The Detroit Pistons defeated the Orlando Magic in the 2008 Eastern Conference Semifinals. Since this victory, each team has gone in very different directions. The Magic advanced to the NBA Finals in 2009, and given the moves made this summer (moves I will discuss in a moment) should be even better in 2009-10.
Meanwhile, the Pistons have essentially blown up their roster. At this moment, only four players – Tayshaun Prince, Richard Hamilton, Rodney Stuckey, and Jason Maxiell – remain from the team that advanced to the 2008 Eastern Conference Finals. And the new players should hardly inspire confidence. With the signing of Chris Wilcox, the first and second string on the Pistons (yes, I know I said this a few days ago, but I am a Pistons fan) is now as follows (with 2008-09 Wins Produced and WP48 [Wins Produced per 48 minutes] reported for each veteran player):
First Team:
PG: Rodney Stuckey [4.0 Wins Produced, 0.077 WP48]
SG: Richard Hamilton [1.9 Wins Produced, 0.040 WP48]
SF: Tayshaun Prince [7.3 Wins Produced, 0.114 WP48]
PF: Charlie Villanueva [3.6 Wins Produced, 0.082 WP48]
C: Kwame Brown [2.4 Wins Produced, 0.117 WP48]
Second Team:
PG: Will Bynum [1.2 Wins Produced, 0.071 WP48]
SG: Ben Gordon [4.9 Wins Produced, 0.078 WP48]
SF: Austin Daye [Rookie]
PF: Jason Maxiell [2.9 Wins Produced, 0.098 WP48]
C: Chris Wilcox [0.2 Wins Produced, 0.008 WP48]
If you add up the 2008-09 Wins Produced by the veteran players all you get is 28.4. So unless Austin Daye is suddenly one of the top five players in the NBA – or these veterans become much more productive players — it looks like the Pistons are in trouble. At a minimum, this team is far removed from the Orlando Magic.
To see this point, here is what the Magic’s first and second team looks like today (again, Wins Produced and WP48 are reported):
First Team:
PG: Jameer Nelson [7.0 Wins Produced, 0.256 WP48]
SG: Vince Carter [9.3 Wins Produced, 0.152 WP48]
SF: Rashard Lewis [3.5 Wins Produced, 0.060 WP48]
PF: Brandon Bass [1.9 Wins Produced, 0.057 WP48]
C: Dwight Howard [24.6 Wins Produced, 0.418 WP48]
Second Team:
PG: Anthony Johnson [2.2 Wins Produced, 0.071 WP48]
SG: J.J. Redick [1.0 Wins Produced, 0.041 WP48]
SF: Mickael Pietrus [1.3 Wins Produced, 0.048 WP48]
PF: Ryan Anderson [1.2 Wins Produced, 0.045 WP48]
C: Marcin Gortat [5.2 Wins Produced, 0.316 WP48]
If you add up the Wins Produced you see 57.2, or an average of 5.7. This mark, though, understates the quality of the Magic.
First of all, Nelson only played 42 games last season. Had he played the entire season, and his per-minute performance didn’t change, he would have produced nearly seven more wins.
And then there is the case of Rashard Lewis. Lewis spent most of his career as a small forward, but was moved to power forward by the Magic. The move, though, didn’t dramatically change his production. Lewis posted a 9.8 Win Score per 48 [WS48] minutes with Seattle. With the Magic in 2007-08 his WS48 was 9.6 and last year it was 9.8. An average small forward posts a mark of 7.3, so was Lewis was above average in Seattle. An average power forward, though, has a 10.3 WS48. Consequently, Lewis was below average these past two seasons with Orlando.
The acquisition of both Bass and Anderson – coupled with the re-signing of Gortat – suggests Orlando is going to move Lewis back to small forward. Had he played small forward last year, Lewis would have posted a 0.177 WP48. If he can maintain that production next year – and the same holds true for Nelson, Carter, and Howard — then the Magic’s starting line-up next year will feature four players with a WP48 mark in excess of 0.150.
Now at this point, ESPN.com says the Magic’s roster only consists of the ten players listed above. Not knowing the rest of the roster limits are ability to speculate about next season. That being said, the Magic look very much improved.
In fact, I think the Magic are at least twenty games better than the Pistons, and that margin could grow to more than thirty. The Pistons could still make the playoffs next year. But a more likely scenario is a trip we haven’t seen since 2001. Yes, the Pistons could actually be visiting the lottery next year.
This means fans of the Pistons are becoming like fans of the Lions. In other words, before the season even starts, thoughts turn to the next draft. For those who can’t wait, DraftExpress has already posted a 2010 mock draft. And judging by 2008-09 numbers, Cole Aldrich might be someone to keep an eye upon. At least, his Wins Score per 40 minutes suggests he might be a productive NBA player (unlike many of the players currently employed by the Pistons).
