that would be more in line with some of the extremely controversial cases and probably reflect the values better.
As it is now, IMO scoring (even efficient scoring) doesn’t count for enough for this reason.
On the usage issue, I’ll give you one quick example.
Chris Duhon is having a terrible shooting year. His inability to shoot is allowing defenses to give him space and take the Pick and Roll away from David Lee. That’s a problem because the P&R is one of the Knicks best offensive options.
The other options are Jared Jeffries (who can’t score if his life depended on it), Wilson Chandler (who is good at getting to the basket but can’t shoot from the outside for sh*t), and Gallo (who the defense is free to tie up because he’s the only decent shooting option left).
If the Knicks had plenty of offensive options, Duhon’s inability to either shoot or finish would probably not matter much, but they don’t. So IMHO, Duhon is a disaster.
You can’t have Duhon and Jeffries on the court together because they both suck offensively and that puts too much scoring pressure on the remaining 3 players, each of which is simply not that skilled and can’t increase his usage without impact. Lee picks up some of the slack with his newly developed outside game, but the rest is a disaster. They wind up with knuckleheads like Harrington on the court because he can score.
IMO, if the Knicks had a solid PG that could both shoot and drive/finish the team would improve a lot more than any formula would be able to measure.
Dre
January 26, 2010
Interesting note, Yahoo video has a talk with Kenny Smith about the most underrated players. Lamar Odom is first listed. As he has produced the most wins for the Lakers this season but is often seen behind Kobe, Pau, and Bynum this seems fair. Oddly Kenny Smith’s FIRST comment is that he has never averaged 20 points a season (his only real remark to dismiss Lamar). He then states Shannon Brown is the real underrated Laker.
mrparker
January 26, 2010
Korver’s only played 150 minutes so that should explain his gaudy wp48.
I’m surprised at Gortat, what gives? He’s down in every category except blocks
Palamida
January 26, 2010
Mr. Parker, off the top of my head: How would you feel if instead of starting for a playoff team (Dallas), you weren’t getting any younger and now binded by a contract for years to come playing 12.8 mins a game behind Ironman Dwight Howard?
Matching that contract offer sure worked magic for the Magic…
Oren
January 26, 2010
“I think if we are going to assume that scoring is almost totally fungible (which I don’t accept at extremes), then we should only credit rebounds, assists, blocks, steals etc…. when someone is bringing something above average to the table that could not be replaced by the other players on the court with him. ”
Admittingly, this doesn’t happen in Win Score. But for Wins Produced, doesn’t the position adjustment in fact ensure that a player only is credited for his rebounds above the average?
Or in other terms, if the average center gets 7 rebounds per 48 minutes and a specific player only gets 5, then wouldn’t his WP48 be below .1(presuming all other stats were average)?
brgulker
January 26, 2010
Gerald Wallace and Marcus Camby should be MVP candidates.
Now there’s a story that no one else will be telling!
brgulker
January 26, 2010
While I’m here, a complaint about the Pistons.
This seems to be the core that Dumars has built:
Rodney Stuckey .038 WP48
Charlie V .053 WP48
Ben Gordon .053 WP48
UGH!
I also noticed Maxiel is horrible right now — is it TO’s that are killing him?
And a big LOL at DaJuan Summers — he’s managed to produce -.8 wins in under 200 minutes! That’s right down there with some of the worst in the NBA.
Lior
January 26, 2010
Oren & IS: I was going to write a long reply but don’t have the time. The position adjustment does ensure that we only consider what the player does relative to what the team could have gotten from an average player instead.
Thus the average PF gets 11.4 rebs/48 minutes. David Lee gets one-third more (he gets 11.4 in 36 minutes).
However, what IS is really objecting to regarding rebounds is the assignment of the “value” of the rebound to the player who caught the ball. I have raised this point in the past; I consider it the part of WP requiring the most further justification. Prof. Berri has published some further analysis as a [technical] book chapter, but I my university library hasn’t bought the book yet so I haven’t read it.
The problem is the following: on the team level, one DRB equals one stop, and hence is worth about 1 point [the a-priori value of the possession to the other team]. On the player level, however, it’s not clear that the rebounder is the player who produced the defensive stop. Perhaps playing on a team with good defenders will make a player look as an above-average rebounder but in fact anyone with average skill at rebounding would have gotten the same numbers?
