He was instrumental in developing sabermetrics, but that's his "using numbers to answer questions when he can". His actual mathematical acumen is limited. In various places throughout the Abstracts he denies being a statistician, and says that he's only interested in mathematics as it helps him understand baseball better. You'll notice that his work involves very little in the way of actual statistics - there're very few regression analyses to be found. And a lot of his work had nothing to do with numbers at all: short biographies of notable players take up a significant chunk of several of the Abstracts.
Win Shares is James' uberstat, but it's not WAR before WAR. It's premised on dividing actual team wins between players. Philosophically, it's the opposite of WAR. The question James wants his stat to answer is "who deserves credit, and how much credit, for each team win?". The question that WAR tries to answer is "how many additional wins would you expect from a team, if this player were to join it?" Win Shares is thus much more context-dependent than is WAR. It provides an interesting record of what happened, but, analytically, it's much less useful. It doesn't tell you much that's of much use for analyzing a trade, or for predicting a player's performance, and so on.
|