– DJ
The WoW Journal Comments Policy
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
Jay
July 19, 2009
Blah blah blah, same old hat, ‘the pistons suck and I’m mad at the team’s decisions’. Why don’t you do everyone a favor and stick to important topics. We all know the pistons suck, you don’t have to keep telling us. And your city, Detroit, sucks even worse than the Pistons.
p.s. I did like your book though.
dberri
July 19, 2009
Jay,
My blog. I get to pick the topics. So from now on… Pistons all the time.
Seriously, I will move on to a new topic on the next post.
Rob O'Malley
July 19, 2009
This blog is his party, and as such I believe he has the right to “cry if he wants to.”
Ravenred
July 19, 2009
… and you would cry too if the Piston’s happened to youuuuuu…..
Jon Cruz
July 19, 2009
Professor Berri, I read an espn article about the ‘American Needle v. NFL’ case that is going before the SCOTUS in the next few months. It talked about how it could have far-reaching ramifications for all the major sports leagues, as the NFL wants to be treated as a single entity (not a group of 32 single entities) and thus be given immunity from antitrust scrutiny.
Since the first few chapters of your book examined the labor and business aspect of pro sports, any chance of a post on this and your take in terms of economic consequences, effects, etc.? Keep the good work up.
TRad
July 20, 2009
Dave
Suppose Odom stays in LA. Lakers or Magic?
Orlando looks better – Carter>Turkoglu, healthy Nelson, and maybe Howard will use summer to work on his post game.
Lakers look worse – Ariza>Artest (at least in the context of Lakers needs), Bryant and Fisher will be one year older, but maybe Bynum has no new knee to injure.
Last year you were propheting Lakers title just after they have lost to the Celtics. And now?
Tball
July 20, 2009
db was backing the Lakers because their anticipated win total was head and shoulders above the competition (I may recall a 70+ win total suggested). The Magic look great, but I suspect the Cavs and Celtics look comparable, which means we should have an exciting 3 team battle for home court in the East.
Also, I’ve seen a some commentators, including one at SI.com, suggest the 4th seed in the East will go to the Pistons. The problem with predictions from many pundits during the offseason is they spend their time focusing on the additions a team makes, ignoring subtractions and unmoved pieces. How could you think a team would trade out Ben Gordon and Chuck V for Wallace and McDyess (and Iverson, if you like that sort of thing, not that theirs anything wrong with that) and move from 40 win team to 55 win team. Joe Dumars is now accepting coaching resumes for the 2010-2011 season.
Italian Stallion
July 20, 2009
Listen all you Pistons fans. I’m a Knicks fan. You have no idea what “pain” feels like. :)
Joel W
July 20, 2009
Why will you never address the absurdity of your SF/PF distinction, especially WRT Rashard Lewis? This frustrates me almost as much as your unwillingness to do analysis in terms of value above replacement, rather than value above average.
That Rashard Lewis can play both power forward and small forward can only be considered good in terms of discussing his value to his team. That his team either doesn’t play him in the correct place, or has injuries which require him to play somewhere else, does not make him less valuable. While positional adjustments are obviously needed, this is clearly a place where WoW needs reform. Unsurpisingly, Rashard Lewis had great plus/minus numbers last year.
Joel W
July 20, 2009
(And I’ll note has had great plus/minus numbers throughout his career)
Westy
July 20, 2009
The interesting thing about Rashard Lewis is that whether he’s actually listed as the PF or the SF, I’d anticipate his actual role on the floor varies much less than when typical players move between these positions.
I can think of many PF who function way more like centers (always under the hoop, constantly rebounding), and others who are like Lewis that float out to the 3-pt line frequently (thus putting them out of position to rebound).
In essence, the positional categorizations would seem to break down somewhat with players like Lewis.
Nick
July 20, 2009
@ Joel W
I think you’re missing the point.
I see this happen a lot around here, the fact is WoW has to do with what the player contributes to wins. Not talent. Not skill. So if Rashard is forced to play PF, and he can’t rebound as good as the average PF, it doesn’t mean he’s not AS good as the other PFs. It means that given his role, in comparison to other PFs, he’s under average.
But as stated, if they play him at SF, and find a better role for his skill set, he’s above average. Not everyone is good at everything, and thus this should be, and is reflected in WoW.
Italian Stallion
July 20, 2009
Joel,
I agree that versatility is a plus even if a player is better at one position than another.
I think this is part of the same issue that leads to some of the other conflicts on this blog. There is a difference between measuring how good a player is and what he is contributing to WINs.
This system measures contribution to WINs.
That’s a very valuable piece of information.
However, most fans think in terms of how good a player is or which of two players is better.
Fans think of Lewis as an above average player whose skill set is most similar to that of the typical SF, but who has enough skills to play some PF when required. That makes him superior to an equally above average SF that could not play any PF at all without it being a total disaster.
