Our next book – Stumbling on Wins (FT Press) – will officially be released next week. But you can already order the book at Amazon.com (in hardcover and the Kindle edition). And now we have some early reviews.
Henry Abbott has actually offered two comments on the book. On the back cover of the book you will see the following:
“This book takes the hallowed traditions of sports decision making and pokes them with a sharp stick.”
That is a great sentence on the book. But 17 words is not enough. Henry has also offered more than 2,500 words at TrueHoop. In “The robots are coming, and they’re cranky”, Henry details what he likes about our latest (quite a bit) and what he isn’t quite as thrilled about. All in all, it’s a very good discussion of our book. So please check it out.
And Chicago Tim – an commenter at this blog – also has had a chance to read our latest (via Kindle). Here are the comments he posted a couple of hours ago:
I’ve now finished The Wages of Wins and Stumbling on Wins on my Kindle. I liked them and recommend them. I’ll admit that I don’t follow all the math, and I don’t understand the disputes between various sports statisticians. But you and your co-authors do a good job of making your theories sound plausible to non-experts like me. And although the statisticians seem to disagree about precisely which statistics to use, they tend to agree that GMs in all sports persistently make many mistakes based on conventional wisdom.
I like to have the theories you cite in the back of my head when I read sports stories, so that I can question the conventional wisdom. Maybe it is okay that the Chicago Bears don’t have a first-round draft pick. Maybe the Chicago Blackhawks don’t need to spend big money on a goalie. Maybe the Chicago Bulls don’t need to spend big money on a coach. And maybe the Chicago Cubs and White Sox can improve their farm system by drafting more college players and position players, and fewer high schoolers and pitchers.
In Stumbling on Wins, your latest book, I like the way you link sports stories to the broader economy and even, perhaps, to broader theories of social and political science. Maybe people don’t act rationally, because they can’t compute all the variables and rely on conventional wisdom. Maybe they act irrationally repeatedly, decade after decade after decade. Maybe some of our heroes don’t deserve their fame, and some forgotten people should be remembered and given their due. If highly-paid experts make these mistakes in sports, a relatively-closed environment with many objective measurements of performance, how much more likely is it that we make such mistakes outside of the narrow world of sports?
I would have liked even more stories, and a more detailed discussion of sports like hockey and soccer. But I guess I’m just greedy.
I will say that having read this blog pretty regularly, there were some stories I recognized and already knew. However, in the books you do present the arguments more systematically and thoroughly, and there is much in the books that I do not remember from your blog. I also found some redundancy between the first book and the second, since you couldn’t assume that everyone was familiar with the first. But you put the more complex equations in the appendices, where I was free to skip or skim them.
Finally, I like your dry sense of humor, of which there are many examples in both books and in this blog. If anything, you undersell your theories. Some of them are pretty shocking, such as the possibility that black quarterbacks are systematically undervalued.
One nitpick: in most cases “use” can be substituted for “utilize.” In rare cases, “utilize” is more appropriate, that is, when you utilize an object for a purpose not originally intended, such as picking your teeth with your knife.
We look forward to seeing much more discussion of our book over the next few months. And we encourage everyone to buy the book and participate. Or if you don’t want to participate, just buy the book :)
- DJ
The WoW Journal Comments Policy
Abbott writes:
i think this rather skews Berri’s argument (though I haven’t read SoW yet, so I might be wrong). As I understand it, there isn’t too much of a drop off between 24 and 28.
But while Abbott makes a cutesy res ipsa statement, I think that his statement doesn’t actually speak for itself. Most elite players are drafted through the lottery, meaning that they join bad teams. It’s thus hardly surprising that their GM hasn’t yet built an elite team around them at 24 when they are peaking in production.
But Abbott presents real evidence that older players are better defenders. Abbott makes a lot of good points in that post, and I was hoping to come here and find a more substantive response to his critique.