A more general version of this is: how do we know that the box-score row actually represents production by the player whose name appears on that row? Perhaps the FGM represent play-drawing by the coach and picks and assists by team-mates, DRB represent defensive stops by his team-mates, ASST represent individual scoring ability by team-mates who would have scored equally well with worse passes, etc etc? Then trying to assign the player a value based on his box-score row would be silly, since what we will be calculating will be a function of the whole team. This problem dominates statistical analysis of American Football — basically everything the team does depends on what all the players do, so it’s very hard to measure the contribution of any individual player.
dberri
January 26, 2010
Lior,
We do talk about the issue of inter-dependency in the next book. And that will be available in March.
Lior
January 26, 2010
Prof. Berri,
I was actually referring to your chapter in “The Business of Sport”, where I thought the details are.
Probably the next book will be in the style of the original one, a popular account of your work rather than a scientific report?
dberri
January 26, 2010
Lior,
In the main text the style is even more accessible than the first book. But we did include an appendix with more details on Wins Produced. And we do discuss the interdependency issue in this book.
brgulker
January 27, 2010
Isn’t this true for every part of basketball?
If Amare Stoudamire catches an alley-oop pass from Steve Nash, he will have done little more than position himself in the right place, let the ball fall into his hands, and move it 2-3 feet.
He will be credited with a made FG and two points.
This is sort of tongue in cheek, which I hope is obvious.
Getting a rebound is more than just letting the ball fall into one’s hands. Perhaps it happens that way once in a while. But those once in a while happenings are mitigated by the large sample size of Dr. Berri’s work (correct me if I’m wrong).
For example, a guy like Ben Wallace might have a rebound fall in his lap once in a while, but even on nights where that doesn’t happen, he’s still leading the team (and often both teams) in rebounds.
Ethan Harning
January 27, 2010
wow. Why won’t you admit you over-rate rebounding?
.448 Camby
.431 Love
.315 Troy Murphy
.279 Noah
.274 Ben Wallace
.214 Nazr Mohammad
versus
.160 Carmelo
.150 Nowitzki
wow! You really think Camby is 3x the win rate of Dirk?
really???
Troy Murphy 2x as good as Carmelo??
really?
Here’s the funny thing:
all those big rated rebounders have negative plus/minus. Melo and Dirk are two of the top plus/minus
these ratings are ridiculous
a defensive rebound for a player is not a defensive stop by that player
only shooting when you have a wide open dunk is efficient but not a sign of greatness
ilikeflowers
January 27, 2010
Ethan, of the hundreds of posts dealing with WOW and rebounding valuation yours has to be the most detailed and persuasive. You have annihilated this model by demonstrating its inherent paradoxes and have uncovered what looks to be a clearly superior statistical method. This ‘plus/minus’ that you refer to sounds particularly interesting. I for one anxiously await your future contributions and expect that they will similarly resolve other longstanding issues that have seldom been addressed in so comprehensive a manner. Indeed, I hope that you will favor us with your thoughts on statistical methods and modeling in general whenever you find it convenient.
Perhaps you can assist me with another related issue? There is a general, and likely statistical, model that says the answer is 42 but I think mesons, whales, and shorts are overvalued relative to corncobs. What is your opinion on this matter?
Tom Mandel
January 27, 2010
“Good” is an abstract term. WoW does not determine how “good a player” anyone is. What it does is measure a player’s contribution to winning basketball games — his contribution *at his position* I should rather say.
Basketball, obviously, is both competition to win games and entertainment to display athleticism and specific skills. I’d rather watch a one-on-one game between Nowitzki and Melo than between Troy Murphy and Marcus Camby — and I’m sure Dave would too! It would be a lot more entertaining.
For that matter, I imagine Melo could take Camby in a one-on-one schoolyard game. So what?
Jim Glass
January 27, 2010
“Clearly this is NOT the year of the Super Teams”
On the up side. What about the down side?
How about them Nets? Worst team in NBA history?