The fact that one player out of the two might have played a lot more PF and not produced a lot of wins etc… is irrelevant to which is better, but significant to determining whether or not he contributed as many wins as the average PF.
IMO, the conclusion that should be reached is that Lewis is an above average player with some degree of versatility, but not good a enough PF to be kept at that position full time. He should be used at SF and the team should try to get an average or above average PF.
I think that’s what Berri is saying.
It’s just frustating for fans to look at the terrible rating Lewis’s earned last year because he played PF when we all know he’s an above average player.
ilikeflowers
July 20, 2009
Don’t have a heart attack Joel. Marginal analysis and players who are (supposedly) poorly defined by the traditional positions is just one of many areas that could use more research. Maybe it’s just not high on the list.
Peter
July 20, 2009
Essentially, the Magic’s biggest acquisition, if we look at all the moves as one continuous whole, was “trading” Hedo Turkoglu for Vince Carter, while making Bass the power forward in place of Lewis. Last year, I think everybody can agree that, while Lewis was listed as a power forward, he played more like a second small forward, like he did in Seattle.
Thus, if we treat Lewis’s PF numbers like SF numbers, then the moves the Magic made might not seem as big a leap as, say, Shaq to Cleveland.
However, the Magic were already very good as it was. And, in the end, the most important improvement it could make is NOT moving Lewis to the true 3 spot, but having Nelson healthy enough to play a full season. Hopefully, all those moves added together will go a long way towards helping propel Orlando to a 60-win season. Having Dwight Howard a year older and with another year of tutelage under Patrick Ewing shouldn’t hurt, either.
Peter
July 20, 2009
By the way, I know using the “second SF” idea in Berri’s model would inflate the Magic’s win total.
Lewis is going to be the exact same player. Win Scores or not, it’s going to be what happens around him that matters.
Rob O'Malley
July 20, 2009
Cavs also offered Jamario Moon a contract. Cav’s might still be on top with that.
DSMok1
July 20, 2009
Berri, have you looked at using a position continuum? For instance, Lebron played 75% of the time at SF and 25% of the time at PF…. is there any good way to adjust for this in the statistics? I did a similar adjustment for NCAA players when evaluating their PAWSmin–I did not use specific positions, but rather a continuum of “positions”.
mrparker
July 20, 2009
IMO,
The cavs still need another .2wp48 full time player to become a true title contender. Regardless, I believe that with a healthy KG Boston is head and shoulders above every one else including the Lakers.
Hopefully Jameer Nelson can make a full recovery. Without Jameer or KG healthy I fear that the Lakers will hoist it again. That is a nightmare scenario for anyone who hates hearing someone say that Kobe Bryant is the best player in the world
JChan
July 20, 2009
Hey DJ,
Are you checking your work email over the summer? I’ve noticed a lot of people asking for a place to look up WP48 numbers. Since I’ll probably never find time to finish the revamp of my site, I found a place that lets you upload data and share it with others. Seems to work pretty well.
I’ve got the season by season data up through ’07, but if you sent me the last two years’ worth, then I could put the whole thing online. Won’t be the perfect setup, but should make a bunch of people happy. Let me know what you think.
Palamida
July 20, 2009
DSMok1, The solution is simple: just use the positional numbers in a weighted fashion.
Use a relative combination of the multiple positions played in accordance with the playing time at each position.
I think this is self explanatory and i’m in a bit of a rush, but if you require an example, I’ll be more than happy to provide one.
Have a nice day.
dberri
July 20, 2009
JChan,
Oops, I did see your e-mail. But I am getting slower and slower in my response times.
About position adjustment… these are all weighted for time at each position.
Sam
July 20, 2009
From what I’ve heard, the Magic plan to continue using Rashard at the 4, while bringing Bass and Gortat off the bench.
Peter
July 20, 2009
If that’s true, I want to see what the team plans on doing with Mickael Pietrus.
JAW
July 20, 2009
My point is that Rashard Lewis contributes to wins in a highly positive manner regardless of how he gets labeled in the system. http://basketballvalue.com/topplayers.php?year=2008-2009&mode=summary&sortnumber=94&sortorder=DESC
His adjusted plus/minus is incredibly good. Isn’t it possible that his versatility, the exact value which gets lost by WoW is the reason his plus minus does not match up with his WoW score?
Look, I love WoW, I think it provides me with a lot more information that I would have otherwise. I understand the regression analysis, and why DB is so committed to it. I just think this application is flawed. It reminds me somewhat of an article I read with respect to the “runs above position” stat in baseball being flawed. It’s obviously important to make statistical adjustments for positions–players at the corners in baseball hit the ball much more, but mostly the value should be calculated independently and then adjusted. Here the value is adjusted and then calculated.