I especially liked this bit from Abbott, regarding Berri’s contention that teams don’t value rebounding and defense: “Common sense, or a peek at their public comments would reveal that every basketball expert thinks every team needs things like rebounding, passing and defense, even if they buy them on the cheap. The proof that 29 other teams wouldn’t have done what Thomas did is that so many of them willingly traded their scorers to Thomas’ Knicks. Even in real time, articles were published explaining how just about the smartest thing a GM could do was make as many deals as possible with Thomas, and more than a few teams did.”
I don’t see any “real evidence” regarding the age issue in Abbot’s review. The only evidence he presents is that rosters of NBA champions tend to be old, but as Evan has addressed, that doesn’t really prove older players are more productive defensively since young players not long out of draft often play in bad teams. Plus many of young players are unproductive but still play because they were picked high and there’s that hope of improvement.
Also most of the greats cited in the article Abbot referenced were in their mid-late 20′s, not near 30′s, when they first won their championships.
I’m not very convinced of WP’s way of apportioning defensive shares equally either, but I just couldn’t find a way to make defensive numbers with good predictive power. Ty at Bucksdiary has a fascinating system with somewhat similar idea as me, with adjustments of opponents’ numbers with 82games and popcornmachine data, so that’s something interesting to see. But there’s always that issue of predictive power in regards to player’s contribution to wins from year-to-year, which I think is still the strength of dberri’s system.
simon,
Here’s the TrueHoop post with the data on age and defense:
http://espn.go.com/blog/truehoop/post/_/id/6025/evidence-that-age-makes-nba-players-better-defenders
The thing that I like about Abbott’s critique is that Abbott appreciates that the statistical revolution in basketball is all about asking questions. We just don’t know enough to get all the answers, but guys like Morey and Dean Oliver and Sam Presti and Mark Cuban are using math to at least ask the right questions. The WoW system is just about answers — there are no interesting questions in this world, because all you have to do is plug in Berri’s formula, and there’s your answer.
Abbott, again: “Given that there’s a lot we know we’re not measuring in hoops, you might think Berri and Schmidt would feel the need to bolster any findings with other evidence — from video, anecdote, other statistics.”
Abbot discusses the limited estimation of defence in WP48. But he goes about it the wrong way. The question is not whether WP48 directly encompasses defence correctly. The question is whether WP/WP48 (a) predicts wins and (b) is attributable to the player. If these hold then it it follows that it WP48 takes defence into account. It doesn’t matter how precisely (a) and (b) take place for this to be true.
Oops .. must have gotten my tags mixed up …
Abbott brings up something that I’ve had an issue with for some time. The post-season scheduling gives older teams an advantage. There are more days off between games; often two ansd sometimes three. I’ve always felt the playoffs should reflect the regular season schedule. Let’s see some back-to-back games and less travel time between games.
I think it’s a pretty fair review by Abbot, all things considered. I haven’t read the book yet (obviously), so it’s hard to know if it’s an accurate review. But I think it’s good-spirited, and based on the previous book and the blog, it raises fair points.
I’d throw my name in the hat that would be interested in seeing some type of formal response (although I admit it may already be addressed in the book, which I plan on reading).
Here are the things I found insightful and would be interested in hearing a response to:
WoW seems to assume that GMs value skills and players by how much they pay them. Obviously, that’s true in a way. But I think the point Abbot’s making (and I think I agree) is that some GMs fully understand the role of rebounding, but they also understand that they can buy that cheaper than they can buy a scorer.
The entire section on “In defense of coaches” is very interesting, especially with respect to defense. I realize that we only have the statistics we have, and we’ve gotta work with what we’ve got. But I think it’s important to admit (and even emphasize?) that current statistics don’t track individual defense adequately.
I’m anxious to get my hands on the book. I’m sure it will be a compelling read.
todd2,
I bet you’d be hard pressed to find many players, coaches, or management personnel that would agree.
And frankly, as a fan, I wouldn’t agree either. I’d rather see healthy, rested players compete to their highest level than tired, sloppy basketball.
“Common sense, or a peek at their public comments would reveal that every basketball expert thinks every team needs things like rebounding, passing and defense, even if they buy them on the cheap.”
We actually address this in the book in the chapter on the NBA draft. ObviouslyI encourage everyone to read the book.
