Allen Iverson and the Wages of Wins
The Wages of Wins Journal (WoW) is a popular sports statistics site based on a book titled Wages of Wins; WoW is frequently cited by mainstream media sources, including best selling writer Malcolm Gladwell and Henry Abbott's popular True Hoop blog (Abbott has also provided links to several of my articles, both from 20 Second Timeout and other publications). While I disagree with the methodologies and conclusions of many "stat gurus," I particularly find fault with WoW, starting with the basic premise of WoW creator Dave Berri that not only can basketball (and other sports) be understood without even watching the games but that watching the games biases the viewer because he is more likely to be influenced by a few spectacular-looking plays at the cost of losing sight of the larger picture. Competent NBA executives, coaches and scouts don't watch games in the way that Berri describes and it is insulting to imply otherwise; I believe that to understand NBA basketball you first must watch games with understanding and only then can you utilize certain statistical tools to add nuance and detail to what you observed--but the numbers can never take the place of watching games with understanding.Berri's statistical system is based on calculating "wins produced"; he claims to be able to determine exactly how many of a team's wins each player "produced." This approach has led to some "interesting" conclusions: Berri applied his metric to the 1995-96 Chicago Bulls team that won a record 72 regular season games en route to capturing the NBA championship and he concluded (page 144 of Wages of Wins), "Per 48 minutes played, Rodman's productivity even eclipsed Jordan. Rodman's WP48 of 0.415 was four times the production offered by an average player in the NBA and even surpassed the 0.386 WP48 posted by Jordan."
Dennis Rodman was a great player, a Hall of Fame caliber player (though he will likely not be inducted for reasons that have nothing to do with his basketball accomplishments)--but to suggest that on a per minute basis Rodman was more productive than Jordan is silly. Why does WoW make such a silly contention? WoW places a very high value on rebounding, a lower value on scoring and completely disregards the value of being able to create a shot (for oneself or for one's teammates); this also explains how Berri could assert--using the same metric that crowned Rodman over Jordan--that early in the 2007-08 season Andrew Bynum was more productive on a per minute basis than Kobe Bryant. The reality, then and now, is that Andrew Bynum is a young big man who has a lot of talent but he is still developing into someone who can be a productive player on a consistent basis; there are very good reasons why L.A. Coach Phil Jackson has tried to lower the public's expectations about Bynum and why Jackson has often benched Bynum down the stretch in games. Of course, Berri would likely assert that he simply understands the real value of NBA players better than Jackson and other NBA decision makers do; don't laugh, because that is not a sarcastic comment: Berri truly believes that by NOT watching NBA games he can more completely understand them than NBA lifers do, which makes almost as little sense as saying that by not working out you can attain the same level of physical fitness of a world class athlete who works out regularly.
Allen Iverson is one of WoW's favorite targets. On page 136 of Wages of Wins, Berri wrote, "At the end of the day, by some numbers Iverson is truly great. By other numbers, though, he is very far below the average player. When you summarize the great and the not-so-great into one metric, the net value of Iverson during his career is a bit below the average NBA player" (p. 136). Iverson's style of play and off court dramas are not everyone's cup of tea but to assert that overall he is "a bit below the average NBA player" is delusional.
Naturally, it would not be very good for the credibility of WoW if Iverson does well individually or if he leads his team to a successful season. Therefore, WoW frequently contains articles "proving" that Iverson is responsible for many of his team's losses and very few of their wins. I've done posts about the flaws of WoW (including one titled The Counterfeit Currency of David Berri's Wages of Wins) but the system is so obviously misguided that I eventually lost interest in even discussing it; however, a recent WoW post about Allen Iverson is so tendentious that I not only rebutted it but I did so at their site:
Really the Answer is Iverson
In the above post, WoW asserts that Iverson is the main reason that the Detroit Pistons struggled a bit after trading Chauncey Billups to Denver in exchange for Iverson and Antonio McDyess. As I noted in my comment at WoW, McDyess is not a mere throw-in with this deal; he was Detroit's leading rebounder last season by a wide margin and without his services the Pistons plummeted from near the top of the league in rebounding differential to near the bottom. The loss of McDyess' paint presence is a major reason that the Pistons did not win at their usual rate, not anything that Iverson did. By NBA rule, the Pistons could not re-sign McDyess for a month but now he is back with the team and Detroit is doing quite well, riding a four game winning streak that includes a victory last night over Orlando, ending the Magic's seven game winning streak.
