Thursday, February 02, 2006

A Review of "The Book on the Book"

The last book review I did was on a seventeen year old book; this one is a little less then a year old, so maybe by the next time I do one it will actually be timely. Anyway, The Book on the Book by Bill Felber is subtitled “A Landmark Inquiry into Which Strategies in the Modern Game Actually Work”. I hate to rip on these, because I realize that the publisher wants to make the book sound dynamic and exciting, but this book is not by any means a definitive book on the Book, nor is it a landmark inquiry into anything.

If you are well-read in sabermetric theory, you will not learn anything from this book. If you are not, you still may not learn anything from this book. The studies Felber presents are wrought with questionable methodology and selective sampling problems. For example, a study on how pitchers who pitch a lot of innings in a given season are affected by fatigue studies the late season performances of pitchers who pitched 250 innings. 250 is a lot of innings to pitch in the modern era. In order to reach 250 IP, you generally not only have to stay healthy, but also pitch effectively. Pitchers who have their effectiveness sapped by fatigue do not reach 250 IP. It’s like studying the people who finish the marathon and concluding that they didn’t collapse in the last five miles.

There are also some misrepresentations of sabermetric concepts, such as Run Expectancy, which Felber at one point refers to as “odds”. The standard RE table does not present odds, it presents expected values. There are tables of course that do print the odds of scoring x runs in a given situation, and these tables are actually more valuable then a standard RE table, because you can derive the RE from them. But the table Felber presents is a standard table. Actually, the discussion on using RE to value strategic decisions that follows really isn’t bad--but it’s at about the same level as what was published twenty years ago in The Hidden Game. So if you’ve read that, you’re not going to learn anything, and if you are new to the field, you would be better served by reading the original.

The part of the book in which Felber evaluates front office decision making has a number of annoying parts to it, like the use of linear regression to estimate W% from payroll. It is never specified whether he is using all the data in his regression or doing a separate regression for each year; whether he uses the absolute dollar figure or compares it to the league average; and most incredibly, what the formula is in case you wanted to use it yourself.

Then a look at what a player “should” be paid, based on TPR (modified to be above replacement), ensues. It does not consider such things as the players’ free agency/arbitration status--this is forgivable to me, because you can still attempt to make a general statement that x RAR equal y dollars or something. But he assumes that salaries should be distributed based on the linear difference between their TPR an the position average TPR. Now maybe they should be, but to evaluate contracts without studying the reality of how the market actually works is a little silly. Furthermore, it appears as if he calculates these values specifically for every season. This could cause a huge change, like ARod getting $25 million/year, to totally skew the evaluation of dollar values of the other players at his position.

There is a look at park factors which points out that they are not very stable from year-to-year and discusses some possible explanations for this…without really touching on the most likely culprit, random chance. There is a bizarre digression on how if Sammy Sosa would catch a flyball with two hands instead of one he might be able to save fifteen runs. He criticizes Win Shares because they compare to a low standard and because they don’t include Loss Shares. I agree with this; however, it seems as if Felber paints the choice as between the WS baseline and the average baseline, when of course there are choices somewhere in the middle that most sabermetricians make.

Even more bizarre then the Sosa discussion is one on switch-hitting, in which he claims that even the 1999 Ken Griffey would have benefited from “judicious platooning”. He shows that Griffey had +57 Batting Runs, but was +55.43 from the right side, leaving him around +2 from the left. He remarkably defines a replacement level player as “[one] who would have hit 5 percent below the average performance level for the league”, and then claims that this replacement level player would have been +4 runs in Griffey’s left-handed PA! This leads to the ridiculous (and incorrect, as somebody 5% below average can’t have positive LW) conclusion that Griffey could have been platooned with some bum versus lefties and the Mariners would have been better off.

On the positive side, I think Felber is a talented writer. I don’t want you to get the impression that I think he’s an idiot--I just don’t think he is qualified to write a book of this sort. Some of the studies with selective sampling issues and the like would be ok as blog entries or something, to generate discussion or provoke thought, but they don’t make for a good book. I cannot recommend The Book on the Book.

1 comment:

  1. For saber newbies, The Book is probably a pretty good read. But for people who are pretty much up to date, there is really little or no reason to read The Book.

    ReplyDelete

Comments are moderated, so there will be a lag between your post and it actually appearing. I reserve the right to reject any comment for any reason.