Win Shares

Win Shares is Bill James' best effort to synthesize everything he knows about baseball into a single measure. A decade after the book's publication, Win Shares has become an accepted tool in sabermetric analysis; by that measure this volume is a success. Nonetheless, the book is a famously difficult read.

I originally read Win Shares just after the book was published, then studied it a year or so later when I adapted its framework for a minor league research project. I found that the practical application was really helpful to understanding how the pieces fit together (although, sadly, it didn't much help my project). This decade-later read benefits from familiarity, now, and from watching other folks apply Bill's methods. Nonetheless, show more this is a difficult book.

The first problem is the book's arrangement. The first dozen pages give some background and present a fairly unhelpful system overview. This is followed by nearly 90 pages of equations and explications, with occasional explanations; you've really got to take this section on faith if you're reading the book from front to rear. This is followed by 20 pages which attempt to explain, not always successfully, how the equations fit together, and the assumptions behind them. These pages also directly address some of the questions James heard from others when he'd discussed the project while he was developing the system. 150 pages of "random essays" follow: Some of these address issues raised during the system's development, but more just illustrate of the sorts of questions Win Shares can help a researcher study. Some of these studies are immensely interesting--in this way Bill's work is always worthwhile. Finally, 250 pages into the work, we get the long lists of organized and comparative career and season numbers which make this a useful reference work; these constitute nearly two thirds of the publication and are the actual product of the system.

The second issue is that the book's emphasis on Defense overwhelms the main message. In terms of the Win Shares project, Hitting and Pitching are treated as Solved Problems; those sections of the system are built around Runs Created and Component Earned Run Average. Since those methods were addressed in detail in earlier books, they are given only slight explication in this volume. Assessing Defense, on the other hand, is a new problem, so James gives it the bulk of his attention. Worse, for many readers, the system looks like a Rube Goldberg machine--or a house of cards. While it's clear that the apparently arbitrary numerical constructs described in this book make large concessions to deficiencies in the statistical record, the justifications underlying the calculations are largely built on research which James describes elsewhere in the work. Since they're initially presented without that context, accepting them on faith seems a lot like a leap of faith.

All this discussion of D overwhelms, in pages and evident intellectual effort, James' description of the overall structure of Win Shares proper. Win Shares is a team-based method for assessing player seasons, and its core is allocating the overall team success to the players who made up the team. Many readers likely lost that message in the relentless march of difficult detail. Personally, I think he should have relegated the defensive discussion to an appendix, and presented the framework without the distractions.

A third concern is that James doesn't convincingly articulate his basic argument about assessing defense. This matters because many analysts seem not to have absorbed his contention that defense is a team function, and individual defensive efforts make sense only within the context of their team's defense. I recognize progress on this front, but we're still far from really understanding how to measure this component of the game.

A fourth matter of concern is the decision to award three win shares for each team win. I fully understand Bill's reasons, but the number itself is peculiar enough to create a barrier to understanding.

A final difficulty, which is less important, is the constant comparison of Win Shares' defensive assessments with those generated by Pete Palmer's Linear Weights. Some comparison was clearly necessary, but Bill's not one for half-measures when he presents an argument. It's pretty clear that Bill was aware that this was problem.

All that said: This is an immensely influential book, and deservedly so. Would it was less to difficult to master, but the effort's worthwhile.








This review has also been published on a dabbler's journal.
 show less

I'm a fan of Bill James work with baseball stats, and for me Win Shares is where he enters the territory of jumping the shark. We all want our data in nice, neat bite sized packages to digest, which is what James attempts to achieve with Win Shares - one simple number that you can assign a player for each season played. The idea is novel, but the execution leaves much to be desired and requires the readers to make some leaps that just doesn't make a whole lot of sense and depends on data that I wouldn't classify as being 100% reliable. The fact that he's still searching for the answer is worth giving him kudos for, but Win Shares is a strikeout for me.