132 of 151 people found the following review helpful
- Published on Amazon.com
I really looked forward to this book after reading the review in The New Yorker. The reviewer's critical skills, evidently, do not extend to evaluating the merits of a logical argument.
There are so many logical problems with the analysis in this book, it is difficult to know where to begin.
I will limit myself to just a handful, among countless possibilities.
1) The authors find a correlation between the stability of a basketball team's roster and its winning percentage, and conclude that roster stability is a factor in producing wins! Classic problem of mistaking effect for cause. Clearly, winning teams are disinclined to make major roster changes, and losing teams are eager to. I was so amazed at this I reread it to see if I missed where they pointed this out. They didn't.
2) The authors show a correlation between more assists and winning percentage and conclude that assists help produce wins. Again, very silly. A team with a higher shooting percentage and fewer turnovers will of course get more wins and produce more assists. But the assists are not producing the wins - the shooting percentage is. Were these factors discounted? Not according to the text.
3) Most problematic, the authors define a way of measuring the value of players to a team, and then "prove" their method by summing these values, per player, across each team, and show that they do indeed predict the number of wins each team will get. What they fail to realize is that their method of apportioning value to a player necessarily sums back to team totals such as points per possession that we know correlate to wins per team. But this in no way proves that the apportioning is wrong. We could just as easily base each player's value "team's points while players is on the floor - opposing team's points while player is in the floor". Sum for all players on team, and you will find the team's points-per-game and the opposition's points per game, and you have just "proven" that your method of measuring a player's value is accurate.
4. In looking at rebounds, there is not even the slightest caveat that a player's rebounds per game are affected by who he shares rebounding responsibilities with. If you are in the front court with Shaq, you will get fewer rebounds, not because the other team gets them, but because your teammate does, so comparing rebounding stats between players across teams is highly questionable.
5. Ridiculously, they conclude that adding great players to your roster makes the other players worse not better (this is written as a great revelation, demythifying the common presumption that great players make teammates better). Adding great players indeed will make other players statistically less productive. They will take fewer shots, get fewer rebounds, fewer assists if a new ball-handler is added, etc. And so, according their evaluation method, the other players become worse. This should clearly tell them there is something wrong with their system, based so heavily on specific metrics of productivity. Instead, they simply accept the merit of their method as fact, and conclude that adding great players really makes other players worse!!!
The most frustrating thing about this book, however, for an analytically minded reader, is its smugness. They understand statistics. They have the answers. They are bringing them down from the mountains and explaining them to us idiot peons. All while their reasoning is so problematic.
I in no way am a supporter of the intuitive and nonsensical drivel one hears from so many sports coaches, players and commentators. I would have enjoyed a good, statistical, analytical study of the game of basketball. Unfortunately, this is not it.
22 of 29 people found the following review helpful
- Published on Amazon.com
I had high hopes for this book but my expectations were not met. The authors are clearly eager to bear the Freakonomics mantle (they say as much in several places), but unfortunately they do not exhibit anything resembling the flair of Levitt and Dubner. Their cute comparison of quarterbacks and mutual funds just sounds like a cheap imitation of the comparison between teachers and sumo wrestlers (which, honestly, wasn't all that clever anyway). Much of the other writing also seems to imitate the conversational style of Bill James, but without as much wit. Overall the writing comes off as alternately condescending and self-congratulatory, and sometimes both.
Style aside, the book contains a number of substantive weaknesses. For example, the chapter on the effects of labor shortages on fan attendance shows clear signs of bias. The authors favorably cite plenty of evidence that supports their hypothesis; and when confronted with evidence to the contrary, they suddenly decide to pick it apart and explain it away. Sorry guys, it doesn't work that way. This clear example of "disconfirmation bias" causes the chapter to lose all credibility. It wouldn't hold up in a peer-reviwed journal.