It is horribly biased and just downright sloppy to purportedly analyze the impact that a trade has had on a team without even mentioning one of the key components in that deal (McDyess); just because Denver had no intention of keeping McDyess (due to their own roster and salary cap considerations) does not mean that he lacks value--far from it. This is what bothers me about WoW and other "stat gurus": they pretend to be engaging in a scientific, objective evaluation of basketball but once their precious systems make certain conclusions about player values then they feel duty bound to slant their articles to make it seem as though they have found the "Holy Grail" of basketball stat analysis. The formulas are considered to be error-free, so any difference between what the formulas say and what happens in the real world is the result of you believing what your "lying eyes" tell you. WoW stated that Iverson is a "below average" player, the Pistons struggled a bit after the trade, therefore WoW decided that this would be a perfect time to do a post reminding the world that for some time WoW has boldly gone against conventional wisdom with its rating of Iverson; data about McDyess does not fit into this "higher truth" and therefore is completely ignored.
I was disgusted after I read this WoW post and this has nothing to do with being a fan of a particular player or team; I really don't care how many games Detroit or Denver win this year and I don't think that Iverson is the best player in the NBA (though he certainly is well above average). What bothers me is that this post is either very poorly thought through or else deliberately deceptive--and although I think that WoW is misguided I don't think that they are stupid, so deliberately deceptive gets my vote. Basically, the post is the basketball equivalent of propaganda, like one of those political attack ads (by either party) that very selectively chooses information in order to not just make the opposition look bad but to completely distort the very essence of what that person is.
In my comments in the discussion thread after the WoW post, I made several of the above points and then offered three predictions:
1) By the end of this season, Denver will drop from second in the West (where the Nuggets were when the post was published) to a battle for the eighth-ninth spots; I expect the Nuggets to fall just short of making the playoffs, despite all of this overheated talk about a "new" Denver team and the ludicrous assertions in some quarters that Billups should be the MVP (which makes as little sense as crowning Gilbert Arenas an MVP candidate a few years ago when the Wizards were in first place in the East for a minute and a half but clearly were not the best team in the East, much like this year's Nuggets are clearly not an elite team in the West).
2) By the end of the season, Detroit will be battling Orlando for the third spot in the East.
3) Denver will struggle to beat plus-.500 teams, which is the main reason that I expect the Nuggets to plummet in the standings (they had been feasting on weak teams in the immediate aftermath of the trade).
There is still a long way to go, but the Pistons are 9-3 this season with McDyess in the lineup (two of those victories happened prior to the trade). Meanwhile, the Nuggets have gone 3-5 in their last eight games, including four losses by at least nine points versus plus-.500 teams; the Hawks drilled them 109-91 last night.
You can rest assured that whatever happens down the stretch, WoW will not budge from their engraved in stone conclusion that Iverson is a "below average player." If the Pistons drop off the map, it will be all his fault (according to WoW); if the Pistons perform along the lines that I expect, WoW will suddenly "discover" McDyess or someone else who will get credit for the wins, leaving Iverson to take the blame for the early losses. As for Denver, the story will be very similar but in reverse, because Billups scores well in WoW's metric: Billups will get credit for Denver's wins, someone else will get the blame for the losses.
Labels: Allen Iverson, Chauncey Billups, Denver Nuggets, Detroit Pistons, Wages of Wins
posted by David Friedman @ 5:42 AM
21 Comments:
I share your disgust at the purported "science" of the WOWers.
