Jeremy Britton, an SF Bay Area interaction designer at ZURB, fell in love with NBA basketball as a kid while watching Eric “Sleepy” Floyd drop 28 fourth quarter points on the Lakers in front of a bank of TVs at a Bay Area Price Club with dozens of Warrior faithful. The Warriors took that thrilling game, but dropped the series 4-1. From Don Nelson (part 1), to Run TMC, Chris Webber, Latrell Sprewell attacking PJ Carlesimo with a 2×4, and their amazing “We Believe” run three years ago, Jeremy has inexplicably remained a part of the Warrior faithful but rarely been rewarded.
Hapless Warrior fans have focused their “we believe” energy almost exclusively on the chance reclusive, litigious owner Chris Cohan might sell the team to a more motivated owner this August. Yet even if the curse of Cohan is lifted, it’s not clear potential new owners Larry Ellison or the 24-hour Fitness guy can put a management team in place capable of assembling a winning team. If these owners aren’t shrewd with their money or their judgment of talent, they might dip too far into their own deep pocket and saddle Warrior fans with a version of the Knicks West.
But Wait…
A funny thing happened the day after the “The Decision” and amid all this new Warrior ownership talk. The current Warrior management led by GM Larry Riley traded Anthony Randolph, Ronny Turiaf, and Kelenna Azubuike to the New York Knicks for David Lee. Negative reactions from Warrior fans focused on how Turiaf and Azubuike had “heart” (they were fan favorites), on Randolph’s wild card “potential,” and on Lee’s perceived lack of defense and new expensive contract. But how does this move change the Warriors’ productivity at power forward?
The move to get David Lee suggests a big jump in productivity at power forward–about a 14.4 win improvement. A quick back-of-the-napkin calculation suggests a 39 win team next year. Not great, but better. The Warriors didn’t have the horses at this position last year, but by replacing classic Nellie “small ball” with Lee — a great rebounder and efficient shooter — their fortunes should improve.
Unexpected Reason for Even More Optimism
What about the Warriors’ other moves? They drafted Ekpe Udoh with the sixth pick, traded Corey Maggette in a ‘contract dump’ for Dan Gadzuric and Charlie Bell, signed free agent Dorell Wright, and are letting fan favorite free agents C.J. Watson and Anthony Morrow sign with other teams. Let’s just say those yellow “We Believe” shirts will stay neatly tucked into dresser drawers.
The Warriors also hold out hope for Andris Biedrins and Brandan Wright to return from injury (they missed 131 games together last season) and for improvement from returning rookies Stephen Curry and Reggie Williams. [Note: I haven’t factored likely second year improvement in yet.] This is the same kind of ‘hope’ Warrior fans strain their eyes to see every summer, so ho-hum, right? But wait–what impact might all these moves have on the Warrior’s ability to win next season? Potentially, a lot.
Wins by position:
As the chart above shows, the Warriors might expect to win about 25.8 more games in 2010-11, making them a 50 win team. They see almost all of this improvement on their front line, first with Lee at power forward, and then at center and small forward.
At center we factor in the (somewhat) healthy return of Andris Biedrins. Specifically, let’s imagine him at his 2008-09 form (a season he missed 22 games and was a shade below his peak productivity). If Biedrins returns to that kind of form, the Warriors will gain better than 7 wins at center.
At small forward the Warriors replace a collection of lesser players with a full season of Reggie Williams and newly acquired free agent Dorell Wright. Williams surprised fans last year, but as Wages of Wins pointed out, based on his college performance his good NBA performance shouldn’t be a surprise at all. Wright, meanwhile, has been an unheralded but positive contributor almost every year he spent in Miami.
Who will get the credit if 50 wins pans out?
If this estimate is actually how it goes next season, the big story in Oakland will be about how the Warriors did it. Will the most productive players take the credit? More likely new ownership will somehow take a large share of the credit off the court (curse lifted?), while the Warriors’ leading scorers will take it on the court.
Stephen Curry, David Lee, and Reggie Williams will receive the lion’s share of the credit due to their scoring (and actually deserve much of it), while the lower scoring contributions of Dorell Wright, Brandan Wright, and Andris Biedrins will take an undeserved backseat for the team’s turn of good fortune.
Either way, most of the credit should go to a general manager no Warrior fan believes in, who doesn’t even know what kind of winning team he has assembled, and who may not even be employed by the Warriors in time to enjoy their success–Larry Riley.
EPILOGUE / ALTERNATE ENDING
How the Warriors Could Still Screw It All Up
The moves being made by current general manager Larry Riley are a mixed bag. While the Lee and Wright acquisitions bode well, the Maggette trade and Udoh pick suggest Riley stumbling through these decisions without a clear measure of player evaluation.
Rumors that the Warriors will trade Andris Biedrins in another ‘contract dump’ could threaten to drop the Warriors right back into the middle of the pack. Coach Don Nelson, both notorious and successful for his unconventional lineups, could shoot his team in the foot with “small ball” by sitting Biedrins and Wright too much, or by relying on Lee too much at center.
However, other rumors of a Monta Ellis trade could improve their weakest position. At shooting guard Ellis produced almost zero wins in a lot of minutes. Let’s just hope the bad moves continue to be more than offset by the good ones as the Warriors stumble into next season.
– Jeremy Britton
Wages of Wins Journal 
robbieomalley
July 14, 2010
This is probably the prettiest post I’ve seen so far. It looks nice. Also, nice analysis.
I would like to point out a few things. First is that, according to ESPN, the Warriors had the lowest team rebound rate of any team in about the last 8 years – by a wide margin.
I also think Don Nelson’s wanting to go small is a bit overstated. I think he went so small so much last year because he really had no choice. The Warriors were as decimated by injuries as any team in the NBA. Also when Biedrins, Randolph, and Wright were healthy they generally were in the starting lineup. This is most true in Biedrins case.
Now that he has Lee he really has no choice but to play him 2800+ minutes if he can.
shawnfuryan
July 14, 2010
Great post. The charts look awesome. I’d been sleeping on the Warriors, well, no more!
arturogalletti
July 14, 2010
That’s awesome. Loved it. I’m going to have to up my graph game.
But seriously, I think the GSW have definite potential, espacially if they can move YayPoints! superstar Ellis for something.
ilikeflowers
July 14, 2010
Me like pretty circles.
Eliot
July 14, 2010
They have 2 big expiring contracts in Rad-man and Gadzuric so they do have the ability to add more talent before or during the season, but they could just as well add an overrated player or not use the contracts at all. Also Monta was pretty solid in the 07-08 season posting a .178 WP48 due much more efficient scoring numbers and far fewer turnovers, in addition to superior rebounding numbers. Perhaps with Curry handling the ball more and David Lee getting touches inside he gets better looks and thus has a year closer to that season than the last.
MegaloArenas
July 14, 2010
A 100 bucks say Ekpe Udoh is gonna get all the credit if they improve by this much
Tommy_Grand
July 14, 2010
Good write-up.
If GS trades Ellis for a value (or if he plays more efficiently) and the injured players return to near peak form, this team could be in the playoffs (and dangerous) for the next 2-3 seasons.
Mike
July 14, 2010
Jeremy,
Great write up! And we didn’t even factor in the fact that Curry had a 0.22 WP48 after the all-star break. If he can continue that trend and be a .2 WP48 player for the entire season, the Warriors can be a pretty formidable team.
Eliot,
Monta’s past 4 out o f 5 seasons have been well below average (< 0.05 WP48), with the exception of the 07-08 season. I'd say it is too optimistic to think that he can produce at that level.
Joe
July 14, 2010
You all seem to be missing who will get the credit actually…
Monta Ellis. Next season is going to be his breakout season when he brings all the intangibles like his leadership and killer instinct. Next season he finally will learn how to make his teammates better.
some dude
July 14, 2010
Isn’t Udoh already out 4-6 months?
Isn’t Tolliver expected to be signed by another team and not matched by the GSW?
Projecting reggie williams at that # of minutes? Don nelson is still the coach. Radmanovic will still get plenty of time.
Plus, Lee is now in a conference with effective big man, so a lot more games against tougher direct competition.
50 wins. If the Warriors win 50 games next year (discounting some massive unforeseen trade), I will start preparing for the 4 horseman of the apocalypse.
Jeremy Britton
July 14, 2010
@robbieomalley You’re absolutely right, injuries really forced Nellie’s hand last year. He didn’t have the horses on the front line and had to go small and play D-Leaguers like Tolliver and Hunter.
That said, even when Nellie has had bigs, we tends to favor smaller, unconventional lineups.
@Eliot That’s my hope for Ellis too, but even with Curry handling the ball at point last year, Monta has gotten awfully cavalier with the ball (turnovers) and with his shot selection. I fear bad habits have set in and he’s addicted to calling his own shot.
@Mike Yeah, you’re absolutely right. Both Curry and Reggie Williams should see a bump in their second years. I left that out on purpose because I wanted to temper my optimism and didn’t want to overcomplicate the analysis.
Jeremy Britton
July 14, 2010
@robbieomalley You’re right, Don Nelson had to go small and play D-Leaguers on the front line. Historically he’s seemed to favor smaller lineups that play up tempo, force turnovers, and create fast break points at the expense of rebounders. As you say, though, the horses he has next year might force him to play a rebounding team.
@Eliot I have the same hope about Ellis returning to his 07-08 form, but worry bad habits have set in. He’s had a lot of positive feedback related to his scoring.
@Mike Great point about Curry and I’d add Reggie Williams, the other returning Warrior rookie, to that equation. Being disappointed so often by this team, I wanted to temper my optimism and left that out of the analysis.
@Joe You might be right if Ellis is again their leading scorer, but I think Curry and Lee have enough notoriety and scoring power that they make get the lion’s share of the credit.
@some dude I wrote this before the Udoh injury and just fudged Tolliver in there (he’s effective a “zero” in terms of productivity). I hope you’re wrong about Radmanovic getting minutes and disagree about Lee’s productivity somehow going down in the West. Get the 4 horseman ready.
Mike
July 14, 2010
Some dude,
“Isn’t Udoh already out 4-6 months?”
Correct, he has wrist surgery, but in the analysis Udoh plays minimal minutes (accounting for his injury) and he’s using a less than stellar rookie WP48 of 0.01 (which is bad, even for a rookie)
“Isn’t Tolliver expected to be signed by another team and not matched by the GSW?”
You and your rumors, such as the Biedrins trade with no evidence. Also, I’d rather the Warriors not sign Tolliver anyways, since he is well below average.
“Projecting reggie williams at that # of minutes? Don nelson is still the coach. Radmanovic will still get plenty of time.”
I suspect that Dorell Wright will get the lions share of minutes at the SF spot. Reggie Williams can back up the 3 and the 2, both positions which he is quite above average at. Radmanovic is an expiring contract, and barring injury, he won’t play very much.
“Plus, Lee is now in a conference with effective big man, so a lot more games against tougher direct competition.”
The conference argument is weak, since every team plays every other team at least twice. So if you are talking about division (4 games against same division teams), Lee will only face stiff competition against the Lakers. Also, some of the West PFs defected to the East: Utah lost Boozer, Suns lost Stoudamire, Denver’s Martin’s out for the season.
“50 wins. If the Warriors win 50 games next year (discounting some massive unforeseen trade), I will start preparing for the 4 horseman of the apocalypse.”
This is just predictions. Much like how you predict the Warriors won’t win 40 games. Except there is statistical evidence to back up this prediction.
Joe
July 14, 2010
Yeah, except your love affair with David Lee ignores the fact that he’s one of the worst defenders in the game, which your statistic doesn’t even come close to accounting for.
Oh, I forgot: these are the people who think Lee’s productivity is comparable to, if not better than, Dwyane Wade’s. Reason and common sense aren’t met kindly here.
Ryan
July 14, 2010
50 seems a bit high. They should trade monte for j-rich, age and knees be damned, and let it all come full circle.
Mike
July 14, 2010
Ryan,
If no one else gets injured (besides Udoh) and miss significant amounts of time, the numbers project to more than 55 wins. This means 50 wins is definitely possible, even if they employ some negative WP players like Chris Hunter or VladRad.