Further, although the authors claim to be "taking measure of the many myths in modern sport" (the subtitle of the book), they actually devote a lot of effort to knocking down strawmen. Is there anyone alive who really thinks that "the best players in basketball score the most" or that "quarterbacks should be credited with wins and losses"? No one with more than a passing knowledge of sports actually believe these things, but the authors act awfully smug after debunking these nonexistent "myths." Yes, we're all aware that offense is at most half of football, and that the passing attack is only about half of that. Luckily no one attributes wins to quarterbacks, except maybe to point out that a team can win with a mediocre QB (e.g., pointing out Trent Dilfer's career winning percentage) -- which is a different issue altogether.
The book also spends a lot of time trying to analyze basketball using methods that are much better suited to baseball. Don't get me wrong, I admire their effort to subject basketball to some analytical rigor. But baseball is largely an amalgam of statistics and can be studied as such. Basketball simply cannot. There are too many events in basketball that clearly affect the game but are not quantified (a pick, a shot that is altered but not blocked, a team deciding not to drive against a particular player, a player drawing a double team and getting a teammate open, the second-to-last pass of a possession). One might conclude, based on the demonstrated strong correlation between wins and the conventional statistics employed in this book, that these events are all relatively unimportant. But this argument ultimately fails because the purpose of the analysis is to measure the contributions of individual players. A team might score two points but the model does not adequately break down individual contributions beyond who scores the points and, if applicable, who gets the assist. Similarly, most of what happens on defense isn't recorded, and the model only takes into account steals and blocked shots. The authors sweep these weaknesses under the rug and proceed to devote dozens of pages to comparing players based on their new, supposedly superior, measures of individual performance. This is an enormous flaw.
Further, I was also struck how a team of economists could write about the value of basketball players without paying attention to the supply curve. They do adjust some of their stats for league-average at the position, but not on a category-by-category basis. In the final chapter, where they purport to show that scorers are paid too much, they fail to examine the issue of scarcity. My wild guess is that the data would support their conclusion, but I was struck by the absence of real analysis here.
Of course, no book on sports statistics and/or economics is complete without the obligatory nod to the genius of Billy Beane and the claim that salary disparities do not lead to competitive imbalance. This version of the story is no more convincing than any of the others. They happily point to the 2003 Marlins as an example of a low-payroll team winning against the odds, but somehow ignore the fact that a number of those players (Derrek Lee, A.J. Burnett, Josh Beckett, Ivan Rodriguez, Alex Gonzalez, Mike Lowell) are now earning big salaries in big markets while the Marlins are under .500. Sure, a team can win with young players who haven't yet become eligible for free agency or arbitration, but is that any way to build a franchise for long-term success? Where is the analysis of that rather obvious question? And where is the point, made quite clear in Moneyball, that no inefficient market can last forever? What happens when the next Billy Beane is hired to run the Yankees?
I will grant that the book is thought-provoking. But ultimately there are many other books on sports statistics and economics that are much more readable and well-argued than this.
11 of 16 people found the following review helpful
- Published on Amazon.com
Berry and his co-authors take a really fresh approach here to the world of sport, and look at the big dollars that get spent - or mispent - by teams that want to win at all costs. The mistake that team owners are making, according to the authors, is that they are looking at the wrong statistics of success. When you buy a player, what are the vital stats that really matter?
The book is at once disheartening (money buys one hell of a lot of the points we see on the various league tables) but then also entirely heartening: individual sportspeople, and b-teams still rock the game and lift their performance way above the odds.
The difference is that many managers and coaches (and us fans for that matter focus on the wrong numbers. We count the goals that a player scores, but discount the fumbles, the turnovers and the other dynamics of team play. A star goal shooter may actually be bad for the team's chance of winning.
Sports these days is dominated by statistics. You listen to a commentary and you hear the win/lose ratios, the all-time earnings figures, the goals, the assists and what have you. Well here the authors put another set of very relevant numbers on the table - and show how they reached these conclusions. Their arguments are pretty convincing and as a result I think we'll be talking about this book for some time to come - and pondering the nature of sports today and whether Player X really is greater than Player Y. What a fresh piece of writing! Buy it - I'm sure you'll enjoy it as much as I did.