Sometimes you'll even hear them reply "the numbers don't lie" - true, but the people who interpret the numbers do.
David - Look, honestly, this is just silly. I can't believe you are bringing up Rodman again. How many times have we dicussed Rodman? At least twice, yet you still insist on trotting out the same old outdated and unoriginal criticism that Berri has addressed numerous times on his blog. I mean, what are you trying to be here? An attack journalist? This is just a long sequence of strawmen that you are setting up and knocking down.
You can attack all you want, but at the end of the day, as a basketball writer you have some very basic choices. Are you going to use offensive and defensive efficiency statistics in your analysis? Are you going to look at differential? Are you going to look at player's ts% when making your analysis? How much weight are you going to give to a player's impact on his team's net possessions?
My sense is that you think you will go farther in your career ignoring those tools. And I think that's a bad bet. Everywhere you look, in front offices, in telecasts, even on ESPN, advanced stats are more and more prevalent. Good luck with that, but at some point, being a die hard old school traditionalist who doesn't use new school stats is going to make you less relevant.
As for Iverson, I don't really hear you defending him. Do you really think Iverson is better than Billups? What is so great about a player with a career ts% of 51.9%? Where are the wins?
Owen
Allen:
The numbers have to be understood in the proper context. I very much support the idea of trying to scientifically analyze basketball statistics but the problem is that some people think that they have already found all of the answers when they don't even know enough about the subject to ask the proper questions.
Owen:
I don't know how often we have discussed the Rodman-Jordan comparison but if you still agree with Berri's take then apparently we have not discussed it enough.
It's funny--and yet sad--that you label me "an attack journalist." My post is a measured and reasoned response to one of the sloppiest and most slanted basketball-related articles that I have had the misfortune to read in quite some time.
I analyze basketball teams and players by watching the games and evaluating players' skill sets, coaches' strategic adjustments and various matchups. The difference between what I do and what Berri does is like the difference between eating a cake to find out if it tastes good and determining how a cake would taste merely by reading the ingredients; Berri can talk about all the ingredients in the "cake" but he has no idea how to bake it.
In your four paragraph comment, you did not address--much less refute--a single point that I made in my post. Point blank, there is no honest, credible way to defend writing an article about the effects of the Billups/McDyess for Iverson trade without mentioning McDyess. The issue here is bigger than whether or not Iverson is a better player than Billups; the issue is that writers should not deliberately withhold information in an attempt to mislead readers. That is my objection to the WoW post.
I realize that posting comments at WoW is kind of like going to the Vatican and arguing that Catholics should convert to a different religion but the Iverson post was such a jumbled, tendentious mess that I had to say something. I suspect that most of my regular readers don't bother to look at WoW at all, which is why I decided to post about the issue here. We can all look back on this discussion at the end of the season and see exactly how Detroit and Denver finish and how well Iverson, Billups, McDyess et. al. play.
Owen,
I am not a WoW aficionado and I don't recall seeing on this site the discussion you are describing about Rodman and Jordan. Would you be able to summarise how exactly Berri addressed this issue, or provide a link to where on the blog he did so?
Thanks.
I enjoyed this post, as I also find WoW frequently infuriating (which I why I stopped going there).
Tho I do find one thing Owen said interesting -- I'd be curious to read an actual eval/comparison of Billups and Iverson.
I definitely would vote for Billups, but I do so out of being a longtime fan and without looking seriously at any numbers or doing a detailed analysis of the players' skillsets. In my view, Billups is an outstanding free throw shooter who is very talented at getting to the line, an excellent "quarterback" of the offense who distributes the ball well (and earns "assists," reserving comment on the accuracy/consistency of this stat) and very, very rarely turns the ball over (almost always near the top of the Ast/TO ratio chart).
Iverson requires a ton of shots to score, tends to dribble the ball to death and forget the game is 5v5, and turns the ball over a fair bit, for all his steals.