And I’m sure all GSW fans would love a J-Rich for Monta trade, although that seems very unlikely, unless the Suns hire a terrible GM.
evanz
July 14, 2010
I’m not quite sure I understand the knock on Riley for the Maggette deal. It seems to me that the expiring contract of Gadzuric is well worth it, since Maggette was clearly not the long-term solution at SF. Also, it seems at odds to praise Riley for getting Dorell Wright, since he is going to replace Maggs at that position. Overall, it’s like we traded Maggs for Wright+$6M expiring. Next off-season, the Warriors will have about $14M in expiring (Gadz+Radman) to go out and get someone decent. Whoever the new owner is, he’s coming in at the right time.
some dude
July 14, 2010
“You and your rumors, such as the Biedrins trade with no evidence. Also, I’d rather the Warriors not sign Tolliver anyways, since he is well below average.”
uh, the Denver post reported a Biedrens to Denver discussion. google is your friend.
“The conference argument is weak, since every team plays every other team at least twice. So if you are talking about division (4 games against same division teams), Lee will only face stiff competition against the Lakers. Also, some of the West PFs defected to the East: Utah lost Boozer, Suns lost Stoudamire, Denver’s Martin’s out for the season”
Amar’e is not a defensive player, so removing him is irrelevant for Lee’s WP48. Two more games vs Pau/Bynum, Griffin/Kaman, and added games vs the Spurs, Blazers, Thunder, Rockets, Mavs, and I’m high on Cousins/Dalembert (but rookies and defense are usually not synonymous).
Lee’s WP48 will drop. mark it, dude.
“This is just predictions. Much like how you predict the Warriors won’t win 40 games. Except there is statistical evidence to back up this prediction.”
Weren’t the Warriors predicted to win more than at least one of the last 2 years by the same measurements?
These “predictions” are based on assumptions on what Don Nelson will do. Personally, I think those assumptions are flat out incorrect based on Nelli’s historical decision-making. Like you claim Radman not playing much at all. I’ll believe it when I see it.
some dude
July 14, 2010
“You and your rumors, such as the Biedrins trade with no evidence. Also, I’d rather the Warriors not sign Tolliver anyways, since he is well below average.”
Except the Denver post reported it for Biedrens?
Tolliver is talking to the Wizards, Mavericks, and Jazz right now. This is known stuff. GSW is also not a luxury paying team as far as I know, though the sale last night might change things. But if they’re not, then they’re not signing anyone past the minimum.
“The conference argument is weak, since every team plays every other team at least twice. So if you are talking about division (4 games against same division teams), Lee will only face stiff competition against the Lakers. Also, some of the West PFs defected to the East: Utah lost Boozer, Suns lost Stoudamire, Denver’s Martin’s out for the season.”
1. Amar’e isn’t doing anything defensively anyway, so who cares.
2. I’ll give you boozer, but he goes from Dwight to Pau/Drew, Griffin/Kaman, and Cousins/Dalembert 2 times a year more. Then more games vs Spurs, Thunder, Blazers, Rockets, Mavs, and others with better defensive front lines than the East. He’ll also have to work harder on the offensive end.
3. Kenyon Martin is not out for the year. Don’t know where you heard that one. As of right now the only thing is a “might” miss training camp.
“This is just predictions. Much like how you predict the Warriors won’t win 40 games. Except there is statistical evidence to back up this prediction.”
This is the fallacy. The prediction is based on assumptions on how Nellie will dole out the minutes and the positions. Predicting what Nelli will do is akin to predicting radioactive decay of sub-atomic particles. IMO, the assumptions made are incorrect and therefore all the predictions are called into question.
some dude
July 14, 2010
“If no one else gets injured (besides Udoh) and miss significant amounts of time, the numbers project to more than 55 wins. This means 50 wins is definitely possible, even if they employ some negative WP players like Chris Hunter or VladRad.”
55 wins. I normally would have thought one would have to be intoxicated to think that of the Warriors. :P
These predictions are so absurd that I wonder if some of you think all of basketball is explained in a single metric. The Warriors are not sniffing 55 wins. sigh…
dberri
July 14, 2010
sd,
not sure I understand your position. Are you arguing…
1. there is no way to make a prediction because you can’t predict stuff like how Nelson will allocate minutes.
If that is your position, then how do you know 50 wins must be incorrect?
2. There is a way to make a prediction, and it says 50 wins is not possible. If that is your position, please detail the sd method. Provide us enough detail so we can see how this method could be tested.
And answer all this in one comment. I am thinking of a new rule inspired by your behavior. No more consecutive comments from one person.
some dude
July 14, 2010
1. Well, I was trying to make a joke about predicting Nellie.
personally, I think the predictions on minutes/positions given were incorrect. I think Radman will play a bunch and Lee will play significant minutes as a C and he could get to 40mpg.
Also I don’t know if Tolliver will be on the Warriors and think it’s unlikely:
Lee, Ellis, Biedrens, Gadzuric, Radman, Bell, B.Wright, D, Wright, Curry, Udoh, Williams.
That’s 11 men on contract and they’re near the LT threshold. Reports from the media are Tolliver will stay if Warriors match a deal offered, but won’t come back for less and Warriors won’t offer more than $1 mil which he rejects. (from bay area news media)
Now, if he leaves, what happens? Probably Gadzuric (negative win producer) plays more and Lee shifts more to C. With Lee playing more C, this opens up more time at PF for poor win producers, no?
Furthermore, Reggie Williams has only played 24 games in the NBA. Small sample set.
2. I never argued you couldn’t ever predict. I am calling into question the assumptions the prediction are based on.
I don’t use a specific statistical model to predict anything. I combine knowledge from advanced stats like WP48 and my own personal and intimate knowledge of the game (even though I know y’all don’t like that approach). Healthy medium, FTW.
I’m sure you’re blast me for that, but it’s all good, prof. You stay rigid, I’ll stay free-flowin’.
Italian Stallion
July 14, 2010
I think the # of wins being assigned to David Lee is probably too high. The Knicks were a poor defensive “team”, but Lee was probably the weakest link on that team.
Golden St fans will quickly learn that Lee’s 20-10 nights are often accompanied by 25-10 nights by his opponent and that he provides ZERO inside help when perimeter players get beat.
If Randolph stays healthy for a full season and improves his efficiency while reducing youthful mistakes over time (he turns 21 tomorrow), he’s going to be a better player than Lee within a couple of years.
He’s already a better individual defender. Plus players like him get very little credit for all the shots they alter because of their length and weak side help.
I think there’s actually a chance Randolph wins most improved player in the NBA this year IF he can stay healthy.
some dude
July 14, 2010
Stop with all that logic, IS. What’s wrong with you!
But yeah, I think you make a lot of sense. I’m curious to see what D’Antoni does with Randolph. He might try to use him like he did with Diaw as a creator within the offense (Diaw acting like a facilitator after Nash). Not a big fan of Mike D, but he does know to utilize the offensive end of the floor.
Italian Stallion
July 14, 2010
some dude,
I think D’Antoni has some ideas about playing defense that run counter to prevailing wisdom.
I can’t give you specifics, but I’ve heard him say things on his MSG show that suggest that in some situations he is willing to give up slightly more PPP over the long haul because it leads to a bigger advantage on the offensive end.
Suppose I told you that maximizing the defensive effort would allow the opposing team to score 1 PPP and for your team to 1.05 PPP.
I think you would be very happy.
Now supposed I told you that in some rare situations if you weakened the defense in order to prepare the offense in the event of a miss it would permit the opposing team to score 1.05 PPP, but your team to score 1.12 PPP.
I think you’d be even happier.
Even though your defense was worse, the things you were doing created a greater point differential.
I think some of his basketball philosophy is built along those lines. He’s not trying to maximize the defense. He’s trying to maximize the point differential.
When he coached the Suns and they would win 128 – 118 he would get criticized for their lack of defense.
His typical response was “We played better defense than they did”.
What he was really saying was that the Suns could have held their opponents to a lot fewer points if they wanted, but it would have come at a greater cost to the offense.
arturogalletti
July 14, 2010
IS & sd,
I’ve shared some of your concerns on bad D being overlooked in the past. One of the coolest things about the scientific method is going out an testing something. David Lee to GSW could provide a great test case for the model that’s being used (granted nellie & the warriors organization could be something of a blocking variable). It’s another great subplot to what should be a great season.
Oh and I put a new post on changing value by position for nba players (click my name if that sounds like your sort of scene)
Jeremy Britton
July 14, 2010
@robbieomalley You’re right, Don Nelson had to go small and play D-Leaguers on the front line. Historically he’s seemed to favor smaller lineups that play up tempo, force turnovers, and create fast break points at the expense of rebounders. As you say, though, the horses he has next year might force him to play a rebounding team.
@Eliot I have the same hope about Ellis returning to his 07-08 form, but worry bad habits have set in. He’s had a lot of positive feedback related to his scoring.
@Mike Great point about Curry and I’d add Reggie Williams, the other returning Warrior rookie, to that equation. Being disappointed so often by this team, I wanted to temper my optimism and left that out of the analysis.
@Joe You might be right if Ellis is again their leading scorer, but I think Curry and Lee have enough notoriety and scoring power that they make get the lion’s share of the credit.
bags fly free
July 14, 2010
55 wins for the warriors? that’s 2 less than lakers won last year….you know sometimes i wonder if ppl actually watch basketball here or just look at the box-score after the game and jerk each other off here with their WoW numbers
some dude
July 14, 2010
IS, I actually believe that is what’s going on to some degree. However, I have 3 challenges to that.
A. This one isn’t so much against his theory, but Phx would often make dumb defensive mistakes. Guys like Amar’e would be way out of position. I understand the theory on relaxing on D, but there’s relaxing and just being dumb. I put a lot of the blame on Amar’e poor D on D’Antoni’s coaching him in his younger days. He, and those Suns teams, made too many simple and fundamental mistakes that should never happen during a game Has he focused more on teaching defense to the younger guys like Amar’e and Barbosa, I think they would have been better off long term.
B. Good defense is good defense. If you give the effort, you’ll almost always accomplish this goal. Offense is more variable. Even if you give top effort, the shot may not fall. The refs may not blow the whistle in your favor (or even against you) for no specific reason. You can overcome poor shooting, but you rarely can overcome bad defense in the end game.
C. I believe you will always run into a team that can exploit your defense if you play poor defense. This is essentially what the Spurs always did to the Suns. You basically killed the Suns by scoring in the paint efficiently. Do that coupled with solid D and you’d beat them in a series. Spurs could score with Duncan and parker/Ginobili penetration all series long, which eliminated the “relaxed” strategy on D.
It’s also how the Lakers nearly upset them a few years ago. Kobe didn’t try to force, outside of game 5, much and let Walton, Kwame (shudder), Mihm, and Odom post up time and again. It actually worked quite effectively and they were a bad Kwame rebound + Tim Thomas 3 away from the upset.
The Clippers essentially recreated this with Brand/Kaman in the post and maggette’s getting into the paint. They too were close to the upset.
I honestly believe that at some point they will hit a team in the playoffs that can post up and drive into the paint and it will always be their undoing. Happened this year in the first 2 vs Lakers til they switched to a zone.
On a side note, I think the numbers are overstated in the regular season. Let me give my explanation.
NBA players don’t play their hardest every game of the season. It’s a long 82 season grind, with b2b and some 4 games in 5 nights in stretches and all on the road. It’s pretty tough, especially for rookies and older players.
D’Antoni’s style of play is very aggressive in that they constantly pushing the ball moving it around, etc. It takes more effort to play D on them than most other teams. What I believe, and cannot prove, is that many teams simply don’t want to bother with it on certain nights. NBA players have admitted that they do try less certain nights, so it wouldn’t surprise me if games vs the Suns or D’Antoni’s teams are often such cases.
However, when the playoffs come around and there are no back to backs and players are playing for pride, a title, and sometimes a contract, they give close to maximum effort and thus the payoff of this “relaxed” D is not only eradicated, but reversed.
In simple terms, my unprovable belief is that teams/players give less than average effort on average versus D’Antoni’s style during the regular season which leads to the numbers we see. When the playoffs come, teams give full effort and thus the regular season numbers are not indicative of what will happen in the playoffs.