- Published on Amazon.com
Format: Kindle Edition
The reality is this belongs on the book of every sports fan. All of us have believed so many things about sports for years. We've believed work stoppages drive fans away, or that Michael Jordan was the reason the Bulls were the greatest team ever. Lost in these narratives are the nitty gritty science and economics that show how sports really tick. Dave, Martin, and Stacey delve deep into sports' underbelly to reveal what you've been missing.
3 of 5 people found the following review helpful
Nathan E. Walker
- Published on Amazon.com
Berri's ignorant dismissal of years and years of research on how to quantify player efficiency/value/production for basketball is one of the stupidest straw-man arguments I have ever seen:
"This is the point we are making about decision making in the NBA. It is
not that people in the NBA are lazy or stupid. It is just that the tools at their
disposal do not allow them to see the value of various actions players take on the
And yet, Berri's purported values of rebounding (among other things) is, by every statistical model imaginable, absolute crap.
The problem here, like many reviewers have noted, is that basketball is NEVER a one-on-one interaction. To base values on the marginal production of a specific statistic continues in the future is absolutely stupid. Berri cites the "repeatability" of rebounding as reason to "believe" in it, specifically. But the confounding variables that make up rebounding are absolutely through the roof.
-Nine other players have the opportunity for the rebound, there are marginal returns on rebounding, especially defensive rebounds(Eli Witus)
-Defensive rebounds are frequently due to chance/non-skill related factors (Dean Oliver)
-Team pace, opponent team pace, opponent field goal attempts, team field goal attempts, position relative to basket: these all explain huge amounts of variation in rebounding (Obviously).
Here are two quick examples of this confounding:
-Imagine there are two players that are perfect clones, and will respond exactly the same given the same situation. If Clone #1 plays Center and Clone #2 plays Power Forward, Clone #1 will be more likely to get the rebound simply due to the position he plays (closer to the basket).
-Imagine an opponent shooting a well-guarded 3-point shot. The fact that the rebound exists is, to a high degree, the responsibility of the one forcing the missed shot (Dean Oliver). Whoever catches the "fly ball" can improve their position to catch it, yes; but players are not capable of predicting very accurately where the ball will bounce towards (or IF it will bounce at all).
-This one is what Berri misses entirely: players that rebound well offensively defend more poorly (SPM model). In fact, in my one-year studies on Regularized Overall Plus-Minus and offensive rebounds, there is (relatively) NO correlation between player offensive rebounding performance and team efficiency margin. This has also been discussed on the team level (by Pomeroy, for college hoops) - that offensive rebounding teams fail to 'get back' on D.
-Dirk Nowitzki physically CAN'T get your overvalued rebound as frequently because he is busy shooting three-pointers, but when he's not taking the shot, he is right there. (Okay, this is less of a variable and more of a rant).
The stupidest result of this system that bugs me to no end is his current 2011 numbers saying that Kevin Love is the most valuable player in the NBA. The adjusted plus-minus model, to a very high degree of cross-validated accuracy, puts Love at around the 170th-most valuable player. Berri's dismissal of the Adjusted Plus-Minus model is mind-boggling. By definition, the standard adjusted plus-minus model predicts OUT-OF-SAMPLE data with the highest degree of reliability. And the regularized plus-minus model does the same in terms of precision.
His other argument that Adjusted Plus-Minus data does not work since it does not correlate year to year is through his lack of understanding of how the statistic works. Yes, a statistical model for plus-minus (which Berri SHOULD have done) would carry over yearly...and Berri's mythical model will obviously carry over with high degrees of year-to-year reliability. But honestly, who cares if it doesn't carry over year-to-year if it predicts any given sample of data with higher accuracy? And the extremely accurate statistical plus-minus models (and the conglomerates of them) still use box scores under the THEORY of the adjusted plus-minus model, and predict more accurately than Berri's "WinsProduced."
"Post hoc, ergo propter hoc" might work somewhat with individual statistics in a sport like baseball, but to apply simple correlation of player statistic A to team statistic A is, and I mean this, one of the most shortsighted things I've ever seen in basketball analysis.