Look at those +/- numbers again. The top players basically conform to expectations, esp. over two years. I think Bill James once said that any statistic which adds value will confirm your expectations 80% of the time and surprise you 20% of the time. That’s how I feel looking at that list. The name that sticks out is Lewis’s, and I think it would be wise of Dave Berri to think about what that says about his positional adjustment system. Versatile big men stand out on the surprising list of the top +/- players.
JAW
July 20, 2009
I’ll illustrate my point with an absurd thought experiment. Suppose a team plays 5 guys who are 6’7 that all put up identical stats. The team shoots well, and plays good team defense, especially because they all rotate well and can guard anybody on the floor, which allows it to win 55 games a season. At the end of the season, we analyze this team. Again, each player shoots the exact same percentage from the field, same rebound rate, same turnover rate, same everything. Now, we have to assign them positions. One gets named C, another SG, another PF. How on earth does it make sense to say that there values are different based on these assignments? Not their “talent” as I was critiqued above for saying, but their actual value to the team. It shouldn’t be different, but it will be significantly different under this system. I know it’s a reductio ad absurdum, but I want to know how it’s logical for that to happen.
mrparker
July 21, 2009
JAW,
I understand the position adjustment and I think of it as something independent of the individual’s player’s talent. I look at Rashard Lewis’ poor performance as a failure of the Orlando organization. I can’t wait to see what he would add to that team as a small forward.
Along the same lines, I prefer to look at players as more than one player. For instance there is Shawn Marion PF, and Shawn Marion SF. One of those guys is much more productive than the other.
Oren
July 21, 2009
“One gets named C, another SG, another PF. How on earth does it make sense to say that their values are different based on these assignments?”
According to your example, one would presume that each player would play all five positions evenly. If so, if there was a position adjustment, it would the same for every player. As such, their values would in fact be the same.
However, if you had one arbitary player ONLY playing Center and one arbitary player ONLY playing PF etc… then if your 6″7′ player is effective compared to the average Center it shows that you should keep your team the way it is. In other words, it shows that you have a good scheme.
However, if your 6″7′ player is ineffective compared to the average Center, then this is clearly problematic. This would show that your team would be better off with four 6″7′ players and a 7″1′ Center.
According to your example, it wouldn’t matter which player you trade, but a model judges players on what they actually do and not what they can potentially accomplish.
A player in the NBA is probably better then a player in college. That stated, a weaker player in college may be more effective then a better player in the NBA because the NBA has better talent then college. In order for the player in college to be good, he only has to be better then the average college player. In order for the player in the NBA to be good, he needs to be better then the average NBA player. In other words, the standards are different at each level.
Likewise, there is one standard for being good as a NBA SF and another for being good as an NBA C. In other words, the standards are different at each position.
ilikeflowers
July 21, 2009
JAW, in your thought experiment why would you arbitrarily assign positions to the players? They all play all of the positions, thus no position adjustment is needed and they will have identical ratings.
I don’t know what the fuss is about Rashard Lewis, wow thinks that he’s a great small forward and a serviceable power forward. If you think that he doesn’t actually play the power forward role for the Magic and that his role is closer to that of a small forward then just use the wow SF measure for him.
ilikeflowers
July 21, 2009
…as Oren has already noted, I see too late.
brgulker
July 21, 2009
Thanks for ruining my day, Dr. Berri! :)
I’m hoping against hope that your model is somehow wrong or flawed — because that’s what we fans do, hope!
But, my more rational, less emotive side tells me that you’re probably right … we need to brace ourselves for the next iteration of the teal years …
Ryan
July 21, 2009
I will say Orlando wins 56-62 games next season ad Detroit (as team basically stands right now) wins 39-45 games- Essentially a 10-20 game wiggle room. I have to believe based on the scoring punch Detroit will bring to the table .500 is almost a given.
Joel W
July 21, 2009
The problem with your responses about even playing time is that “well they’ll each play each position equally” isn’t how things work in the scoring of the league. One player gets marked at each position.
The issue isn’t Rashard Lewis, it’s an issue with the system and Lewis is a good example of the issue. By plus/minus, he clearly contributes to wins as though the positional adjustment should be for SF, but Berri insists that he hasn’t been good, because he’s putting him in as a PF.
The reason there are basic positions is because basketball generally has a division of labor among its players. Now, what happens when a team changes how to divide that labor from the traditional 1/2/3/4/5 divisions? Perhaps, the team has one guy, oh I don’t know, play the wing on offense, and shoot a lot of threes like a traditional 3, but on defense he’s big enough to guard the 4? So their offense has one big guy and four outside guys? Should we compare this person who plays the 3 on offense and the 4 on defense to a SF or a PF? Aren’t their offensive rebound numbers going to look more like a 3’s, whereas their defensive rebound numbers are going to look more like a 4’s? And their assist rate? and personal foul rate? etc. etc.