Edmond
July 15, 2010
New York Won 29 Games last year. Using only the box score stats and minutes played, WP “predicts” that they should have won 31.28. Two games. Not bad right? That’s got to be within the margin accounted for by sheer randomness. If Lee’s defense was so terrible (or let’s say terrible enough to cancel out a statistically significant proportion of his positive, measurable contributions) shouldn’t we be seeing a greater disparity between the model’s prediction and the actual outcome?
Explain.
Michael
July 15, 2010
Great work. I love the graphics.
some dude
July 15, 2010
Edmond, the prediction is nearly 7.5% off!
But no, it’s not enough because the WP48 also underscores Jeffries and Gallo’s defensive abilities as well, so it can make up for Lee’s issues to the point of still arriving at 29 games.
Since WP admittedly underrates defense, some of Lee’s WP would surely be passed on to his teammates were defense better accounted for in the measurements.
ohreally
July 15, 2010
This is where the rubber meets the road on the limitations of box score stats, and stats like WP that are derived from box score stats…
Let’s assume NO improvement from Biedrins and that the Warriors C position is still a shambles. WP is saying that just Lee and D. Wright are worth 19 more wins.
That’s not going to happen. WP is great at accounting for inefficiency (like Ellis’ rough year trying to do too much), but it’s no good at accounting for poor defense, which Lee has in epic proportions.
If Lee plays big minutes the Warriors will not finish 19 games ahead of last season. I’ll be back then to either eat crow or ask WP fans to revise their view.
todd2
July 15, 2010
I didn’t see David Lee play last year. Is it possible his numbers were inflated by tempo? The Knicks assist numbers were better than their opponents, did he benefit from good setup guys? He posted about 4 fta’s/game, which is low for a 20 ppg guy. Does he take quick shots, shun contact or benefit from teammates drawing multiple defenders? The Knicks dfg% and rb%/margin were abysmal, should he take some of the blame?
dberri
July 15, 2010
For some reason, Jeremy’s comments on his own blog post were sent to spam. I found some other comments there as well. Please let me know if comments are not appearing. I don’t tend to look at the spam filter that often.
Joe
July 15, 2010
Do the doubters of this not see the 500 ifs?
Many of these ifs won’t happen. Nelson isn’t going to allocate the minutes like the author states most likely, they never do.
Injuries and possibly trades are other factors that will come into play.
The point is, if they stay healthy, and Biedrins remembers how to play, 50 wins is a possibility. And that makes sense to me.
marparker
July 15, 2010
I’m working on trying to predict 50 win teams.
Here’s what I found.
There were 4 guys who were worth 20 wins individually. Only one of their teams(Charlotte) won less than 50 games. They had Flip Murray playing nearly 1000 minutes of negative basketball.
There were 9 teams with 3 players who produced 7 wins or more. 3 of them failed to win 50 games. 7 of them won 50 games. The 2 teams that failed to win 50 games(Mia, Tor) had Barganini and Joel Anthony who each played over 1000 minutes of negative basketball.
GS already employs a 7 win basketball player from last year(Steph Curry) and employs Monta Ellis(10 win season high from 2 years ago) and Adrins Biedrins(13 win season high from 2 years ago). They have also added David Lee who no matter what anyone thinks of him has to be worth at least 7 wins though he has produced at least 15 wins for 4 straight seasons. However, Radmonovic has serious torpedo potential and he played over 700 minutes in only 33 games last year. Of course, if all 4 of those guys perform near their career highs then Radmanovic won’t be able to torpedo their efforts by himself. However, I also think that Steph Curry has some serious rookie slump potential.
Because of VladRad I’m going to have to say that I think GS is going to fall short of 50 wins.
Now, this has to concern Heat fans a little as maybe there is some evidence that negative players can kill an otherwise well put together team. However,Cle seemed to cancel out Ilgauskas’ 1300 negative minutes from last season. And Boston won the title in 08 with Glen Davis giving 900 plus minutes of negative basketball.
arturogalletti
July 15, 2010
sd,
WP does undervalue defense but it really isn’t wrong to do so. I did a step by step win regression from offensive & defensive stats ( see here http://arturogalletti.wordpress.com/2010/07/01/win-regression-for-the-nba-2/). Just using offensive boxscore stats I can get a basic model model that explains 85%+ of the variation in Wins. Defensive stats kick me up to the 90%+ level but they clearly are less important. David Lee is clearly a sub par defender as a Center and just ok as a PF but the stats are telling us that hey that’s not such a huge deal.
Italian Stallion
July 15, 2010
Since it’s virtually impossible to predict all the injuries, trades, and minutes, I think it’s sufficient to say that the author did a very good job of demonstrating that GS has the potential to be a much improved team this year. I certainly don’t disagree with that.
arturogalletti
July 15, 2010
Link to regression is
http://arturogalletti.wordpress.com/2010/07/01/win-regression-for-the-nba-2/
Sorry about that marparker.
marparker
July 15, 2010
no apologies necessary, I definitely wasn’t complaining. I appreciate your hard work.
Italian Stallion
July 15, 2010
I wish some of those peak Suns teams coached by D’Antoni weren’t plagued by key injuries and suspensions during the playoffs so we could have seen whether it would have changed anything.
Let’s discuss this offense/defense issue again in two years after the Knicks sign two more max players and their young talent is starting to peak. ;-)
Sheed World
July 15, 2010
CJ Watson signing with another team? Is it with the Lakers? So its true then that Warriors will trade Watson for Lakers pair of 2nd round picks.
ryder
July 15, 2010
Larry Riley let Morrow go mainly because they want to bring back CJ Watson. With CJ returning plus some improvement from Curry (I’m not buying the sophomore slump theory), I don’t see them losing any ground at the PG spot (we also have to remember that it took Curry about half a season to take off).
Monta will no longer be the #1 option with what should be a deadly Curry-Lee pick-n-roll. Maybe we won’t see Monta play as efficiently as in 07-08, but he should improve his efficiency significantly IMO.
Vlad will definitely not play as many minutes. I would’ve had a hard time seeing him get on the court at all if it weren’t for Udoh’s injury and Wright’s injury-proneness. Vlad also had an unusually bad shooting year.
But thanks for the write-up. I knew WP would project the Warriors to do really well and I’ve been waiting for this site’s take on it.
Both the new-look Warriors and Heat should be great test-cases for WinsProduced.
nerdnumbers
July 15, 2010
Jeremy,
Awesome post! I sincerely hope the Warriors do something foolish like trade Biedrins for K-Mart.
Also I do want to address the WP48 “underrating” defense. I disagree with this. Wp48 factors in defense but does so based on the team level. It then distributes this among the players based on their minutes played.
So people claiming well David Lee was bad at defense and other players were better on the Knicks IS a valid claim.
Last year the Knicks lost 10 games due to their poor defense. If you blamed Lee for every one of these losses he would still have generated 7 wins and you’d have to make the claim that no other player on the Knicks played defense. Clearly blaming one player for an entire team’s defense (including when he is on the bench) is foolish. Now Arturo is right, next year we’ll get a better chance to see if Lee’s “poor defense” is as a huge a factor as some people want to believe.
ilikeflowers
July 15, 2010
Mike Miller to the Heat is official.
“Game over man! Game over!”
Edmond
July 15, 2010
Yes. Great post. Great design. The Bay Area may have mediocre sports franchises, but you can’t beat our graphic design.
ohreally
July 15, 2010
nerdnumbers, I elided some detail in my earlier comment. The real way that WP48 doesn’t count defense (aside from assuming that everyone on the team defends equally, which is clearly false) is that it rewards players who skip defense for the purpose of gathering box score stats. Lee and Troy Murphy are both examples of this. Murphy has had some ridiculously big WP numbers because he ignores interior defense and just positions for a possible rebound. That boosts his rebound numbers but the defensive pain is distributed across all his teammates, so he comes out hugely ahead in WP rankings. It’s the same for Lee, who grabs all the easy rebounds while actually not helping his team’s rebounding percentage when he is on the court. He never boxes anyone out.
The excuses are going to fly next season when the Warriors don’t make the big leap. WP will assign the blame somewhere and arguing will begin. I just want people to step up now and admit this is a huge flaw in WP, or put their money where their mouth is take a stand on the Warriors really improving 19 games between Lee and Wright.
bduran
July 15, 2010
I’ve often wondered about how WP48 handles defense. If, in genearl, good defenders do things like get defensive rebounds, steals, and blocks then they will be rewarded for this in WP48. Sure, WP48 doesn’t assign value to how good you are at staying in front of your man or contesting shots specifically, but it assign value to things that seem like they would be highly correlated with these actions.
That just leaves possible outliers like a Bruce Bowen. Someone who everyone says is a great defender but generates few stats. It seems like this would be a rare situation and even then do guys like this really add more than they substract?
nerdnumbers
July 15, 2010
Ohreally,
Your logic is a little flawed. WP48 was built using linear regression to attribute how much given box score stats correlated with wins. So you may be right that David Lee and Troy Murphy “shirk” on defense to get good positions for a rebound.
However, a rebound is valuable! So the issue many people have when they point this out is that they have assigned a value to one on one defense and have ranked it higher than other stats. The numbers don’t back this up.
This is an interesting argument to use on WP48 and rebounds. Essentially you are saying that players are playing poorly to pad their rebound numbers. Oddly this is similar to a common theme on this forum that many players shoot poorly while padding their points numbers. As we’ve shown that WP48 doesn’t pay the bills like scoring, it is a very odd strategy for players to take.
Also you’ve put WP48 on the line based on the performance on one player in one year. I have no problem agreeing with Jeremy in his optimism for the Warriors. It is ridiculous to grade a metric on such a small sample though. Barring a major injury though I am willing to bet on the Warriors doing better than the Knicks by 10 games as it stands.
ilikeflowers
July 15, 2010
ohreally,
That’s some impressive backtracking.
Just to make your position concrete, you claim that there should be an inverse link between Lee and Murphy’s rebounding numbers and their respective team’s rebounding percentage, correct? This sounds testable.
Be sure to put your money where your mouth is as well, what’s your prediction for next season, including expected minutes played, and position? We already know that you think the Warriors’ DRB will decrease.
ohreally
July 15, 2010
nerdnumbers, those regressions are across the league. So if a lot of players are playing the game the right way, trying to win rather than stat whoring, then the value of a rebound in WP will be much higher than it really is with players like Lee and Murphy.
That is exactly why WP gives Lee and Murphy so many win shares.
WP rates Lee ahead of Wade, and WAY ahead of Dirk and Bosh. It gave Murphy 19 wins one season.
Sure, a rebound is valuable. But the next Warriors season is about to demonstrate that Lee’s rebounds in particular are not as valuable as WP states.
ohreally
July 15, 2010
ilikeflowers, you are off in the weeds. I never said any of that.
The Warriors were so horribly bad at rebounding last season that Lee will help them there. He just won’t help them as much as people think.
What do you think their team rebounding percentage will be next season? Do you think they will be average? Because I will bet money they will get < 50% of the boards.
bduran
July 15, 2010
arturogalletti,
Nice work. Really interesting data. One question though. You say offensive stats, but you include defensive rebounds, steals, and blocks which seem like defensive stats. I’m guessing you mean traditional box score stats vs. opponents stats?
This is something that Dave Berri has looked into before I think. That the traditional box score explains most of the variation in wins. Good to see again.
J. Scott
July 15, 2010
A quick perusal of the history of the Warriors (going back to the ’46-’47 season) reveals no previous instances of a +24 turnaround. That’s what a 50-win season would require following last year’s 26 wins. My guess is that in the history of the League +24’s are extremely rare. Memphis garnered nationwide attention for their +16 last year. Out of curiousity I checked the Milwaukee Bucks of 1969-’70. They were a 2nd year expansion team which added Kareem Abdul-Jabbar. They were a +29 that year. I know, I know…David Lee/Kareem…Same Difference!!! I’m … skeptical.
some dude
July 15, 2010
“WP does undervalue defense but it really isn’t wrong to do so. I did a step by step win regression from offensive & defensive stats ( see here http://arturogalletti.wordpress.com/2010/07/01/win-regression-for-the-nba-2/). Just using offensive boxscore stats I can get a basic model model that explains 85%+ of the variation in Wins. Defensive stats kick me up to the 90%+ level but they clearly are less important. David Lee is clearly a sub par defender as a Center and just ok as a PF but the stats are telling us that hey that’s not such a huge deal.”