Brgulker, while yes coaches and management understand that you need rebounding, passing, and defense, it’s been shown repeatedly in this blog not to be the case when signing players.
It’s not just a question of Player A being a great scorer and Player B being a great rebounder, and player A makes more than player B. What’s at issue is that Ben Gordon and Charlie Villanueva got paid major dollars when all they do is score. Both are underwhelming in virtually every other category. Al Harrington makes $10M this year, does that sound reasonable?
Carmelo Anthony’s regularly listed as one of the top wing players in the league, yet he is lightyears behind Lebron and Wade in everything but scoring. The same can even be said about Kobe.
Too many Steves, one possible explanation for “Age vs Defense” is extremely simple. Every year, several players are taken in the NBA draft who simply can’t play. With the rookie scale contract it can take up to 4 years to get rid of those players. Given that those players are 19-22, and their first contracts run until they are 23-26, is it any wonder that teams are better defensively when they’re able to drop these poor selections? I submit that the pool of players 26+ are better defensively because the pool of players under 26 is loaded with wannabes. As soon as the Adam Morrison’s of the league are gone, their “aged” teams miraculously get better.
Xavier Q,
Yes, I know. I’ve read WoW, and I read the blog. You’re basically just summing up major conclusions, which is fine, but I already know those things.
I’m thinking something more specific. Yes, Denver employs Melo, and Melo is a sort of poster child for WP. I find WP convincing, so I tend to agree that Melo is overrated because of his PPG averages.
But, in defense of Denver, they also employ Nene and Birdman, and Birdman is pretty cheap. They also traded Iverson for Billups (and McDyess, who was brilliant last year but didn’t end up playing for them).
So, when I look at a team like Denver, it strikes me that they might actually get it. They might overestimate Melo’s worth, but they also seem to have a very clear grasp on the value of efficient scorers who do lots of other things (Billups) and good rebounders who score at efficient rates around the basket (Nene, Birdman).
The fact that they pay the latter less than the former doesn’t necessarily mean that they value those skill sets less. It might simply be shrewd business practices — I have to pay more for a scorer, because that’s how the market values him, and I can pay rebounders less, because that’s how the market values them.
too many steves//
Ah thank you for the article. That’s a interesting read that I haven’t seen before. I’ll have to go look at Pelton’s study a bit more.
I don’t know if I agree about “(GMs) math to at least ask the right questions” whereas dberri’s method is just “plug in the numbers” mundane-ness. I feel they are inherently designed to address different issues; one is more for decision making in very specific in-game situations whereas the other system is for player evaluation with season-to-season predictive power on performance. I guess what people want is, like what baseball sbmetricians did, to get %s of success/failure on very specific in-game situations, but I don’t think that’s what dberri&co aimed with this model at all.
xavier//
I’ve also said the same thing above about poor young players who get minutes only because of their high draft status. But I wonder if there’s an indication of older units performing better specifically on defense only.
too many steves//
Your comment about valuing the +/- crowd’s questions over WoW’s answers seems perverse. Even if we know WoW doesn’t take non-box score variation in individual defensive ability into account, it still makes much more sense to use WoW as a guage for assembling teams and paying players than some adjusted +/- system that claims to measure individual defensive contributions, because WoW has a close correlation with wins and, as far as I’m aware, no adjusted +/- system does (the reuslts vary so much from year to year the models have basically no predictive power). If someday adjusted +/- does get predictive power, then yes, of course it would make sense to use it. But privileging an unproven but promising system (+/-) over a proven one (WoW) is just not smart.
As for Abbott’s defense of traditionalism, once the status quo has been shown to be horribly flawed (and I think WoW/SoW has conclusively demonstrated that most GMs vastly overvalue qualities that correlate very poorly with wins, like inefficient scoring, appearance in the Final Four, etc), it is very strange to nonetheless reflexively claim allegiance-until-proven-otherwise to that same status quo. The more reasonable hypothesis would be to assume that if teams have been shown to go so consistently and disastrously wrong on the things that matter most to wins (who to give max contracts to), there is no good reason to suspect that their other decisions are any better. I.e., Oliver’s ‘null hypothesis’ shouldn’t be cheered. It seems more likely that Oliver finds it professionally and personally advantageous to take that position than that he really believes it, but even if he really believes it, so what? It’s still wrong.