The issue is simply that there is really good data that says Rashard Lewis was a very productive basketball player last year. WoW would agree if we made him a SF, but not a PF. I don’t care about Rashard Lewis though, I care about better statistical evaluation of basketball, and I think those contradictions are a big problem with one of the better systems for value-evaluation.
dberri
July 21, 2009
Joel,
You appear to like adjusted plus-minus because it fits what you believe. I wondered if you ever looked past the evaluation of Rashard Lewis on the Orlando Magic. Because looking at adjusted plus-minus it appears that Dwight Howard was only slightly better than Tony Battie last year. And Marcin Gortat, who just got a significant raise (motivated by the team that originally started the adjusted plus-minus approach) was the least productive player on the Magic last year.
Now I would argue that you can’t evaluate a model based on whether it fits popular perception. But this was your approach. When you look over the rest of the Magic (and I am not even looking at other teams) do you still want to maintain that position?
DSMok1
July 21, 2009
Adjusted +/- is a remarkably noisy metric by which to evaluate anything. There is simply too much coupling in the player interactions to make heads or tails of anything precise. Primarily, this is because of the limited group of players that are on the same team–if a player played a few games on each team, it would be (relatively) easy to quantify his true +/-. However, with the limited group of players on a team… it gets really hard to extract much meaningful data. Just too much noise. That said, one can probably glean a little from that–for instance, which players should be looked at further (via scouting, etc) and perhaps considered a bit more highly than the Wins Produced would say. (Due to the Bruce Bowen effect–significantly reducing opponent FG%, which can not be measured with WP, or any common stat.)
Peter
July 21, 2009
By the way, one interesting trend, is that Lewis’ effective shooting percentage has actually increased while in Orlando. However, his rebounding average from 2000-01 to 2006-07 (when his numbers went up in Seattle) has dropped from 6.3 all the way down to 5.6.
Regardless of how he should be classified in the model, his numbers in Orlando have changed for a reason. A 6’11”, cape-wearing, defensive player of the year reason.
While I’m at it, I should also mention that his assists have spiked from 1.9 to 2.5 in that same period.
Peter
July 21, 2009
Okay, so saying it’s because of Howard is overly simplistic. But I still felt it was important to mention.
Oren
July 21, 2009
“One player gets marked at each position. ”
I don’t think that’s true. I believe Berri has stated that he got his numbers from 82games. They mark players at multiple positions.
“Now, what happens when a team changes how to divide that labor from the traditional 1/2/3/4/5 divisions?”
We see whether doing so is effective. If a player is effective at the 3 but not at the 4, then this would indicate an error in the scheme.
“Should we compare this person who plays the 3 on offense and the 4 on defense to a SF or a PF? Aren’t their offensive rebound numbers going to look more like a 3’s, whereas their defensive rebound numbers are going to look more like a 4’s?”
That’s an interesting question. I’m not sure. I understood Berri’s argument as stating that it wouldn’t significantly change his numbers whether he played at the three or the four or if he played with Dwight Howard or with Mikki Moore.
However, I disagree with the standard that you use. If he plays instead of a traditional PF, then the standard which you’d compare him to is other PFs. If he plays instead of a traditional SF, then you’d compare him to other SFs.
Why do you insist on comparing him to the average player when you can make a much better comparison and compare him to a PF or SF?
“WoW would agree if we made him a SF, but not a PF.”
Rashard Lewis isn’t the best player in the NBA. But if you compare him to College Players… well he’ll be fantastic. And that’s at both SF and PF(and probably SG and C as well). So, if we just made him a college player, he’d be the best in the nation!
But he’s not playing in College. Ultimately, this year he played at PF and not SF. It’s almost like saying that a coach should let Mike Miller play Power Forward because he is a good Shooting Guard. After all, how could you bench a “PF” that could score 15 to 20 pts a game? WoW would agree that if we made him a SG or SF, he’d put up excellent numbers.
Joel W
July 21, 2009
Dave Berri,
My perception of Rashard Lewis isn’t that he’s good at all. I’m predisposed against him actually. I really have no idea where you got that I took the approach of trying to confirm my expectations, and it’s rather annoying to be accused of that. All I was attempting to do was analyze an issue I had with the WoW model: that positional adjustments make a player go from valuable to not valuable under its constraints, given the same numbers. So then I said, “Well, let me look to other statistical analysis that I think are valuable to see if they agree that Lewis hasn’t been valuable, and I’ll reassess my thinking.” Then, to my surprise, because I think WoW is really good, Lewis had great +/- numbers. Shockingly good ones in fact. So I thought, “Ok, maybe I’m on to something with the positional adjustment thing,” and thought I’d put forward an argument about it. How this turned into me trying to confirm expectations I do not know.