So far 3-40, on course to break the record, and one of their best players, Chris Douglas-Roberts (0.4 WP in 1112 minutes) just told the local papers he thinks “it could get worse”.
Really bad teams can be just as interesting as really good ones.
Will
January 27, 2010
@ Lior
I think another good question is: how do we know we can credit someone else for the DRB (ie the ‘stop’)? Wide open 3s are routinely missed. So are wide open 2s. Even wide open layups are sometimes botched. So how do we know that when someone misses a contested shot that is was really the contest that caused the miss?
I’m not discounting defense, but can you really be so sure that you can credit one individual?
We do know that some players are consistently better rebounders than others. It’s plain as day when you look at per minute stats.
Can you say the same for good defenders? Yes I know some players (Battier, Artest) are considered great defenders by the media and fans. But can this be quantified empirically?
Italian Stallion
January 27, 2010
I know I get way too repetitive for most people’s taste and appreciate that D Berri has been very patient with me to date. I am to a large extent just trying to understand the game better myself and am looking forward to the next book.
Anon
January 27, 2010
I would be curious to see how the ratings change if every defensive rebound was credited equally to every person on the floor at the time.
kinda goes against the whole box score only thing though.
Italian Stallion
January 27, 2010
Lior,
“The problem is the following: on the team level, one DRB equals one stop, and hence is worth about 1 point [the a-priori value of the possession to the other team]. On the player level, however, it’s not clear that the rebounder is the player who produced the defensive stop.”
I agree with what you saying, but my issue has more to do with the fungibility of the rebound (maybe that’s the wrong term) than who accounted for the stop that lead to the rebound.
I think a rebound is worth something.
I think a point scored by a player of even average efficiency is also worth something.
However, I think if you are going to argue that usage is pretty irrelevant (other players easily make up the difference for one or two low usage guys on the court with them), then I think you have to concede the same is true for rebounds and probably to a lesser extent assists, and an even lesser extent blocks and steals etc…
I would not argue that rebounds are vervalued.
What I am arguing (actually speculating) is that everything is overvalued relative to”efficient scoring”. (key word efficient because IMO PER is way off in that regard)
When you do that, it pushes the rebounders up the list and the scorers down because rebounds tend to count for more than assists, blocks, steals because there are so many more of them.
That’s why you get the list that Ethan objects to.
Raising the value of efficient scoring or lowering the value of everything else because it’s also fungible would produce a list that would be more in line with strongly and almost universally held perceptions and still identify hidden gems, overvalued players etc…
So instead of calling Dirk slightly above average this year, we’d recognize him as a great player that is essential to Dallas.
At the same time, we’d recognize that Camby is a lot better than most people think, but not one of the greatest or most valuable players ever.
I hope that’s clearer.
And again, I am looking forward to the book in March.
simon
January 27, 2010
Ethan Harning//
According to adjusted +/-, Carmelo is in the negative range this season: http://basketballvalue.com/teamplayers.php?year=2009-2010&team=DEN
According to adjusted +/-, Nene, Martin and Chris Anderson are all far far far more valuable than Carmelo. You still like +/-?
Italian Stallion
January 28, 2010
Perhaps what everyone needs to recognize is that every player contributes either some positive or negative value on every single offensive and defensive possession.
Picture these plays:
Duhon dribbles the ball up against a press and gets it over mid court, David Lee sets a pick for Gallo who gets loose on the elbow, Duhon gets the ball to him, he knocks down a 3.
On that play, Gallo gets credit for the 3 pointer, Duhon gets credit for the assist, but Duhon gets no credit for beating the press and Lee gets no credit for setting the pick that freed Gallo. IMO both should get some marginal credit for those other contributions even if they are smaller than the score itself.
In addition, I even think the other two players on the court deserve some very marginal credit for being enough an offensive threat to prevent their defender from helping out on Gallo.
Picture the same exact scenario except that when Gallo goes around the pick, Jeffries’ defender leaves him open on the perimeter (because he hasn’t hit an outside shot since the stone age lol). He helps out on Gallo and prevents the success.
We don’t measure the negative impact of a poor shooting low usage offensive player allowing his defender to quickly help and prevent a success on some occasions during the game.