For those looking for more of a response from me, can I just say “see Patrick Raulerson’s comment”? I think he essentially gets it right.
Thanks for your comment Patrick.
I don’t think the “allegiance-until-proven-otherwise” thing is strange at all. It almost always happens in any field, be it a field of “hard” science or a field “soft” science. Just think of what basis theists were supposed to be “the” authority of morality.
Rookies can be great right after the draft, but those are once in a life-time players, or situations. For instance, Tim Duncan ending up on a power Spurs team, or Magic being drafted by a Lakers team with Kareem already starting on it. So the age thing is still questionable, but usually, by the time the team is assembled to win a championship, much of the core is past their early years and their supporting cast is much older, just consider the CLE team this year.
Also, http://courtsideanalyst.wordpress.com/ attempts to “capture” defense, and to some degree it seems to work, and is often apparent when you witness the game itself.
Anyway, can’t wait to read the new book, thanks!
Thank for that link Alvy,
If I’m reading marginal win score correctly, then according to it Ginobili and Wade still produce at a higher rate than Kobe. It also says that Kobe is indeed the best overall player on the Lakers both by production and production rate.
Brgulker, we’re looking at this from different perspectives. I’m not saying that the Nuggets, for example, don’t value rebounding despite having Birdman and Nene.
The issue is that scoring (and volume shooting) is vastly overrated, hence lumping Melo in with Lebron and Wade even though he performs substantially less. The going rate for Andrew Bogut (16/10/2.5 52%FG) is $10M. The rate for Melo (29/6 46%FG) is $15M, the same as Lebron (30/7/8 50%FG) despite being more on par with Bogut.
The argument is not that NO ONE values rebounding, etc. but rather that money is being tied up when it really shouldn’t be (i.e Melo should not be paid like Lebron).
Re: +/- for what it’s worth. Anderson Varejao is 2nd in the league in +/- according to the ABC telecast this weekend.
brgulker, I won’t argue your point. It does make for better basketball—for television/marketing. I feel the current scheduling format favors older teams and I’d like to see the younger squads get their due. Let the playoff format match the regular season scheduling and keep it consistent.
Patrick Raulerson,
re: “Even if we know WoW doesn’t take non-box score variation in individual defensive ability into account, it still makes much more sense to use WoW as a guage for assembling teams and paying players than some adjusted +/- system that claims to measure individual defensive contributions, because WoW has a close correlation with wins and, as far as I’m aware, no adjusted +/- system does (the reuslts vary so much from year to year the models have basically no predictive power).”
Please show where the WoW metric … or, any other metric, for that metric … has the ability to predict a team’s future wins totals with a high degree of accuracry.
Thanks, in advance.
Thanks for reprinting my review. I hope it helps people decide.
By the way, the next book I read was The Greatest Trade Ever by Gregory Zuckerman, looking at the people who defied conventional wisdom and bet against the real estate market, thereby making billions of dollars. You can make money defying conventional wisdom! However, Zuckerman shows that it can be very difficult to go against the flow.
I found this quote from John Maynard Keynes, which Zuckerman used in his book, particularly telling. “Wordly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally.” Perhaps people are rational after all — and they have rationally decided that they would rather fail conventionally than risk their reputations by succeeding unconventionally.
Chicago Tim,
Thanks for providing this terrific quote from John Maynard Keynes, concerning the value of what is properly termed “unconventional thinking” within a specific and relatively homogenious environment.
Speaking of rationality …
In a world gone completely mad[d] the only sane man is the one who appears to others to be wholly irrational.
khandor, if you wish to predict future win totals, you can simply look at the WP48 for each player on a given team in a given year and then, based on those WP48 numbers, multiply by the number of minutes played and divide by 48, and add up the net contributions of each player. For example, if I wished to see how well the WP48 metric would predict the total wins of this year’s edition of the LA Lakers, I would first acquire WP48′s from the previous season for all players on the Lakers’ current roster, and then multiply each WP48 by the number of minutes played by a player and divide by 48, and add all these numbers up. If Kobe posted a 0.24 WP48 last year, and we wish to predict from that the number of wins he would produce this year, we go through the above calculation. By doing this procedure, we can determine the projected wins for every team and compare it to the actual number of posted wins.