Both methods basically do confirm expectations most of the time, and surprise me in others, which I think is a relatively good indication of a valuable metric. Obviously the best indication of the value of a model is how well it predicts wins going into the future, but I have never seen a comprehensive study comparing the predictive value of WoW starting at the beginning of a season versus using adjusted plus/minus numbers. Adjusted +/- is obviously hard to do because of the limited data, which is why they provide standard error on that website.
I look at adjusted +/- because I think despite the very high usefulness of a regression based analysis, WoW has three problems:
1) Positional adjustments.
2) Value above average versus value above replacement (a below average player who is an above average bench player is a valuable player to a good team, and still provides value to a bad team).
3) The model explicitly does not capture individual defense. It adjusts on a team level, which captures a decent amount of the value. Steals, rebounds, and blocks capture another good piece of the value, and obviously correlate with defensive skills that are not captured without play-by-play data, so the R^2 is very high. And yet, I want to see if we can figure out the residual in the regression analysis better: what sorts of players do not have their value captured by WoW correctly. That’s not looking to confirm my biases, it’s looking to get deeper into the analysis.
In baseball, you can create a really good regression based analysis to go back and figure out, with fairly high accuracy, a player’s defensive value over the length of his career. However, UZR and other play-by-play metrics, while still flawed, are better. The regression based metrics and the play-by-play based ones have a high correlation, suggesting that they both capture defensive value pretty well. Still, I find it important to look where they disagree.
Oren,
The issue I have with the positional adjustments is essentially this: In baseball, you have defined positions. In football, you also have defined positions. It’s really easy to say what position somebody is playing. In hockey, you have goalies, and relatively defined positions. In basketball, that’s just not the case. Teams can go big, they can go small. They can switch up who is guarding who constantly.
When somebody plays shortstop in baseball, I know that they stand between the 3rd baseman and 2nd base. What does it mean to play the 3 in the NBA? Is Dirk a Center or a PF? Tim Duncan? The concepts are much more fluid. This isn’t Dave Berri’s fault, it’s just a really difficult question: In a game where positional adjustments are necessary and inherently difficult, what does one do? Well, I don’t know really, but I’m not willing to say a player’s actual value changes because of how we label their position.
Boney
July 21, 2009
hey dberri
Get in your time machine and work your little magic potion on the gathering of assholes that eventually won a ring in 2004.
Pull up the stats for:
Chauncey
Rip
Tay
Ben
You can even plug in any of the many PFs that played as well.. and then do the bench.
Tell me if that team has 50 wins.
brgulker
July 21, 2009
Lol, they do, Boney, (or at least approximately) because of Ben Wallace and CB mostly. Sheed helped too back then.
ilikeflowers
July 21, 2009
Now we know why they call him Boney.
ilikeflowers
July 21, 2009
Joel, so it’s not just that you find the SF/PF distinction absurd but you also find the entire idea of positions in basketball absurd? I think that you are the one far removed from most basketball observers on this point. I imagine that for most people to take your position seriously you would need to provide evidence not just opinion. Furthermore, I think that the predictive power of adjusted vs unadjusted wp48 was examined some years ago, but I may mistaken.
Alex
July 21, 2009
Hi Prof. Berri – I saw that the Magic picked up Matt Barnes today so I was looking up his stats in the archives and it looks like there’s an inconsistency. The 06-07 page (http://www.wagesofwins.com/GS0607.html) says he played about 1800 minutes and had a WP48 of .159, but the 07-08 page (http://www.wagesofwins.com/Warriors0708.html) says that year he played about 1400 minutes and had a WP48 of .047. Which one should it be?
Joel W
July 21, 2009
No that mistates my point, to say that I find the idea of positions in basketball absurd.
My point is that positions in basketball aren’t easily defined like they are in other sports. We observe coaches doing things like “go big” and “go small” which means, essentially “players that are meant for position X go to position Y.”
My point is that positional adjustment is just more difficult in basketball, and we should accept that.
Anon
July 21, 2009
going small means that you’re going to put a smaller player in the place where traditionally someone bigger would play, hoping for some sort of matchup advantage. But if playing these players at a position other than their natural one hurts their productivity then that’s what happens. I thought the example of Mike Miller at PF was a good one. Take a more extreme example like Muggsy Bogues playing center. He’d have a huge quickness advantage and would get a ton of steals probably, and how would anyone guard him? But at the same time having him as your center is going to hurt your teams rebounding, so ultimately its probably not productive to the team. Similarly, if Shaq played PG he’d probably average 20 turnovers a game.
I don’t see why it’s not believable that Rashard Lewis is a good SF whose skills aren’t those of a PF, so that when he plays PF he’s not as valuable to the team.
It reflects well on him as a player that he is capable of playing two positions, but ultimately if you want to get the most value out of him, you should play him as a SF. It doesn’t matter if his stats are identical at the two positions if an average PF puts up better stats.
And as I understand it the model doesn’t really measure stuff as value above average. It has an amount of production over which the person is a positive contributor to the team, and then another amount of production which is the league average. Below average players still provide value. Just, obviously, not as much as an above average player would.