Joel:
Of course, Owen can speak for himself and offer his perspective but the crux of the discussion involved Dave Berri's contention (on page 144 of his book, as I mentioned in this post) that on a per minute basis Dennis Rodman was more productive for the 1996 Bulls than Michael Jordan was. This statement attracted a lot of attention and criticism and inspired some back and forth between Berri and various critics. One of the big problems with WoW is that it overvalues rebounding; that is why Berri insists that Rodman was more productive than MJ and that Bynum was more productive than Kobe early in the 2007-08 season. Rodman was a HoF caliber player but I don't agree that he was more productive on a per minute basis than MJ; the idea that Bynum was more productive than Kobe is not even worthy of serious discussion--if Phil Jackson starts benching Kobe in fourth quarters to run the offense through Bynum and if Bynum becomes the main signal caller on defense the way Kobe is then we can talk about this.
As for comparing Billups and Iverson, I may very well do that in an upcoming post or article but the main point here is not so much which player is better but rather that the WoW post about Iverson is horribly tendentious, biased, one-sided--choose your adjective.
I've made my predictions about how both teams will do, so we'll see what happens.
I must admit I'm not a huge fan of formulas that seek to ascribe a specific value or rating to every player - especially when the creator of said model uses it as the basis for all basketball-related analysis. Advanced stats like TS%, eFG%, and offensive/defensive efficiency are all objective and easy to understand. Ratings like PER and WoW, whether people want to admit it or not, are based on one person's interpretation of what stats are important and how they should be weighted. Just because that person is a trained economist or mathematician doesn't mean his method of evaluating NBA players is foolproof.
I would love to know where in front offices, in telecasts, or on ESPN you will find someone who agrees with the assertion that Allen Iverson has been a below-average NBA player throughout his career.
Joel - See "On Jordan and Rodman, Again" at the Wages of Wins Journal
But the important line is this:
"This is the quote Kaufman takes from page 144 of our book. “Per 48 minutes played, Rodman’s productivity even eclipsed Jordan. Rodman’s WP48 of .0.415 was four times the production offered by an average player in the NBA, and even surpassed the 0.386 WP48 posted by Jordan.”
If Kaufman were to read the very next line he would see: 'Of course when one looks at standard deviations about the average, Jordan was still more productive than Rodman.'"
David - This is not the first time I have posted the above sentence on this blog. And I find it frustrating you are taking the exact same line of attack that Kaufman took two years ago.
Owen
Owen:
The "standard deviations about the average" business is simply a comparison of MJ's productivity to the productivity of other shooting guards, according to WoW. The bottom line is that WoW overvalues rebounding and thus concludes that Rodman, on a per minute basis, was more productive than MJ for the 1996 Bulls. Since Berri overvalues rebounding across the board (no pun intended), Rodman was not as far ahead of other power forwards as MJ was ahead of other shooting guards--but none of this changes the fact that Berri maintains that Rodman was more productive (and, thus, presumably more valuable) than MJ. As I have said, I think that Rodman was a HoF caliber player--but the Bulls could have had another good rebounder/defender at that position and still won championships with MJ/Pip, but if you took MJ (or Pip) off of that team the Bulls would not have won those six titles; MJ and Pip were the only two players who played for all six championship teams, while everyone else was interchangeable to some degree. That does not mean that those players lacked value--far from it.
Owen,
Thanks for the reference. I was able to find the post and read Berri's explanation which cleared things up a bit.
To Owen, so what? Are you saying there are different types of WP?
EL
Joel:
You will not find an informed NBA observer who agrees with the assessment that Iverson is a "below average" NBA player. Iverson is a high usage player or, in colloquial terms, a bit of a gunner. He shoots a lot and that is not everyone's cup of tea, whether you are a GM or a fan--but Larry Brown and the Sixers proved that Iverson can be the centerpiece of a championship-contending team. A below average player cannot lead a championship contending team in minutes, scoring, steals, etc. Iverson is a unique player because of his physique, durability and playing style, so he is hard to categorize but "below average" is not an intelligent or accurate description of his place in NBA history.