Finally, imagine the same scenario except it’s Michael Jordan going around the pick and getting doubled by the help. However, in this situation Jordan simply beats the double team gets to the hoop and posterizes the paint player.
In that case, Duhon deserves some but less credit for the pass, Lee deserves some but less credit for the pick, Jeffries deserves some “negative credit” for making the success more difficult, and Jordan deserves some scoring credit + a credit for having such overwhelming skills he could offset the Jeffries negative.
That’s what people see in those great scorers that the stats don’t measure. They can turn a failure into a success.
Another example is that some small value should also be assigned to any player that draws a double team that eventually leads to an easier basket than would have otherwise been the case over and above the potential assist.
I could go on and on.
I also think these values are variable:
Bringing up the ball is worth almost nothing if the defense drops back. Beating a full court press has some marginal value.
Making an incredible pass that leads to a layup may have more value than the basket itself, but making a routine pass to a wide open player may have less value than the typical assist.
I think once you recognize that the values are not limited to the things we measure and that the values of each thing we measure varies depending on the circumstances, you start to understand why some models break down in some places and why the perceptions of experts sometimes disagree very strongly with models.
It’s because these experts are seeing the things that have positive or negative value that we are not measuring at all or that we are assigning fixed values to when they are really variable.
I still think that stats should be used as a tool within the analysis because perceptions are also often flawed (sometimes wildly), but very good perception + a very good overall understanding of the game is required to iron out the edges on those stats until such time that we measure and value everything properly.
Xavier Q
January 28, 2010
IS, I’m fairly confident everyone does realize that more than a single person contributes to each FG, rebound, steal, block, etc. the problem is in measuring it. In a single 48 minutes game you could go through and watch every single play, rewind, and determine who helped the play and who didn’t. But could you determine that a pick is worth 20% of the FG that resulted from it? Or 40%? Or that someone “hurt” the play, and by how much? The analysis becomes incredibly subjective after a certain point, leaning more toward opinion than analysis.
Assuming it were done objectively, there are 1230 games in an NBA regular season. Watching all of them simply isn’t practical for one person to do at that level of scrutiny. Whether you believe in WP or not, it does have the benefit of correlating to the wins a team gets in a year, even though it focuses on individual players and their stats. That lends a certain level of creditability to the model.
Ethan H, +/- is not an accurate measure of, well, anything. Yes it’s possible that in WP that rebounds might be overvalued and scoring undervalued, the general concensus is that the public, and team officials who pay players, VASTLY overvalue scoring.
The only example really necessary is the moves the Pistons made in the offseason (which Berri discusses here :https://dberri.wordpress.com/2010/01/12/a-history-lesson-from-detroit-that-is-repeated/). If scoring was as important as you seem to think, then adding Ben Gordon and Charlie Villaneuva should have catapulted the Pistons to the Finals. It doesn’t look like that’s the case.
Even +/- proponents do not suggest using it to compare players. It is for measuring trends.
Tom
January 28, 2010
Jason Kidd twice as valuable as Dirk Nowitzki?
http://www.82games.com/0910/0910DAL.HTM
mrparker
January 28, 2010
I hate when I am this late to the party.
re, Efficient scorer vs. great rebounder
According to the four factors, fg shooting = 40% of the game, turnovers 25, reb 20, and fta 15. You can cut those numbers in half to account for offense and defense.
Efficient scorer only excels in 20% of the game. Awesome rebounder accounts for 20% of the game as well because efficient scorer is only allowed to shoot into his own basket while awesome rebounder can rebound on both ends of the floor.
dberri
January 28, 2010
Tom,
Adjusted plus-minus in 2008-09:
Jason Kidd: 15.23
Dirk Nowitzki: -1.51
And those numbers are supposedly control for every other player and the opponent. Of course, the numbers completely change this year (suggesting the controls are not exactly perfect). But if you like plus-minus, that is one piece of evidence Kidd is worth more (and of course, with plus-minus you can generally find evidence of the opposite argument as well).
khandor
January 28, 2010
IS,
re: It’s because these experts are seeing the things that have positive or negative value that we are not measuring at all or that we are assigning fixed values to when they are really variable.