That’s why Iverson would be a good pickup for a number of teams even though his stats would indicate he’s slightly below average as a player. Slightly below average is still better than way below average, and his contributions to a team are positive overall. Only extremely bad players contribute negatively.
Oren
July 22, 2009
“My point is that positional adjustment is just more difficult in basketball, and we should accept that.”
I agree that positional adjustment is much more difficult to define in baseball then in basketball.
That stated, in a large majority of cases, I suspect you can tell the difference. Just looking at a list of players(very unscientific), I’d say at least that 2/3s of players can be listed in one specific position.
“Well, I don’t know really, but I’m not willing to say a player’s actual value changes because of how we label their position.”
I suspect that the only argument you could make is that players whose position we can’t label shouldn’t be used for the position average.
But it’s the always the case that a players actual value changes not necessarily based on his own value but on the value of the players around him. Even if a players performance stays the same, if the players around him get better, then he becomes worth less.
todd2
July 22, 2009
Losing Turkoglu and gaining Carter is more or less a wash. Carter shot a slightly higher AFG% but took more shots. The Magic need to improve their rebounding. IF Gortat can develop a face-up game and play with Howard, the Magic will be a monster. Time will tell.
todd2
July 22, 2009
As far as SF/PF and Rashard Lewis goes, he doesn’t have a PF’s game. Pull up his shot charts—he’s a perimeter guy. 3.4 FTA’s per game are low for a PF. If he does take shots near the lane they’re turnarounds and fadeaways. He can’t defend PF’s either, too light. SF on both ends of the court.
Italian Stallion
July 22, 2009
I’d like to throw a monkey wrench into the positional adjustment debate. Currently, I think it’s a required adjustmen. But I think it “may” also expose a flaw in EVERYONE’s thinking.
Let me explain.
According to most models Cs and PFs tend to be the most productive players in basketball. Let’s ask ourselves a question about that that has already been asked numerous times. But this time, let’s ask it for a different reason.
What would happen if we created a team of all productive Cs and PFs?
Most people cocede that the team would be flawed and not do very well because it would probably lack some ball handling, passing, outside shooting, and other skills in sufficient quantities.
Well, if you can’t substitute more productive PFs and C’s for less prodcutive PGs, SGs and SFs and improve a team, doesn’t that suggest that some of the skills of these smaller players are being UNDERVALUED.
We correct via the backdoor of positional adjustments, but that opens up all the definitional problem. Is a player really a SF or PF, a PF or C, what if he plays both, is versatility a positive etc…
Wouldn’t it be better if we actually identifed what these smaller players are bringing to the table that makes the team better properly so that regardless of which position a player plays and how often he plays it, he gets credit for it.
Let’s face it, there are some big men that don’t rebound, block shots, and score in the paint at super efficient rates, but they can pass, shoot from the outside as well as some of the better small men, space the floor, and even have the offense run through them at times. Guys like that will often rate poorly on most models, but they may be adding a lot of value that is not being captured properly just as the value of small men typcially needs to be adjusted relative to typical PF or C. They have a lot of the same skills.
Art
July 22, 2009
The Pistons = 28.4 wins sound relatively accurate estimate. After seeing Dumars turns Billups (one of the best all-around PGs) and R.W. and McDyess (two of the best all-around PFs) into B Gordon (a defenseless, selfish and one-dimensional small ballhog) and C.V. (can’t comment much on him because I didn’t watch much the Bucks), my immediately thought is the Pistons will have a hard time to win 30 games next year. Yes, they become younger, but it doesn’t matter unless the younger players are better or have a lot of upside than the older ones. Billups in the next year (and beyond) is still much better than BG because Billups makes everybody better, whereas BG makes everybody worse and he has no upside.
I wouldn’t be surprised that the Pistons will be the bottom 3 teams in the East next year. Only the Knicks have a worse assembled roster than the Pistons (at least in the Knicks case, it’s not their current GM’s fault, and they will have cap space next year). The Nets are too young to going anywhere (at least the Nets have a lot of young players with good upsides).
I’m just so glad the Bulls finally get rid of BG (who hurt the team winning more often than helped), of course with a lot of help from Dumars.
Italian Stallion
July 22, 2009
“Well, if you can’t substitute more productive PFs and C’s for less productive PGs, SGs and SFs and improve a team, doesn’t that suggest that some of the skills of these smaller players are being UNDERVALUED”.
By the way, I realize that the reverse is also true. You can’t create a team of all SFs and SGs and expect a team to be successful either.
Remarkably though, small ball seems to work quite well at times and has probably even grown in popularity in recent years.
Kind of strange that you could create a team of supposedly LESS PRODUCTIVE players that lack some of the key important skills of most big men and still be successful.