No, they are not telling you that. You’re trying to apply the macro to the micro level.
Using offensive numbers may do 85%, but it doesn’t explain WHY those offensive numbers exist. Individual defense creates offensive production. Tim Duncan creates offense for Parker and Ginobili from his defense. David Lee doesn’t produce offense for anyone via his defense.
If you have a fundamental misunderstanding of how we get to those numbers, then you can’t begin to explain why those numbers exist and can predict.
So yes, while your model can predict wins to 85% (which seems low, but wtv) with just offensive production, it does not explain how we got there and how much defense led to that production.
some dude
July 15, 2010
“A quick perusal of the history of the Warriors (going back to the ’46-’47 season) reveals no previous instances of a +24 turnaround. That’s what a 50-win season would require following last year’s 26 wins. My guess is that in the history of the League +24′s are extremely rare. Memphis garnered nationwide attention for their +16 last year. Out of curiousity I checked the Milwaukee Bucks of 1969-’70. They were a 2nd year expansion team which added Kareem Abdul-Jabbar. They were a +29 that year. I know, I know…David Lee/Kareem…Same Difference!!! I’m … skeptical.”
Boston did it in ’08. Of course, they added Kevin Garnett, Ray Allen, James Posey, and others to a sub 20 win team.
J. Scott
July 15, 2010
Some Dude, thanks. I’m much too lazy to do any sort of in depth research.
Kevin Garnett…David Lee…Same Difference!
And… I’m still skeptical.
Twinkie defense
July 15, 2010
I’ll just mention Vlad Rad, for those who think he’ll get a lot of minutes for Nellie this season: Nellie HATES Vlad Rad. Says terrible, public things about him. He’s the resident Nellie doghouse guy, and is very unlikely to see big minutes this season.
ilikeflowers
July 15, 2010
ohreally,
This is what you said,
Lee and Troy Murphy are both examples of this. Murphy has had some ridiculously big WP numbers because he ignores interior defense and just positions for a possible rebound. That boosts his rebound numbers but the defensive pain is distributed across all his teammates, so he comes out hugely ahead in WP rankings. It’s the same for Lee, who grabs all the easy rebounds while actually not helping his team’s rebounding percentage when he is on the court. He never boxes anyone out.
So they negatively impact defense while positively impacting their rebounding numbers, correct? Wouldn’t better defense and boxing out tend to produce a higher DRB? I’m in the field right next to what you yourself posted, have you left the arena?
I don’t have any opinion on their DRB. wp48 isn’t tied to DRB. The post says that if the Warriors use their best players at their best positions using their current lineup and Biedrins is healthy then they will win around 50 games. In other words that 50 games is their likely ceiling barring unexpectedly high young player improvement.
Put your money where your mouth is is what you said. Talk the talk and walk the walk, my man. Predictions?
ohreally
July 15, 2010
Only thing I’ve heard Nellie say about Vlad was that he was trying to play through an achilles injury and that it wasn’t working, so he was shutting him down to heal.
ohreally
July 15, 2010
ilikeflowers,
Yes, they negatively impact defense while helping their own rebound totals, correct.
Would better defense and boxing out produce more rebounds *for them*, no. Incorrect.
arturogalletti
July 15, 2010
sd,
I regressed ten years worth of data using Team & Opponents stats. I found that value is skewed towards what you do as opposed to your opponent. So defensive variables (outside of steals,blocks, & your rebounding stats) hurt but not as much as you would think. So if you get scored upon at a high rate but generate possessions you’re a good player. I agree that Lee probably contributed a large part of the Knicks negative values in WP at Center but not all and I’m assuming he’s playing pf.
The questions are good though(I’m probably going to try to segregate more between pure defensive and offenssive stats in my next regression) so keep them coming.
ohreally
July 15, 2010
And I *already made* my prediction. Warriors will not improve by 19 games. This is the WP prediction for improvement based *only* on Wright plus Lee. So when they don’t improve this much, I don’t want to hear about how so many guys were injured (they were injured last year too), or how Biedrins didn’t return to form. I’m holding all those problems constant and not factoring in any improvement there.
Mike
July 15, 2010
ohreally,
I don’t really understand. So if Lee and Murphy play poor defense, then won’t their opponents score more on them, which leads to less opportunities for defensive rebounds?
And is there any evidence that shows their lesser defense costs their team more than the positive impact their defensive rebounding provides?
ilikeflowers
July 15, 2010
ohreally,
Would better defense and boxing out produce more rebounds *for them*, no. Incorrect.
Whew, good thing that I didn’t say that then. This is what I said,
Just to make your position concrete, you claim that there should be an inverse link between Lee and Murphy’s rebounding numbers and their respective team’s rebounding percentage, correct? This sounds testable.
Does this not follow from your claims? Isn’t it testable league wide for players that you identify as DRB saboteurs?
And we have you on tab for a prediction that the Warriors will win from 0 – 45 games. Unfortunately this won’t separate you from wp48 since if 50 wins is optimistic then something around 40 wins certainly sounds likely. Thanks for your insight.
ilikeflowers
July 15, 2010
Just to clarify, I’m referring to the team’s DRB.
arturogalletti
July 15, 2010
And the warriors were just sold:
http://sports.espn.go.com/nba/news/story?id=5383261
New owners are:”Boston Celtics minority partner Joe Lacob and Mandalay Entertainment CEO Peter Guber won the bidding. The deal had not been formally announced.”
If they bring in the Celtics management style they’ll definitely have stat guys (they might not listen to them but they’ll have them).
ohreally
July 15, 2010
ilikeflowers,
No, saying there is an inverse correlation, if I am understanding you correctly, is not implied by what I am saying.
Saying that their rebounding numbers are exaggerated over their true worth is not the same as implying an outright inverse correlation.
What I’m saying is very simple: their rebounds are not worth as much as WP thinks they are. And this is testable by watching the WP prediction for next season’s Warriors record.
ohreally
July 15, 2010
and ilikeflowers, the WP article is estimating 56 wins from the WP values. So your claim that 40 is reasonable seems to imply you don’t really believe the WP totals. Good to know. I’m with you on that one.
ohreally
July 15, 2010
Mike,
Yes, there is such evidence. The Pacers this season were worse when Troy Murphy was on the court, by a lot. And even looking *just* at team defensive rebounding percentage, and area which is supposed to be a Murphy strength, they were worse when Troy was on the court.
Lee I think is better than Murphy. The Knicks were worse when Lee was on court, but he at least slightly helped their defensive rebounding percentage, even though he hurt their overall rebounding percentage. But the short story is that his rebounding was not as good as it seemed, and it did not overcome his poor defensive contribution.
Edmond
July 15, 2010
ohreally:
But explain again why it wasn’t (isn’t) testable by last season’s Knicks record?
Also: I think everyone here understands that saying that the team could win 50 games according to WP doesn’t mean that injuries or rotation issues couldn’t decrease that number. If A.B’s pelvis falls off, then no, they will not win 50 games. I believe that is what Ilike means. A good stat “describes”, it does not “predict”.
ohreally
July 15, 2010
Edmond,
I think that WP correlates pretty well to team records in general. I just don’t think that it succeeds there for the reasons some think. There are guys like Murphy who break the mold a bit.
It gets hard to demonstrate this flaw in WP unless the right situations present themselves. One of the guys like Murphy or Lee has to switch teams for this to occur.
The only way I know to demonstrate this flaw in WP is to attack it’s predictions. If you are saying that predictions are off limits, to me that is the same as saying WP is meaningless, so I can’t follow you down that path.
Edmond
July 15, 2010
That’s fair. There was an interesting article here a while ago about the predictive power of basketball and baseball statistics. Basketball players tend to be much more consistent between seasons than baseball players, but baseball statistics are very accurate at “describing” player performance nevertheless. Or something like that. Anyway, find the article–its great. Its funny how alot of the most bitter arguments here tend to boil down questions that sort of look…I don’t know…epistemological. Where’s David Hume when you need him.
Mike
July 15, 2010
ohreally,
Are you getting this from 82games.com? How do you separate Lee or Murphy from their teammates? Are you strictly looking at plus/minus? Just wondering what metric you are using to make these claims.
Also, I’m not seeing the “stat whoring” claim as legitimate. Research shows that players get paid by their scoring, so it makes little sense that players would purposely play poor defense just to get rebounds, when it won’t get them paid. I’m not sure what the correlation of rebounds to pay is, but I’m sure Dr. Berri has some old articles about it.
ohreally
July 15, 2010
Here is a possible case study where we already have the results: what did WP predict when Danny Fortson got traded to the Warriors a decade ago? Fortson was another empty stat guy, and the Warriors did not improve with his presence. I wonder what WP predicted?
ohreally
July 15, 2010
Mike, yes, I’m using 82games. plus minus is far from perfect, but it does have one great virtue: it counts everything. Yes, it counts a lot of things that you don’t want to count. If player A has a high correlation in his minutes with player B (or an inverse correlation, like B always subs for A) then plus minus ends up counting a lot of stuff you don’t want. But that curse is also a blessng, in that if there is something important about a player that isn’t captured in box score stats, it has a chance of showing up in +-.
Regarding stat whoring not being legitimate: players do what they can. Troy cannot defend the paint, *and Troy knows this*. So rather than try to do that and end up surrendering a lot of and1 plays and picking up a ton of fouls, he just goes for the rebounds. It doesn’t matter what his motives are. He probably thinks he is doing what is best for the team given his abilities. But the effect is that his defense hurts the team mightily, and his rebounding does not make up for it. In short, his abilities are not up to the value that WP indicates, and this is because WP spreads his horrid D over the other guys on the court instead of giving it all to Troy.
some dude
July 15, 2010
“I regressed ten years worth of data using Team & Opponents stats. I found that value is skewed towards what you do as opposed to your opponent. So defensive variables (outside of steals,blocks, & your rebounding stats) hurt but not as much as you would think. So if you get scored upon at a high rate but generate possessions you’re a good player. I agree that Lee probably contributed a large part of the Knicks negative values in WP at Center but not all and I’m assuming he’s playing p”
The problem with this claim is the assumption that Lee is generating possessions. There is no evidence that he’s taking rebounds from his opponents rather than from his teammates. And he’s also contributing to the Camby effect where he gives up higher % looks on the hope that he gets the rebound.
The thing is, to really hit the nail on the head, we’d have to go through a synergy style method where we break down the plays into how they’re done.
every stat, besides FT% for the most part, has a potential negative to them. There are such things as over-rebounding or doing too much to get a rebound. I know people don’t want to think so, but it’s true.
ilikeflowers
July 15, 2010
ohreally,
If they would play defense more and cherry pick rebounds less, then presumably either their individual DRB’s or DRB percentages would decrease and the number of available team DRB’s and percentages would increase.
Specifically,
[1] There are more available defensive rebounds for the team.
[2] The team grabs a higher percentage of the available defensive rebounds.
[3a] Their individual DRB’s go down since they’re now fighting for harder rebounds and helping their teammates get a larger share.
[3b], Their individual DRB’s stay the same or increase since they’re causing more available rebounds but just getting a smaller percentage of them. In this case, their individual DRB percentage decreases.
So, the relationship will be either more individual defense/boxouts and less individual total DRB’s or more individual defense/boxouts/DRB’s and a lesser DRB percentage. Take your pick.
Now we just need to identify teams that had a relatively stable roster while losing a cherry-picker and replacing them with a comparable non-cherry picker. Any team past or present fit this bill anyone?
So your claim that 40 is reasonable seems to imply you don’t really believe the WP totals.
No, this is incorrect. I was simply using the 50 wins that figured prominently in the post as a starting point and assuming that the minutes won’t be well allocated or the players will be played out of position, etc… Since the ideal situation is actually 55 wins, then I’m on board for a ceiling of 55 if everything goes their way as presented in this post, and something like 45 wins as more likely (I’m sure we’ll get a more detailed analysis of the various likely scenarios later). And we have you on board for 0 – 44. So 45 (+/-4) sounds good, still a bit of overlap there, so you too might agree with a wp48 approach in spite of the dang cherry pickers.
ilikeflowers
July 15, 2010
The flaw in this test of course is the eyeball model of determining who’s a cherry picker, but you have to start somewhere.
ohreally
July 15, 2010
ilikeflowers,
Yes, you are capturing my meaning very well in your numbered bullet points.