The good thing about WP48 is that it has more predictive power than say, PER. Performing a similar calculation based on PER’s would, on average, produce projected win totals that are more off than the WP48 calculations. Obviously, because of things like player improvement and plain old randomness, no such metric would be perfect, but WP48 is still better than the other tools we have.
Finally, I’d like to ask you in the future to do your homework and actually understand the metric before taking shots at it. It’s fine that you don’t necessarily agree 100% with it, as all good research goes through heavy criticism at some point (and is never 100% perfect), but all your critiques of it seem to stem from the fact that it doesn’t agree with your preconceived theories of what should win basketball games. It seems like 99% of your posts here, khandor, in response to a dberri WP48 analysis of any given team, are something to the effect of “oh no, numbers can never tell the story. *scoff* Here is what REALLY happens. *giant wall of text* Obviously, I’m right, simply by virtue of repeating my opinions repeatedly. I don’t need any real empirical evidence. You’re absolutely wrong, just because. *scoff*”
You believe all these theories about basketball, but you never really bother to test them thoroughly. You mention on your webpage that taking the operation points scored – points allowed + rebounding differential is the best measure for the quality of a team, but if you actually bothered testing that notion against, say, regular season records or even eventual NBA Finals winners, you would see that your calculation is not, in general, as effective as simply taking the standard efficiency differential of points scored – points allowed per possession (which is what dberri and most other people do) and basing predictions on that. So I’m kind of surprised that you feel like you’re so much smarter at basketball than the other readers on this blog.
brian,
1. re: the predictive accuracy of WP48 for a team’s future wins totals
A. The question I asked is concerned with the actual correlation of the process you outlined and the future wins totals of teams in the NBA.
If you actually know what this correlation is it would be most helpful if you posted it here.
If not …
B. Whether the correlation which exists between WP48 and a team’s future wins totals is “stronger” than the correlation which exists for PER and a team’s future wins totals is of little/no concern to me.
e.g. if the correlation to PER is, in fact, low, and the correlation to WP48 is, in fact, higher but, still, relatively low and not especially strong in itself
2. re: “You believe all these theories about basketball, but you never really bother to test them thoroughly. You mention on your webpage that taking the operation points scored – points allowed + rebounding differential is the best measure for the quality of a team, but if you actually bothered testing that notion against, say, regular season records or even eventual NBA Finals winners, you would see that your calculation is not, in general, as effective as simply taking the standard efficiency differential of points scored – points allowed per possession (which is what dberri and most other people do) and basing predictions on that. So I’m kind of surprised that you feel like you’re so much smarter at basketball than the other readers on this blog.”
Please tell me where you’ve ever read something which I’ve written that said the following?
“The best method of evaluating ‘Team Quality’ is points scored – points allowed + Rebounding Differential.”
as, IIRC, this is not something which I’ve ever said.
http://en.wikipedia.org/wiki/Regression_analysis
dberri basically performed one of these. WP48 is capable of predicting, for example, how many extra wins a team would produce (our dependent variable) on the average if so-and-so player were to, say, get 2 more rebounds per 48 minutes and all other independent variables (the other box score stats) were fixed. This kind of calculation can be extended to more complicated situations.
I’ve explained to you the gist of everything already. If you’re still not clear, then you should try reading the book and/or the first few blog entries, as well as the wikipedia entry.
what I quoted was your QIR, I think. If you actually bothered to test it rigorously against teams’ win/loss record as well as against the eventual NBA champions you’ll find that it falls short compared to efficiency differential, which is the gist of what most people use save for a few extra bells and whistles like team adjustment.
Read the book. It’s there. No one is obligated to re-post Dr. Berri’s published work for you.
Pingback: Cut-and-Paste Guest Post from Tony « Courtside Analyst