To me, this also suggests that some of the skills of smaller players are not captured properly and they are actually more productive than people think- just in different ways that sometimes aren’t in the boxscore.
Oren
July 23, 2009
“Well, if you can’t substitute more productive PFs and C’s for less prodcutive PGs, SGs and SFs and improve a team, doesn’t that suggest that some of the skills of these smaller players are being UNDERVALUED. ”
No. It suggests that if you don’t play players at their proper positions that they’ll suck. I don’t think that Shaq will get very good numbers if he plays PG.
A model tells you what happens. It doesn’t tell you why.
Italian Stallion
July 23, 2009
“No. It suggests that if you don’t play players at their proper positions that they’ll suck. I don’t think that Shaq will get very good numbers if he plays PG.”
I agree . But the key reason is that players at different postions tend to have different but “still valuable” skill sets.
So the question is:
“Are PFs and Cs actually more productive on average the way most models suggest or are the smaller men equally productive but doing things that are not being captured properly in the boxscore?”
I am simply pointing out that when teams try to substitute supposedly more productive Cs and PFs for SFs and SGs it doesn’t seem to work, but when they substitiute supposedly less productive Gs and SFs for Cs and PFs, sometimes it does. That’s why small ball has become more popular in recent years.
So intuitively to me that suggests that the smaller guys are more valuable than is being captured by the boxscore.
I have no statistical proof of that, but I am wondering if the positional adjustments are getting to the right answer for the wrong reason.
If so, then rewarding the various skills properly would eliminate these controversies about SFs rating poorly when they play PF etc…
Skills would be skills, end of story. The actual position played would mean nothing.
A coach/GM would still have to build a team with diverse skills. But it wouldn’t matter if they had a 7 foot guy with an excellent 3 point shot shot, good ball handlingand passing skills etc… that wasn’t as efficient a scorer or rebounder as the typical PF /C . He would still rate highly. The team might simply have to find a SF that was an above average rebounder and highly efficient scorer around the basket to get the proper balance of skills needed to be successful.
Get it?
Both have required skills, but the roles would reversed from the more traditional ones and the ratings would reflect the actual value of what the players were contributing regardless of the position they played.
The key question is whether PFs and Cs are actually on average more productive than the SF and Gs or whether there is a bias in the model because the boxscore is not capturing some of the value of smaller players that typically shoot from the outside more often, space the floor, pass better, handle the ball better etc…
Westy
July 23, 2009
For what it’s worth, I agree with you, IS.
The problem is, I’m not sure these other ‘skills’ are easily measured in a box score.
Oren
July 23, 2009
“I am simply pointing out that when teams try to substitute supposedly more productive Cs and PFs for SFs and SGs it doesn’t seem to work, but when they substitiute supposedly less productive Gs and SFs for Cs and PFs, sometimes it does. That’s why small ball has become more popular in recent years. ”
Let’s suppose that argument is true. I’m not sure it proves anything other then playing a PG at C is a better idea then playing a C at PG although neither are good ideas.
“So intuitively to me that suggests that the smaller guys are more valuable than is being captured by the boxscore. ”
So, you’re arguing that if a play a PG at C, he’s going to have really bad stats while his teammates stats will remain constant. But if you play a C at PG, his stats may remain constant but his teammates’ stats are going to become much worse.
“The key question is whether PFs and Cs are actually on average more productive than the SF and Gs or whether there is a bias in the model because the boxscore is not capturing some of the value of smaller players that typically shoot from the outside more often, space the floor, pass better, handle the ball better etc…”
If this was true and there was a bias, then the way to correct this bias would be by solely comparing PFs to PFs. Each PF is judged on the same scale, so any unfair weighting would be spread out evenly.
Italian Stallion
July 24, 2009
Oren,
I tend to agree that currently the best way to evaluate players is to compare them to other players at the same position.
However, I do it differently.
I look at the player’s individual skill set and try to categorize him that way regardless of which position he is actually playing on the court.
So when I evaluate a guy that is a really productive SF, I don’t penalize him for playing some PF during the year. I consider him a SF. In fact, if he was average or so at PF, I will actually reward him for versatility.
If I see a 7 footer that plays offense like a SF and his stats stack up favorably to other SFs, that’s fine by me. I don’t care if he’s classified as a C or PF by others. I compare him to other SFs. It’s up to team to try to get the other skills on the court with him.
I’m not saying this way of doing things isn’t flawed. It is very flawed. But I think it’s less flawed than totally miscategorizing players and making them look bad when they are actually good.
I realize that the model here is not trying to say which player is better than which. It’s just trying to calculate Wins Produced. However, I am equally interested in evaluating player’s skills and ability. For that, I think I need to think a little independently. I hope that makes some sense.
David Michaels
August 11, 2009
Detroit isn’t that bad….
Bill Bartmann
September 3, 2009
Excellent site, keep up the good work