And yes, I will take a stand on 44 wins or less, assuming no more major changes in personnel on the Warriors.
WP is on record for 45 to 57.
May the best man win! ;-)
ohreally
July 15, 2010
ilikeflowers,
Fortson played limited minutes for the Celtics before coming to the Warriors. I’m curious what WP48 predicted there. It might be a good test case. I’m not sure he was efficient enough though to demonstrate the problem… maybe WP didn’t love Fortson.
arturogalletti
July 15, 2010
Here (I can’t believe I just looked up Danny Fortson):
Name Year Team Age Sum MP Sum WP Total WP48
Danny Fortson 1998 DEN 21 1811 -0.2 -0.006
Danny Fortson 1999 DEN 22 1417 10.1 0.342
Danny Fortson 2000 BOS 23 856 6.4 0.359
Danny Fortson 2001 GSW 24 203 2.4 0.577
Danny Fortson 2002 GSW 25 2216 11.5 0.250
Danny Fortson 2003 GSW 26 223 0.1 0.030
Danny Fortson 2004 DAL 27 625 2.6 0.202
Danny Fortson 2005 SEA 28 1047 5.7 0.262
Danny Fortson 2006 SEA 29 276 0.1 0.019
Danny Fortson 2007 SEA 30 158 -0.1 -0.039
some dude
July 15, 2010
“Now we just need to identify teams that had a relatively stable roster while losing a cherry-picker and replacing them with a comparable non-cherry picker. Any team past or present fit this bill anyone?”
If I am correct, the Pistons had the same DRB% the year after Ben Wallace left replacing him with Webber, Nazr, and Maxiel.
But they’re defense suffered because they committed more fouls than Wallace did.
ohreally
July 15, 2010
Another test case is to look at players who leave teams that have a Fortson or Murphy on them. I picked the first that came to mind: Larry Hughes. He played with the Warriors during Fortson’s “prime”, and then went the Wizards. His WS rating on basketball-reference.com jumped when he got traded. It’s easy to say that improved, but I suspect escaping Forston’s aura of WS depression played a part.
Interestingly, one of the components that jumped the most after the trade was his defensive rebounding percentage. Maybe Fortson was no longer stealing his boards?
some dude
July 15, 2010
their*
ohreally
July 15, 2010
Danny Fortson 2000 BOS 23 856 6.4 0.359
Danny Fortson 2001 GSW 24 203 2.4 0.577
Danny Fortson 2002 GSW 25 2216 11.5 0.250
So the first season after the trade Fortson played only a handful of games before getting hurt, so toss that one out.
the question is, if you knew a guy with a 0.359 WP48 rating was coming to the Warriors to play 2200 minutes, what would the prediction have been for their record? Because their actual record was 19-63.
Mike
July 15, 2010
ohreally,
The Warriors had 19 wins in 2000 before Fortson. In 2002 they had 21 wins. It looks like Donyell Marshall’s wins got replaced by Danny Fortson, while the rest of the team was still miersable. I don’t have the numbers and hope Arturo can come to the rescue here.
ohreally
July 15, 2010
Mike, thanks for the correction. I’m going from memory here sometimes, and when it comes to bad years for the Warriors, it’s easy to get confused.
The good news for me as a Warriors fan is that the team just got sold. But that’s a story for a different thread…
arturogalletti
July 15, 2010
So GSW 2001 (17W):
Name Year Sum MP Sum WP Total WP48
Mookie Blaylock 2001 2352 7.52 0.154
Antawn Jamison 2001 3394 3.41 0.048
Adonal Foyle 2001 1457 2.65 0.087
Danny Fortson 2001 203 2.44 0.577
Marc Jackson 2001 1410 1.92 0.065
Adam Keefe 2001 836 0.85 0.049
Bob Sura 2001 1684 0.69 0.020
Larry Hughes 2001 1846 0.46 0.012
Chucky Brown 2001 339 0.24 0.034
Randy Livingston 2001 7 -0.03 -0.238
Chris Mullin 2001 374 -0.24 -0.031
John Coker 2001 32 -0.28 -0.421
Bill Curley 2001 192 -0.34 -0.085
Chris Mills 2001 493 -0.44 -0.043
Chris Garner 2001 149 -0.75 -0.241
Chris Porter 2001 1147 -1.59 -0.067
Erick Dampier 2001 1038 -1.62 -0.075
Vonteego Cummings 2001 1495 -1.86 -0.060
13.03
So GSW 2002(21 W):
Name Year Sum MP Sum WP Total WP48
Danny Fortson 2002 2216 11.52 0.250
Bob Sura 2002 1779 5.46 0.147
Antawn Jamison 2002 3033 3.86 0.061
Gilbert Arenas 2002 1155 3.24 0.135
Larry Hughes 2002 2050 2.55 0.060
Adonal Foyle 2002 1485 2.17 0.070
Jason Richardson 2002 2629 0.90 0.016
Chris Mills 2002 1237 0.88 0.034
Mookie Blaylock 2002 599 0.43 0.035
Cedric Henderson 2002 70 -0.14 -0.095
Dean Oliver 2002 139 -0.56 -0.193
Erick Dampier 2002 1740 -1.68 -0.046
Troy Murphy 2002 1448 -1.70 -0.056
26.92
It a fairly large outlier but I figure that the Presence of Fortson and Murphy on the same roster probably warped the space time continuum.
ohreally
July 15, 2010
arturogalletti, thanks for running those numbers, but that isn’t what I meant.
Those are after the fact numbers (and interesting that they give 11 wins to Fortson, who I’m pretty sure actually hurt the team).
The critical question is what did WP *predict*, before the trade, that 2200 minutes of Fortson in (and Donyell out) would do. That is the context that we have here in this thread, with WP predicting 14 more wins from Lee coming in.
ilikeflowers
July 15, 2010
ohreally,
He’s a young 350, gets injured and can “only” manage 250 for the rest of his career in between being injured. The prediction would be based upon a 250 player, but we need the rest of the team’s info.
For the cherry-picking this is about as bad a test case as there can be from the GSW side due to the injury coming in between the team switch. Who took his spot in Boston?
ilikeflowers
July 15, 2010
The injury between the team switch blows away any predictions. We need some different players.
ohreally
July 15, 2010
ilikeflowers,
So you believe that his 2002 season was depressed by lingering effects of the 2001 injury? I disagree. He certainly had a strong season from a box score perspective, and played plenty of minutes – more than Larry Hughes for instance. I saw no signs of limitations from injury when I watched him that season.
I think his WP48 went down precisely because he no longer had better defensive players around him to prop him up.
arturogalletti
July 15, 2010
ILS,
Tell you what, I’m willing to find you data if you’re willing to write a post. I’ll even publish it if DJ won’t (although I’m sure he will).
ohreally
July 15, 2010
I’m a newb here, so apologies if this is a dumb question:
Who is ILS?
arturogalletti
July 15, 2010
Sorry I meant (ILF or ilikeflowers) and actually oh really as wel.
Mike
July 15, 2010
ohreally,
I didn’t know you were a Warriors fan! No wonder your hatred for T-Murph runs deep (me too). However, I think you may be pleasantly surprised at what the current lineup can do. I choose to remain optimistic, the Warriors really never had 4 out of 5 starters have career WP48 above average in recent years (at least not to my knowledge).
ohreally
July 15, 2010
Mike,
I will be very happy to be wrong, and will happily eat crow here if Lee propels us to 45+ wins.
I am optimistic, but my optimism is longer term, after the new owners clean up the mess.
khandor
July 15, 2010
Somewhere up above, the following words can be read: “Put your money where your mouth is.”
In this regard, and relative to the content of this specific article, I have $100.00 in my Palpal account which I would be more than willing to wager against anyone who claims/predicts that the Golden State Warriors will win 50 [or more] games during the 2010-2011 NBA regular season. i.e. My bet is that the Warriors will, in fact, win something less than 50 games this coming season. If there is anyoone here who is willing to “put their money where their mouth is,” on a call of this type, made in advance, on this site, please, just let me know.
[NOTE: Italian Stallion, do you think there will be any takers?]
ohreally
July 15, 2010
Italian Stallion’s post on D’Antoni playing for point differential instead of maximal defense is something I have thought before about Don Nelson’s approach to the game.
And when I play basketball, it’s something I engage in myself. Only in my case, the strategy runs the other way: I cheat out on defense instead of offense, because me and my peers play so badly that fast breaks are about 4 times as likely to result in a basket, and it’s important to prevent that if you want to maximize your scoring differential.
ohreally
July 15, 2010
Clarifying: I meant above that if you allow the other team to fast break, they are about 4 times as likely to score then if you don’t allow that and they have to attack a set defense.
Suomynona
July 15, 2010
lol khandor I don’t normally participate in these discussions, although they are interesting. but the wp48 people are saying that around 50 wins is the warriors ceiling assuming that everything works out. no injuries, everyone does what they did last year, except monta ellis who does what he did 3 years ag0, minutes assigned logically, and so forth. so pretty much they’re saying that 50 wins could happen if the warriors get incredibly lucky, or at least that’s how i read it. so saying you’re willing to bet that they don’t win 50 isn’t going out on a limb too much.
Owen Ellickson
July 16, 2010
Awesome article. Having said that, ohreally has this nailed, and I’m surprised a diehard Warriors fan like Jeremy wouldn’t be wise to the hollowness of Lee’s numbers.
It is almost never the case that a player with excellent production statistics doesn’t help you win… but it does happen. There are a couple players whose systemic effects on team defenses are so bad that they essentially sink your chances of competing. And the banner-carrier for that group is Troy Murphy, a guy who rates brilliantly by WP48, and a guy who’s dragged down every team he’s ever played for.
The ’09-’10 Pacers were outscored by 5.4 points per 48 minutes when Murphy played, and outscored their opponents by 0.5 points per 48 when he sat; they were a horribly ineffective team with him, a slightly above-average team without him. Murphy’s -5.9 (actually -6.0 with rounding) net plus-minus was one of the worst of any NBA starter. This wasn’t exactly a new phenomenon for Murphy, either. His net plus-minus by year, going back as far as 82games has data:
’02-’03: -3.5
03-’04: -2.8
’04-’05: -4.4
’05-’06: -2.6
’06-’07: -6.7 (Warriors), -3.8 (Pacers)
’07-’08: +0.0
’08-’09: +0.7
’09-’10: -6.0
Troy Murphy’s production has always been impressive. And yet, in eight NBA seasons, only *one* of his crappy teams has actually played better with him on the floor, and then only slightly; in one other year, his presence didn’t make a difference either way. The other six years, Troy Murphy’s teams suffered from his presence — they were crappy in general, but extra-crappy when he was out there. And the reason was his defense: Murphy is so intent on securing rebounds that he lets opponents traipse straight to the hoop, murdering his team’s chances of success. If that’s a good player, I’m a lug nut.
David Lee is better than Troy Murphy, but he’s the same breed: a guy whose defensive rebounding belies his pernicious effects on that end, a guy whose team plays better without him as often as not (each of the last two years in Lee’s case). WP48 completely misstates the contributions of guys like these, and as such is wildly off about the potential of the new-look Warriors, no matter how much I wish it wasn’t. They won’t win 40, and they won’t sniff 50.
marparker
July 16, 2010
Khandor,
Why would anyone take your even bet at 50 when the vegas line is going to be much lower and shaded towards the under?
Owen,
I was looking at Murphy’s +/- and a couple of things came to mind. I’m not saying your incorrect. Murphy played a ton of center last year. He is not a center. Secondly, when I go look at his most used lineups I don’t see anything like that huge number of 112 per 100 possession defense. What I do see is that some of his least used lineup(the ones the coach probably doesn’t want to used but gets forced to) have horrible defensive numbers like 123 and 132 next to him. Doesn’t that skew the numbers?
reservoirgod
July 16, 2010
I think the moral of the story in this thread is that white PFs suck.
khandor
July 16, 2010
Suomynona,
What exactly is the point of saying something like this:
“If team X gets ‘incredibly lucky’ and every break tips in their favour, then, they are likely to win 50 games next season,”
in the first place?
Secondly, I did not say that a wager placed on the Warriors failing to win 50 wins next season was akin to “going far out on a limb”. Did I?