I still think that stats should be used as a tool within the analysis because perceptions are also often flawed (sometimes wildly), but very good perception + a very good overall understanding of the game is required to iron out the edges on those stats until such time that we measure and value everything properly.
Amen. :-)
Tom
January 28, 2010
dberri, I’m getting different numbers than you for last season.
Nowitzki: +8.0
Kidd: +12.4
http://www.82games.com/0809/0809DAL.HTM
Sam Cohen
January 28, 2010
Tom- I’m pretty sure the numbers from 82games that you’re looking at are regular plus/minus numbers, while the numbers Professor Berri posted are the adjusted plus/minus numbers.
Oren
January 28, 2010
Tom,
I don’t know if 82 Games offers adjusted plus/minus or just plus/minus. This site is what Berri used I suspect.
http://basketballvalue.com/teamplayers.php?year=2008-2009&team=DAL
dberri
January 28, 2010
Yes, I was looking at the basketballvalue numbers. Like the raw plus-minus numbers, the adjusted plus-minus numbers jump around quite a bit.
Italian Stallion
January 28, 2010
Xavier Q,
I agree with everything you said.
The problem is:
Sometimes a player measures well on one model but poorly on another.
Sometimes a player is highly regarded by stats guys but dismissed by coaches, players, and GMs as just OK.
Sometimes a player is highly regarded by players, coaches, and GMs but dismissed by stats guys as just OK.
Without those available obscure stats and the associated values, it is impossible for one side to prove to the other side who is right at a level that would satisfy academic circles.
But common sense and intuition should not be dismissed just because there is a lack of academic proof when we have decades of practical experience and observation to go by.
I guess what I saying is that right now the stats guys are trying to catch up to and explain what the very best players, coaches, and GMs already know as a result of their experience even though they may not understand it all.
At the same time, the stats guys are providing nuggets of new info and explanations that allow the very best players, coaches, and GMs to improve on their own efforts.
So when either side points out a “value issue” in a model or the thinking that almost certainly matters based on even a marginal understanding, I think the goal should be to improve the thinking and/or model and not dismiss it or get defensive about it.
simon
January 28, 2010
IS//
“But common sense and intuition should not be dismissed just because there is a lack of academic proof when we have decades of practical experience and observation to go by.”
No, but the problem with “common” sense is that it’s usually just a passed-down old wisdom that perpetuates itself without much evidence. Science is full of common senses which have been proven false over the years, ditto for “observation.” Basketball or any social science is different in that sense the variables are infinitely harder to control for, but at the end you do have to show some sort of evidence to claim something.
Westy
January 28, 2010
Lior,
Good summation above of the potential problem with assigning total value (to the team) for a defensive rebound to the individual who procures it.
khandor
January 28, 2010
David, IS, simon, brgulker, and all others who would like to participate,
If you were forced to choose only 1, which specific statistical category, currently associated with the game of basketball, would you identify as being the most vital, in terms of having “an actual effect[s]” on the winning and losing of important games played in the NBA against high calibre opponents?
i.e. You are not allowed to choose more than 1.
If you’d like, please feel free to provide an explanation of your rationale.
PS. Those who’ve read my work on-line should already know my answer to this simple question.
mrparker
January 29, 2010
You guys are wrong about how a total rebound is assigned. You don’t understand the team adjustment.
todd2
January 29, 2010
It would appear rebounds are over-rated using this model and IS makes some good points. There is some value in a player that can distribute and take care of the ball and/or be a scoring threat and draw double teams. I think we’re seeing the Rockets bog down some lately because they’re a team of role players. The 24 second clock limits the effect of x’s and o’s and at some point teams need a guy that can make plays—whatever that means.
Tom Mandel
March 2, 2010
I’m worried about the integrity of your raw numbers here, Dave.
E.g., after 41 games, you have NO’s Darren Collison w/ 571 minutes. Today, after 55 games, the nba player file for him shows 1301 minutes.
Has he really played 630 minutes in the last 14 games? 44+ minutes per game?
dberri
March 2, 2010
Tom,
You need to go back and look at Collison’s numbers. After 41 games Collison had only played 35 games. He has now played 20 games since the midpoint.
Tom Mandel
March 2, 2010
Duh, you are right. And he is playing more — almost 140 minutes in his last 3 games.