Bill James tries to defend his overrating of Sandy Koufax, and accidentally sabotages it. James ranked Koufax tenth among pitchers in the new version of the Historical Baseball Abstract. This is really indefensible using current statistical methods. Koufax was a dominant pitcher for a short period, but his career numbers aren’t that impressive, and other pitchers have been just as dominant for just as long. One is Greg Maddux, who after all won four straight Cy Young Awards in the Nineties, who rates 14th. James himself rates Maddux as the best pitcher in the game four times in that decade (not always in his Cy Young years); Koufax rates as the best pitcher in baseball three times. Koufax probably rates some extra points because he pitched so many more innings in those seasons, but for his career he had more than a thousand innings less than Greg and a similar ERA. Greg is eight hundredths of a run higher, but (a) he has mostly pitched in hitter’s parks in his career, and in particular in his best seasons, while Dodger Stadium in Koufax’s prime was maybe the best pitcher’s park ever, and (b) many, many more runs were scored in the nineties and today than in the sixties. Maddux’s career ERA is 45% better than the league, Koufax’s 31% better. (This is by ERA+, and I know I’m simplifying.) That’s without adjusting for the parks.
It’s actually more extreme when you look at individual seasons. Koufax, in his best years, was about 90% better than the league, and only hit that level twice. Maddux was 173% better than the league in his best season, 159% better in another, and bettered Koufax’s best marks twice. Now, it’s a lot easier to be twice as good as a league with a 4.12 ERA than one with a 3.63 ERA (league averages during their careers) but it is very hard to see how Koufax was much more dominant than Maddux.
James uses his win shares system to try to prove that Koufax had more impact on pennant races than any other pitcher. And if you buy his methodology, Koufax ranks first among 20th century pitchers. But Maddux ranks second! Maddux has far better career stats, was as dominant or more dominant in his best years, and had a big impact on pennant races — James doesn’t say how much, but if he’s second to Koufax they’re probably pretty close. So how in the world can James justify rating Koufax four spots ahead of Maddux, or ahead of Maddux at all? He can’t, not and be at all consistent. James says he rates active players “as low as he possibly can”, but (leaving aside that there’s nothing Maddux can do at this stage to significantly hurt his career standing) he can’t possibly rate Maddux below Koufax, not with any intellectual vigor. So Maddux has to rate ahead of Koufax.
But then, what about Maddux’s near-contemporary, Roger Clemens, fourth on the list of big-time pennant-race pitchers? (Which will certainly surprise his admirers in the Boston media.) James has been saying for at least a decade that Clemens rates ahead of Maddux. I don’t know that I agree with that, but there’s no doubt in my mind that they’re very close; they’re probably right next to each other on the career list, or at most once removed. But James ranks Koufax ahead of Clemens, too, by one spot! So Clemens has to pass Koufax, too. He rates Three Finger Brown 20th, and Brown is on that pennant-race impact list as well; I don’t know if he should rate ahead of Koufax, but if Koufax gets a boost, so should Brown, who after all was the best player on the best team of all time.
Wrapping things up, I’m a big admirer of James’ work, but he’s way off base here. He’s doing something he used to warn against, allowing his personal memories and biases color his judgment. Fighting that is what sabermetrics is supposed to be all about.