———————
marparker,
I might be amenable, if any would-be takers were interested in lowering the Warriors “predicted total wins number” by a nominal amount.
What number do you [or, other ‘supporters’ of an article of this type] have in mind?
arturogalletti
July 16, 2010
khandor,
I think more than a few of us are interested. I’d like to see how the roster finally shakes out though before committing to anything. Let revisit this in a couple of months.
khandor
July 16, 2010
arturogalletti,
No problem. Whenever you are ready, you know where you can reach me.
Cheers
Jeremy Britton
July 16, 2010
Is David Lee so bad on defense, his wins produced numbers lie?
That’s essentially the thread I’m seeing here. (I can’t believe that the ghost of Danny Fortson was resurrected for this discussion!) I’ll going to go back to earlier comments by nerdnumbers to make a simple point:
Taken as a whole, distributed across whole teams and the entire league, rebounds represent evidence of “good defense.” Somebody missed a shot and your guy (mostly Lee in this case) got the ball back.
I know, I just grossly oversimplified the problem. But that’s my point. Wins produced teaches you to do exactly that when evaluating how to win basketball games. Less matters than you think. Rebounds happen to matter a lot.
There’s this very strong perception that Lee is extremely “bad on defense,” but this seems to get back to “lying eyes” simply not believing what the analysis tells them. That kind of skepticism is healthy and I’m not going to say don’t do it, but the fact is right now that the wins produced model says the Lee-bad-defense assumption is the wrong thing to focus on if you want to tell how many games he helps you win.
And I do like viewing this new version of the Warriors as a test case for the model, but know too many outlier factors will also influence the season.
Jeremy Britton
July 16, 2010
@Suomynona
You misread a few things about the article:
1) 50 wins is not the “ceiling” for the Warriors next year, it’s a number that actually rounds down a moderate guess for what this roster is capable of.
2) The “logically assigned minutes” were just rough guesses rolling forward what players played last season, for the most part.
3) It factors in Monta Ellis playing the same as he did last year–no change, not very productive.
And 4) I’ll add, it factors in a partly healthy Biedrins (plays 60 games like 08-09) and doesn’t factor in any improvement from Curry or Williams, which you’d expect.
@khandor
I’m actually putting my money where my mouth is by buying tickets to Warrior games this year. ;)
Owen Ellickson
July 16, 2010
“There’s this very strong perception that Lee is extremely “bad on defense,” but this seems to get back to “lying eyes” simply not believing what the analysis tells them.”
It’s not about not believing analysis; it’s about using more than one type of analysis. Can’t a WP48 adherent be accused of simply not believing what plus-minus analysis tells them?
By WP48, Troy Murphy was the 12th-best player in basketball last year — better than Kevin Durant, better than Rajon Rondo. And yet the Pacers were hapless when he played and downright competitive when he didn’t… their chances of winning soared when he sat. This has been the case throughout Murphy’s career, and it’s because his defense is far worse than can be captured by simple statistics.
Any system that fails to flag a player like Murphy — or Lee — is an incomplete system.
ilikeflowers
July 16, 2010
ohreally,
It’s simply a very bad test case. Why on Earth would you ever choose as a test case the situation where a player had an injury between teams? It just doesn’t make any sense. Injuries add a great deal of uncertainty and no one’s eyeball model reassurances are going to alleviate that issue. Certainly there are more suspected cherry pickers who switched teams than Fortson? It’s only good analysis to choose cases in which the number of unaccounted for variables are minimized.
ag,
Hah! That sounds like a challenge! I have to decline though. I’m actually thinking about implementing the model myself so that I can identify situations in which teams are under or over performing their efficiency diffs. I already have the data, I just need to spend the time to write the code. I have another personal project that I really should work on instead, but we’ll see.
ohreally
July 16, 2010
ilikeflowers,
So there are two reasons I chose it. First, it was the first name that popped to mind. So it was unbiased… I didn’t know what the stats would show: I didn’t cherry pick, looking through WP48 stats to find the biggest outlier.
And second, the basic issue here is that it’s rare for a player to be in this situation, where they generate tons of box score stats while hurting their team mightily. So I don’t get many guys to choose from. I have to take what I can get and if Fortson had a year injured right when he was traded, then that sucks but it’s still the best I can do so far.
The whole problem with using WP48 to make the kind of predictions this article makes is causality vs correlation. Typically coaches and GMs won’t play terrible players big minutes. This prevents those players from ever accumulating big box score stats to start with, which means the players who do accumulate those stats will do so in a way that correlates well with winning, and in a way that lines up with WP48 as a predictive tool. But when you get the rare exceptions to this rule (Fortson, Murphy, and to a degree Lee), you end up having a team that doesn’t get many wins, and where that one guy gets way more than his real portions of credit from WP48. The only way to demonstrate that WP48 misvalues them, that I can think of, is to observe them switch teams and note that the new team doesn’t do as well as predicted.
If I’m right, Lee will come into GS, GS won’t do as well as predicted here, and Lee will steal WP from other GS players due to his ability to accumulate box score stats while not playing very well.
I’d give you a better test case than Fortson if I could. I guess we’ll have to wait for the results from Lee.
ohreally
July 16, 2010
One other possibility to look at is when Troy Murphy went to the Pacers, but I suspect that one may be problematic because so many players were involved.
ohreally
July 16, 2010
“That kind of skepticism is healthy and I’m not going to say don’t do it, but the fact is right now that the wins produced model says the Lee-bad-defense assumption is the wrong thing to focus on if you want to tell how many games he helps you win.”
Jeremy, that reads to me like “WP is right because WP predicts it will be right.” I think you need a little more of that healthy skepticism!
WP was wrong about Fortson. It was wrong about Murphy. And I think we are about to see it be wrong about Lee. And this time, we have folks out in front of the event, to make & test the predictions and see just how much WP is worth.
Burney
July 16, 2010
David Lee is a pick n roll guy who can finish with either hand at the rim, and he also hits his jumpshot on a kickout. Get ready for a steady diet of pick n roll offense with Lee (and Biedrins) and Steph Curry. This excites me. A properly executed pick n roll offense is virtually unstoppable (see Stockton/Malone). Pick n roll WINS GAMES and championships (see Gasol).
Jeremy Britton
July 16, 2010
@ohreally
“Jeremy, that reads to me like “WP is right because WP predicts it will be right.” I think you need a little more of that healthy skepticism!”
No, it’s really more to say you roll with the best you’ve got and keep an open mind that something will improve on it. Skepticism can be healthy or paralyzing. You’ve gotta ask, skepticism to what end?
“WP was wrong about Fortson. It was wrong about Murphy. And I think we are about to see it be wrong about Lee. And this time, we have folks out in front of the event, to make & test the predictions and see just how much WP is worth.”
All I’m hearing here is a lot of declarations about a foregone conclusion: Lee and Murphy and Fortson are bad on defense and pad stats by calling their own number on the glass. I’m open to that possibility, but it’s not certain. The burden of proof is on you guys. Show it! I want to see it.
ilikeflowers
July 16, 2010
ohreally ,
WP was wrong about Fortson. It was wrong about Murphy.
You have demonstrated none of this. These are strong statements of opinion as fact from a random internet dude. Show us why this is true. You haven’t provided the slightest bit of hard evidence for anything that you’ve said – not even for the case of Fortsom. It’s your claim, back it up, or just present it as a possible issue instead of a fact.
Ultimately, systematically examining team efficiency diff or wins against expected diff/wins from wp48 may (or may not) uncover specific cases in which there are additional predictable factors at play. As for this cherry picking rebounder hypothesis, it’s just that, a hypothesis awaiting evidence.
ohreally
July 16, 2010
What would be a demonstration that you would believe?
The plus minus data is pretty convincing if you know how to read it (which I admit is difficult).
The team rebounding percentage differentials on/off court are pretty good for showing these big rebounders are really fake rebounders.
And the shifting team record projections, and WP values of the players themselves and the players around them when switch teams are another indicator (in Fortson’s case, and possibly soon in Lee’s case).
You aren’t giving me much to work with here when you just dismiss these as if they don’t matter. If the only thing you care about is correlation value from the latest WP regressions, then I can never prove anything, right?
ohreally
July 16, 2010
that last was to Jeremy.
ilikeflowers got in a response while I was typing…
ilikeflowers
July 16, 2010
reservoir god,
We also need to look into why white PF’s are being played at center. It’s a conspiracy!
ohreally
July 16, 2010
ilikeflowers,
I’ve laid out the method for demonstrating that WP likely flubbed the prediction on Warriors wins when Fortson was traded. I don’t have a tool to crunch those numbers.
You seem to have already indicated that you won’t believe the numbers prove anything if I did have them, so I have no incentive to try to generate them.
ohreally
July 16, 2010
Or, to be more accurate, Jeremy has laid out the method. ;-)
Just copy what he did, and see what you get.
reservoirgod
July 16, 2010
ILF:
Yes. Yes, it is. Lmao.
arturogalletti
July 16, 2010
ILF & reservoirgod,
Funny how it’s always white big men. Selection Bias anyone?
marparker
July 17, 2010
I missed alot by actually working all day yesterday. It would seem to me that the easiest way to debunk a method is to use the method to debunk itself. All methods produce a curve of success. Wp48 is predictive model with relatively thin tails. Its debunkers try to point to the tails as proof it doesn’t work. I contend the debunkers must find the meat of the curve and debunk that if they want to actually prove anything.
What we have is a method without very many followers. This opens wp48 to alot of criticism and in this world of one person one vote that means wp48 is not likely to ever be cast in a positive light.
I’m with Khandor. If you like it so much put some money behind it. I won’t be betting on the Warriors unless a really attractive line is available around October. However, that Miami Heat bet is looking very attractive.
some dude
July 17, 2010
I don’t think the tails matter.
What matters is that it’s predictive rate is better than other advanced metrics. If it’s isn’t, then why should we use it for predictions if there’s a better method out there?
I’m not claiming there is one, but that’s the truth.
Though, I do lend some credence to the earlier argument if even in the tails is ridiculously way off base. Say Warriors win 25 games this year. That’s a bit too far off to not think WP48 isn’t doing something wrong at some level.
marparker
July 17, 2010
Some Dude,
Lets say David Lee’s production measured by WoW falls off a cliff. Are we going to use one player to debunk an entire model? I don’t think that would be entirely fair. Noone is claiming the model to be perfect. Yet it seems its detractors are trying to prove that it’s not perfect. What model is perfect?
I’m curious to see what David Lee and all other movers do relative to what they did with their old teams. How many players move to better teams and see their wp48 drop? It would seem that shouldn’t happen. Then for the guys that are moving to worse teams, how much of a drop if any can be expected?
some dude
July 17, 2010
If WP predicts at a worse rate than another tool, then it becomes a worthless tool.
I wouldn’t use the term “debunk,” as debunk refers to shooting down a theory or argument, but it would render it obsolete.
And never did I say Lee’s production dropping off (actually, I don’t know that it will, just saying he might be the same and Warriors no better off really), then it just tells us we need more adjustments in the model, not that the model is debunked.
I would probably look at the team defensive stuff and try to adjust it to the individual players in some way.
marparker
July 17, 2010
some dude,
I keep a percentage of wp48 for offense and defense for individual players. Its a long formula put through my spreadsheet where I cant remember where half the calculations came from but it kind of points me towards an idea. For instance I have David Lee as a high offensive percentage player. His offensive rating suggests that his wp48 should be higher. So he must get most of his productivity on the offensive end(hypothetically by my thinking).
Now take someone like KG who has a much higher wp48 career wise but a lower offensive rating. KG is a higher percentage defensive player. Rondo is the king of defense of my rating. He has had a below average offensive rating and yet had an very good wp48. These two inspired my spreadsheet.
These players tend to do tremendously well when switching teams or having their teams switched on them. At least according to my limited data. And, I consider these players to be more reliable. We just reached a point where reliable end of year wp48 data was available readily. So I haven’t had time to research as deeply as I would like.
Anyway, NY was a terrible defensive team last year and yet Golden State was even worse. My point being David Lee already gets minimal help from the team defensive adjustment so him going to Golden State shouldn’t kill his wp48. However, he definitely won’t be creating any excess team adjustment to be redistributed to his teammates. This cannot be good when we realize that this is the exact same problem that Biedrins has. Now, that doesn’t mean that their offense won’t be awesome. But, I don’t see how this helps a player like Monta Ellis who only had a good year when Baron Davis decided to grace us with playing up to his potential.
ohreally
July 17, 2010
The main issue to me is that WP hamstrings itself by limiting the inputs to boxscore data.
