Sometimes a player makes an adjustment, or suffers an injury, or recovers from an injury, or undergoes or gets over a personal problem, at some specific point in the season (which usually has to be inferred from the numbers) that then renders everything previous irrelevant. Johnny Damon’s second half of 2002 and first half of 2003 (divorce) had no predictive value, ditto Bill Mueller’s first 6 weeks of 2002 (ST injury), big chunks of David Ortiz’s tenure with the Twins (misc. injuries), and Dustin Pedroia’s first career 166 PA. In every case, predictions would have been better had the clearly aberrant season chunks been tossed out (and in all these cases I did exactly that; in Ortiz’s case it was verifying that his ’03 reason was for real sometime in mid-summer, when sabermetric wisdom had him as just fluky hot) .

In almost every case you can run simple significance tests (e.g., chi-square) to confirm that the odds of the observed before-and-after split are very small. Lester, for instance, had an improvement in HR / Contact that you’d expect to see in 2.2% of random simulations *plus* an improvement in BB rate you’d see in 7.6% plus an improvement in K rate you’d see in 17.9% (and even in improvement in BABIP you’d see 30% of the time). The odds against all of these things happening together might be something like 11,000 to 1.

]]>Lester, average game score: 54.76

Matsuzaka, average game score: 56.79

Lester, standard deviation of game score: 17.76

Matsuzaka, standard deviation of game score: 15.72

Matsuzaka has the higher mean and the lower variation in terms of Game Score. GS isn’t perfect, and doesn’t account Dice-K’s lucky BABIP, so I’m not concluding that Dice-K is better than Lester, but it is tough to argue that Lester’s been more consistent, results wise.

]]>Lester’s had 7 starts with a Game Score of less than 40 this year – that’s one in every five starts where he could charitably be described as lousy. He’s also had 11 really good starts (his best 11 performances have an average game score of 74, which is excellent), and then everything else falls in the decent-but-not-spectacular category.

7 lousy starts, 15 okay ones, and 11 excellent ones. To me, that’s a guy who is occasionally excellent but still frustratingly inconsistent.

]]>