You are absolutely correct that PER and WoW offer specific (which is to say by definition, biased) interpretations of what boxscore numbers mean. Those interpretations may be 50% right or 25% right or dead wrong but they are not scientifically tested and verified theories. Stat analysis is an interesting but limited and flawed tool for evaluating basketball players and teams.
My issue with WoW is its "militant" attitude (I don't see that attitude in Hollinger, for instance). I am all for deeper statistical analysis, and I firmly believe that almost everything in a game shows up somewhere in the stat sheet; but that does not mean that one specific formula is the holy grail. Specially when it is based on notoriously unreliable and incomplete data as box scores.
I do not have a problem with the WoW "tool" per se, many teams use statistical tools and I am sure they are truly educational. But I do not like how any discussion ends up in a "my way or the highway".
What's wrong with using statistics as a tool but not as the sole basis for analysis?
Thanks for your article, I share most of your concerns. But I would've liked to see more argument focused on Iverson being a below average player, versus the general attacks of stat-based observation, which while valid in many ways, is a well-worn topic. Instead, prove their theories wrong by disproving their specific example, which you only began to do with McDyess.
Arthur:
The main point of my critique is that the WoW article was biased, tendentious and poorly written. WoW stated that Iverson is "below average" but did not present a well reasoned argument to support this contention; the burden of proof is on WoW to prove that Iverson is "below average," not on me to prove otherwise. That said, there does seem to be some interest among 20 Second Timeout readers for me to do a post comparing Iverson and Billups, so I may end up doing so--but that comparison has nothing to do with WoW's article, which was sloppily put together.
TP:
I agree with you completely about WoW's attitude, which I would call "arrogant." You are right that Hollinger is much more willing to admit the limitations of a stat based approach in general and even the limitations of the methods that he uses.
WoW not only refuses to take such a sensible approach but very obviously and intentionally aims to mislead casual readers into believing that the WoW metrics are much more accurate than they really are.
NBA teams have been much more interested in what Roland Beech, Dean Oliver and Dan Rosenbaum have to say than in anything that WoW has produced. As you said, stat analysis is a tool, not the holy grail; those three guys understand that and that is why the NBA takes their work seriously.
If you based your system on some statistical algorithm that said Billups a far inferior player to Iverson, and you wanted to be like Wages of Wins, you'd now talk about how Denver is going to lose games because of Billups, while conveniently not mentioning Carmelo's broken hand.
Free Cash Flow:
Exactly. If Melo misses several games, what WoW will do is use his absence as an excuse to explain why Denver lost to some plus-.500 teams that the Nuggets--based on their track record--were going to lose to anyway. So Melo's injury could end up masking how off base WoW's predictions about the Nuggets are--but if nothing changes in Detroit then everyone will still be able to see that adding Iverson hardly damaged the team. The Pistons are 10-3 in their last 13 games (the Nuggets are 10-5 during that same period of time).
David - you are too civil to say it, so I will do it: that Tball character over on the Wages of Wins Journal commenting section is an imbecile.
Regardless of the outcome of tonight's game your analysis would have been sound and valid.
I'm only happy that a moron like Tball does not get to have the satisfaction of seeing his prediction ("Denver will beat Detroit, regardless of Carmelo Anthony's availability") come true, which only feed fuel to his fire of ignorance.
I think it's best to leave such lost souls in the wilderness, especially when you've given them a terrific compass which they refuse to use.
Free Cash Flow:
I'm responding to him not so much because I think that he will understand and/or agree but so that other, more open minded people who find that thread can see an alternative point of view.
Most of the time I just ignore what is going on at WoW because their basic methodology is so flawed but this post really irritated me because not only is the math bad but the writing is very tendentious.
If someone wants to make the case that Billups is better than Iverson, fine, but he shouldn't do the measuring with his thumb on the scale by completing disregarding McDyess' impact.
Post a Comment
<< Home