Why not build a model that tracks on/off data? That way a correlation between player A and the opposing teams offense could be tracked better.
This isn’t rocket science. As soon as I read how WP worked, I knew immediately that it would be broken for Troy Murphy. And wa-la, it says Troy is a superstar. If I can see this the moment I read Wages of Wins, before even looking at any data, then WP is not good enough to be useful to me.
marparker
July 17, 2010
Isn’t whatever model you choose gonna severely overrate/underrate some player? If you pick and choose different models for different players aren’t you essentially just using the eye test?
marparker
July 17, 2010
Also, I’m not sure that its being fully understood that this model was built via regression(i.e its based on your basic laws of mathematics, i.e if this isn’t correct than the roof over your head is gonna fall on top of you).
I will add that I came here in search of profit. I could never build a model that could beat my instincts which by the way weren’t worth a damn. Then I built a model based around WoW’s consistency and that hasn’t been a problem since. Obviously, as with any model, there needs to be a way to fish out the bad results.
Again, I think you are overrating the Troy Murphy rating. I don’t think you are giving the model enough credit for what it gets right. I don’t think its fair(or smart) to point out the Troy Murphy flaw but then not also point out the Ben Wallace’s of the NBA. It may overrate a handful of players but I wouldn’t hesitate to claim that the model has at least 80% of the player correctly rated.
If you design a model to resist ridicule then you are essentially just building a sentiment indicator and one only needs to poll to discern that.
some dude
July 17, 2010
80% is too low for me. One starter on every team is not aptly rated by that.
Now, 80% might be more than another metric, but the thing we have to do is get it closer to 100%. My issue is that we won’t advance unless we accept the limitations and then try to eradicate them.
On a sidenote to the Lee thing, I disagree with NYK being a terrible defensive team outside of Lee. At least pre-Jeffries trade, they were average outside of Lee. But when you have post defense that atrocious, there’s nothing you can do. As I mentioned earlier, 26% on average his counter did better than their average against Lee. That’s a staggering number.
marparker
July 17, 2010
And fwiw, I’m just using Troy Murphy as an example of others thoughts. I personally think the guy has never gotten to play with another .2 player his entire career save a 19 year old Biedrins who only played 1000 minutes for the Warriors that season. Then he was traded midway through the next season.
marparker
July 17, 2010
80% is just a number I threw out there. But, that doesn’t matter. And there’s is plenty I could do with that number as a GM. As long as the pareto principle holds true I have a very good chance of putting together a productive squad if I know its a given that a model is 80% correct.
Eliot
July 17, 2010
It is reasonably well known that WoW does show preference for big men with good rebounding numbers, high shooting percentages, and lowish turnover rates, while most other advanced statistics and basketball in general assign a lesser value to these kind of players for various reasons (low scoring numbers, poor defense, lack of “skills”). While WoW may overrate these players in some respect I feel that it is closer to their true value than most other metrics and the general perception of these players. This disconnect seems to stem from watching the game, as these players excel in the most mundane parts in the game, while the most overrated players tend to be terrible at these aspects but great at the much flashier parts of the game. It is after all more exciting to see a player score a boatload of points than to see a big man gobble up rebounds and take high percentage shots (expect if that man is Dwight Howard). That makes it harder for many fans to believe that Ben Wallace can be a more valuable player than Carmelo Anthony, and that leads to much of the backlash against WoW as it goes against the common knowledge and perception of the game.
ohreally
July 17, 2010
Also, I’m not sure that its being fully understood that this model was built via regression(i.e its based on your basic laws of mathematics, i.e if this isn’t correct than the roof over your head is gonna fall on top of you).
Bachelors in Mathematics, highest honors. I understand how the model was built.
Warriors fan, 15 years. I understand the nba game, and particularly overrated players. ;-)
This makes me well suited to sniffing out problems with basketball player ranking metrics.
marparker
July 17, 2010
Ohreally,
Then why are you questioning possible outliers of a mathematical model that explains 90 plus percent of what it is trying to measure?
And lets all not get into a knob measuring contest of who has which kind of degree. I’m 100% positive there are some extra fancy degrees from extra fancy higher learning factories in this comment thread.
nice homes
July 18, 2010
Also, I’m not sure that its being fully understood that this model was built via regression(i.e its based on your basic laws of mathematics, i.e if this isn’t correct than the roof over your head is gonna fall on top of you).
ohreally
July 18, 2010
Then why are you questioning possible outliers of a mathematical model that explains 90 plus percent of what it is trying to measure?
This is a recurring refrain here. Anyone who points out problems with WP is told it’s “good enough”. Maybe for you.
I don’t care how well it correlates with winning. If it tells me a bad player is a star, I can’t trust it, and I can’t use it.
And lets all not get into a knob measuring contest of who has which kind of degree.
You were the one who questioned my understanding. “I’m not sure that its being fully understood that this model was built via regression”. So don’t complain when I tell you that I do understand.
It doesn’t matter (to me, at least) that the weights of the box score components are tuned via regression to be the best that they can be, if box score stats aren’t sufficient to reliably differentiate good players from bad.
I do wish that we had both been GMs all these years. I would have acquired Forston, Murphy, and Lee on the cheap from the other GMs who understood their value, and then traded them to you and looted your team in the process.
marparker
July 18, 2010
Ohreally,
I meant that regression as a joke. I wasn’t trying to be an ass though I can see how it sounded like that, I apologize. I was trying to make light by stating something obvious not point out that you don’t know that something obvious. It just seems to me you don’t like the model because
of the kind of player it “overrates”. If the model overrated Carmelo Anthony or Danny Granger you wouldn’t have a problem with it? Thats a serious question not a nit.
I find it confusing because implicitly models can’t be 100% predictive. So we know there is going to be something wrong with every model. In any basketball model that something wrong is going to be underating/overrating players. Its the nature of the beast that every model is going to be “wrong” about the same types of players.
I don’t think the system overrates rebounders I think it overrates easy shots. I think its impossible to attribute enough credit for easy shots to the right players. I will point out JJ Redick’s wp48 as one I don’t believe is sustainable unless he always plays with specific teammates.
The thing about these “overrated” rebounders is they also seem to play on underrated teams when they aren’t the only highly rated player on the team. Neither Murphy or Lee has ever played with another player rated high by wp48. The model thinks teams that are built like that suck. Now, the model is saying Golden State won’t suck anymore since they finally have 3 players the model finds promising. Lets test it. That trio(Curry, Biedrins, Lee) if somewhat healthy is worth at least 35 wins in my book. However, I don’t think 50 is possible with the bench guys some I don’t want to be lumped in with anyone predicting 50 wins when I think it’s got less than a 50% chance of happening.
Why don’t you have such a problem with Landry’s ratings over the year? Or Mutombo’s ratings over the year. Remember when Houston went on the 20 game win streak a few years ago after Yao Ming and Tracy McGrady got hurt? The Rockets were forced to playing all their highest wp48 players and they went on a tear.
How did the Pistons win with Ben Wallace? I’m pretty sure measurements like Wp48 were the only ones giving the Pistons a shot to match up with the Lakers.
Wp48 saw Kevin Garnett as a franchise changer way before I ever heard anyone else say this. Its funny because with revisionist history its like everyone else saw that he could be a necessary championship piece.
I can go on and on but I think that’s pointless. I have a grand time debating with people who actually give an f so maybe the both of us will keep banging our heads with our palms and keep this going for a while.
ohreally
July 18, 2010
marparker,
To answer your question: it’s not important to me what kind of player gets overrated.
No, wait. It is, in a way. But not the way you are thinking. What bugs me most about WP48 is that I can tell from the moment I read the definition of it that it is going to overrate certain players, and I can name some of those players in advance. And then when I check them, I find I was right. It bothers me to use a player metric that has such an obvious flaw.
I’m glad to know you weren’t serious about the correlation comment. If you were, I was going to point you at my new player metric, which I call UWP, short for “Ultimate Wins Produced”. For any player, UWP is defined by the teams record for whatever portion of the season they played with that team. You’ll find it correlates with wins 100%. ;-)
I completely agree with you on your comment about easy shots.
I just thought of two more players that I bet WP48 overrates: Corey Maggette and Jamal Crawford. Neither of them plays any defense and they both generate a lot of offense. I wonder what WP predicted for the Warriors when Baron left for LA and Maggette came to the Warriors?
Is there anyplace I can get WP and WP48 historical data for players? I can get Win Shares on basketball-reference, but my ability to point to WP48 examples has been hindered by not beig able to find the data.
I have an idea for a player metric that I’d like to try out. As a first cut I’d like to generate a list of the top 50 players by my metric, and by WP48, among big minute players for the last two season. Then we can eyeball the results and see what we think. But to do that I need the WP data.
dberri
July 18, 2010
ohreally,
It is amazing how many comments you have left here before
a. understanding the measures discussed here first (your comment on Jamal Crawford reveals you do not understand as much as you think)
b. reading the discussions we have had on plus-minus (and measures of this nature) before. As has been noted many times before in this forum, plus-minus is very inconsistent across time. Hence, it does not capture the productivity of individual players very well.
This is something I teach my students. Read first, write second. I think you could have saved yourself some trouble if you followed that advice.
ksalenet
July 18, 2010
Nice site and useful information
I wish you continued development
And I’ll be keen follower of your site wonderful
ohreally
July 18, 2010
dberri,
My comments on Crawford are off the cuff, without having looked at specific data. So it’s possible that they are off the mark.
If you think I am misunderstanding something, you can advance the discussion by telling me what it is you think I am misunderstanding.
As for plus minus, I’m not sure what your point is? I’m trying to point out a problem with WP. What does it matter if plus minus is inconsistent?
ohreally
July 18, 2010
dberri,
Do you think that WP is useful for making the kind of predictions that Jeremy is making in the article above? And if so, do you agree with his conclusions?
dberri
July 18, 2010
ohreally,
My point is we have already had this discussion. This site has existed since 2006. We have already discussed the role of defense. We have already noted that on-off stats, counterpart defense stats, etc… (i.e. stuff you see at 82games.com) are inconsistent and therefore not reliable in projecting future performance. In other words, Jeremy’s projections (which he notes makes a number of assumptions about playing time, etc…) are more reliable than the stuff you are noting. You simply do not know that the defensive stats you are citing for Lee are really all about Lee (or if those will translate to the Warriors, or that those really have that big of an impact on outcomes).
Furthermore, Stumbling on Wins addresses conistency across time, the value of defense, etc… And Stumbling on Wins (and other published work) address a number of objections raised to WP by various people (objections raised since 2006).
I think the problem here is that you simply haven’t read very much. And yet you have formed very strong opinions. You are free to leave whatever comments you like. But I think you would be offering more insightful observations if you just took some time to read something first.
ohreally
July 18, 2010
You simply do not know that the defensive stats you are citing for Lee are really all about Lee
We don’t even know that the offensive production included in WP is all about Lee either. What if a player gets easier offensive opportunities because of other players?
I’m still curious about your feelings on Jeremy’s predictions… other than telling me you prefer them being based on WP than something else.
ohreally
July 18, 2010
btw, consistency of a stat does not necessarily constitute an argument in it’s favor over a less consistent stat.
If I rate players by their height, that will be very consistent, but not very valuable.
David Lee can come to GS and put up the same kind of lines he has in NY. But if we don’t see a jump in wins for GS, where does that get us? His WP will be consistent, but it also will have been a failure at anticipating his impact on GS.
marparker
July 18, 2010
Every formula used to measure height is consistent. Not every formula that measures player production is consistent. Its all relative.
ohreally
July 20, 2010
Right. In my example, I was measuring performance by height. Not a very smart way to do it. But extremely consistent.
However, a ton of less consistent ways would actually have more predictive power.
Edmond
July 20, 2010
Right, but height hasn’t been shown to correlate strongly with team wins, whereas WP has.
The problem I have is that these “overrated” players aren’t as good as their stats indicate, then you have to come up with a story about how other players on their team “secretly” got those wins back for their team so that the numbers would all work out–Wilson Chandler plays great D, the shine off Danilo Gallinari’s forehead blinds the opposition etc. You could invent any number of more or less plausible accounts. Over the last 4 seasons Lee has produced between 15 and 17 wins a season while the roster around him has gone through some fairly significant changes.
Edmond
July 20, 2010
Rather: “The problem I have with claiming that…”
ohreally
July 20, 2010
The problem I have is that these “overrated” players aren’t as good as their stats indicate, then you have to come up with a story about how other players on their team “secretly” got those wins back for their team so that the numbers would all work out
We already know that WP does not align with wins perfectly. Therefor it’s likely that it overrates some players, and underrates others. How the other players got the wins back isn’t a secret. It’s simply the other side of the same coin. If it’s true that Lee is overrated from his defense, then there will be some combination of:
a) teammates being underrated from their defense
b) difference between real wins and projected wins.
{I’m simplifying, pretending there are no other problems with WP.}
WP is basically accounting for scoring opportunity and scoring efficiency. So the sum total of WP across a team is very likely to correlate strongly with real wins. But that doesn’t really tell us that the player that got the box score stat was the one who should truly be assigned the credit. There are systematic forces at work that distort the credit – it’s not always fully earned by the person who gets it on the box score.
Cliff Robinson had a long career, and one of his stops was as a PF/C for the 2003-04 Warriors. He was well known for not grabbing any rebounds.
Cliffy’s defensive rebounding percentage that year: 9.5%
Lee’s last season: 25.2%
Lee is a monster compared to Cliffy, right?
But I watched those games, and I observed that Cliff religiously boxed out the best big on the other squad, and let his guards come in and grab the ball while his opponent was neutralized. Meanwhile David Lee is a PF/C who never boxed out anyone for the Knicks.
team rebounding % differential for Cliffy:
+2.4% (from 67.4% to 69.8%)
team rebounding % differential for Lee:
+0.7% (from 69.9% to 70.6%)
Which one got their team more possessions from their rebounding?
And just because WP correlates with wins at a certain rate does not imply that it’s really “that right”. There are GMs and coaches out there who are actively trying to get the most out of their team. They do things like compensate for player weakness all the time. If there is a weak defensive player, they may try to play him more often with stronger defensive players.
To get a more meaningful correlation between WP and winning, we need to be able to try arbitrary mixes of players and see if the WP relationships still hold. Of course, in real life, that is not possible.
So we are left with the much more limited test of watching players change teams and checking the results WP predictions. That’s what is going on here.
I’m told I’m not well read and have strong opinions. Maybe that’s even true.
But I have watched and re-watched a lot of basketball, while studying various statistical measurements, and I have an open mind. I’m willing to put any theory I have to the test, and I will adapt my thinking as best I can based on the results.
And my current thinking is that you have to look at *team* measurements while a given player is on and off the floor to understand the value of a player to that team. Individual box score stat measurements will not get you that information.
jbrett
July 20, 2010
ohreally,
Everyone on the planet believes they have both a sense of humor and an open mind–but you may be the only person on Earth who knows, beyond a shadow of a doubt, that David Lee is not very good at playing basketball. I had a certain admiration of your dogged determination on the subject, in the face of contrary opinions on this blog, as well as countless other metrics; I was compelled to respect your convictions–until you claimed to have an open mind. That’s a good one.
I have nothing to contribute to the technical side of the debate; just wanted to thank you for the chance to demonstrate, for my wife, the classic comedy ‘spit take.’ I’ll go clean the coffee off the fridge now.
ohreally
July 21, 2010
jbrett,
Will you be around this coming April? There will be a final exam on open-mindedness at that time. Get your #2 pencil ready! ;-)
Evanz
September 9, 2010
As a 20-year fan of the Warriors, I can’t even explain how exciting it is to think about the possibility of having two 10+ rebound guys playing this season. Predictably, the guy who was mostly left out of this discussion, Dorell Wright figures to have a huge impact on the defensive side of the ball. My hope is that Reggie Williams supplants Ellis at some point, and that Ellis is traded for another big expiring contract or another defensive stopper who is underrated by traditional metrics (i.e. low ppg). There has been much talk in Warrior land about the possibility of trading for Melo, but at this point, it looks like we may be able to dodge thay overrated bullet, I mean Wizard, I mean Nugget. More than anything else that has taken place since this article was written, the changeover in ownership from Cohan to Lacob has got to be the best news. I’ve liked everything I’ve heard from Lacob so far. It’s not just that I think he wants to win, which is definitely important considering Cohan’s legacy. More than that, though, he appears to be a guy who actually has a good idea already of how a winning team should be built (given his experience as a minority owner of the Celtics). He has said that a winning team should have three main guys, and that Curry and Lee appear to be 2/3 of that equation. He definitely does not appear to be doe-eyed over Ellis (although he loves watching him play – diplomatic, eh?). That and other statements he has made lead me to believe that he understands and maybe is somewhat even familiar with the advanced metrics discussed on WoW.
This season will be a good test for WP48. If Lee doesn’t produce, his WP48 will be low and the team won’t win so many games. But that’s not a failure of WP. In my mind the only way that WP48 “fails” in the case of the Warriors is that the team does amass 50 or so WP-calculated wins, but, say, only manages 35-40 actual (real-world) wins. I think there’s a rather small chance of that happening, though. As discussed elsewhere, the thinga that WP values, like rebounds and efficiency, are exactly the things that Warriors fans know we need.
One more thought about test cases…I’ve heard, although not verified myself, that the Warriors actually had a much better winning percentage when Ellis didn’t play last season, than when he did. If true, this serves as a nice counter-example to the WP-naysayers, who might not believe guys like Morrow or CJ Watson could help a team win by getting more minutes.
reservoirgod
September 27, 2010
Keith Smart basically admitted at media day that Don Nelson didn’t do much of anything but show up last year. Smart said he had more freedom than any other asst. coach in the NBA – he ran training camp, practices, etc. I know Dave said he didn’t call coaches “deck chairs” last year when Abbott posted it on Truehoop, but Nelson really was useless last year for GSW.
ohreally
October 1, 2010
“If Lee doesn’t produce, his WP48 will be low and the team won’t win so many games. ”
No, that’s not it at all. Lee is going to produce, and he’s going to get his WP48. The question is “will the Warriors jump to 50 wins as a team?”, as per the WP-based predictions in this article.
If WP doesn’t work well for measuring Lee’s real impact, Lee will still get good WP numbers but the Warriors won’t win anything like 50 games.
Evanz
October 1, 2010
ohreally, it’s been shown over and over that WP correlates extremely well with wins. If Lee puts up his numbers and has a high WP, and say, Curry and Biedrins put up good numbers, they will win a lot of games. Take that to the bank.
Evanz
October 1, 2010
To follow up my last point, David Lee led the Knicks last season with 0.25 WP48, followed by a bunch of average to below average (<= 0.10 WP48). WP retrodicted 31 wins for the Knicks. How many did the Knicks actually win? 29. WP was only off by 2 wins. Why did the Knicks lose so many games? Not because of David Lee. His teammates were all average to below-average players. The numbers really do speak for themselves.
ohreally
October 9, 2010
“ohreally, it’s been shown over and over that WP correlates extremely well with wins.”
Of course it correlates well with wins. At it’s heart, it’s basically totalling up possessions and efficiency, and subtracting out that of the opponents. It’s going to correlate very well with wins, the same way scoring differential correlates well with wins.
Your whole example you provide has the same problem. I could just make a table of how much each player outscored his opponent player in the minutes he played. Because the sum of that across the team would be the scoring differential, and because the scoring differential over the season correlates very well with wins, my player ranking method would correlate with win very reliably. But it would be a pretty bad way to rank players, and that could be proven by watching what happened when a guy like Maggette gets traded.
The only way I know to test whether the credit gets divided up correctly is through trades. Lee is a WP star. Lee was traded. Lets watch what happens to the Warriors win total. Same for Murphy to the Nets. It’s a shame that the deck is stacked in WPs favor for these tests (Warriors were decimated by injuries last year, Nets were about the worst team possible last year), but you gotta take what you can get.
evanz
October 9, 2010
ohreally, that’s a nice strawman you’ve constructed (well, others have done it before you, of course), but it doesn’t actually invalidate WP.
ohreally
October 20, 2010
In that case nothing can invalidate WP. If you exclude making predictions about changes in team wins from any testing scenario, then WP is infallible.
And also useless to me as a tool that says anything interesting about basketball players. But hey, still infallible.
ohreally
December 14, 2010
the early returns not looking good for WP.
Randall
January 25, 2011
So what’s the deal? Why aren’t the Warriors good?
ohreally
January 27, 2011
Andris is healthier than last year. D. Wright exceeding all expectations. Monta more efficient. Curry doing his thing. And yet they still suck. hmmmm.
btw, turns out Warriors rebound worse when that premo rebounder, David Lee, is on the floor.
Jeremy Britton
January 30, 2011
It’s time for a follow up on this post, isn’t it? I’ve been following the team closely and noticed a few things:
1. David Lee has been playing the worst basketball of his career (he’s also been hurt)
2. Dorell Wright and Reggie Williams started off way below what we’d expect (though Wright has been picking it up again lately)
3. Andris Biedrins hasn’t returned to pre-injury form at all and has, in fact, tailed off a lot lately (he’s been hurt off and on too)
4. Stephen Curry got off to a slow start (ankle injury too) but is picking it up
5. Monta Ellis started off really strong around that 6-2 start, but his shooting has tailed off quite a bit since
ohreally
February 5, 2011
Let the excuses begin!
Dorrell Wright leagues the lead in made 3’s and has been playing better than most expected. Reggie’s reduced minutes are due to Wright’s success, so this is not a bad thing for the SF spot on the team.
Lee did have a blip downward from an injury, but is playing as well now as he did in NY. He’s just not getting as many touches because GS has more weapons than NY did.
Biedrins may not be as good as he was two years ago, but is better than he was last year, and has been on court a fair bit and helping the Warriors rebounding situation tremendously when there.
Curry has been doing fine.
Monta’s cold shooting streak is exactly 5 games long and otherwise his year has been his best ever, with all the scoring from before but at increased efficiency.
The simple fact is that we enough of the ingredients that were predicted. Lee just isn’t that good. He grabs boards, but does not actually help the team rebound. We rebound worse when he’s on the court than when he isn’t. And he can’t defend a chair.
On the offensive side he’s very good. His jumper has come back and he passes well. If he weren’t a fake rebounder and non-defender, the Warriors might have something.
Jeremy Britton
February 5, 2011
@ohreally, well, that’s just like your opinion, man. It’s obvious from your comment that you don’t think there’s anything to this Wins Produced analysis.
Lee already wasn’t going to be any good, so the Warriors losing just validates this. Monta scoring a lot of points in spectacular fashion proves he’s a great player. Same old, same old.
What I think is interesting is how the Wins Produced shows how fans and observers of basketball persistently undervalue possessions and overvalue individual player point totals in assigning credit or blame for team wins and losses. Sticking to that analysis, some more interesting themes emerge from the Warrior’s season so far. I’m looking forward to a followup post.
ohreally
February 6, 2011
Hey, don’t blame me. It’s WP that made the prediction. It’s WP that deserves the blame.
There are two options: admit that WP is testable in these situations and chalk this one up as a failure for WP; or simply deny that WP can be used as a predictive tool, and, by implication, confirm that it’s not a very good way to measure player performance.
Napoleon
March 26, 2011
Currently Golden State is 31-42. Looks like ohreally knew what he was talking about all along…
Jeremy Britton
March 27, 2011
@Napoleon, it’s fun to think like you and ohreally do. It’s like, “Gotcha WP! The proof is in the pudding and WP is wrong, see?”
I catch myself falling into that same intellectually lazy I-told-you-so trap from time to time. I’m working hard to get better at this. You cannot forget that thinking with data feels odd a lot of the time. Correlation does NOT equal causation. Or better, our first instincts are often bad judgment.
jaber
May 16, 2011
So what’s the deal? Why aren’t the Warriors good?
Bigmouth
June 16, 2011
Little late to this discussion, but I find it a little hilarious that ohreally predicts WP and WP48 would overrate Corey Maggette and Jamal Crawford. If that comment doesn’t suggest a lack of understanding the model, I don’t know what does.