clock menu more-arrow no yes

Filed under:

Matchup/Game Thread

New, 9 comments
3 - 2, 4.91 2 - 4, 3.29

Astros pitching hasn't given up a run since the third inning of Roger Clemens' victory over the Giants, a span of 24 innings.

Andy Pettitte looks to extend that streak tonight vs. the Giants, which got me thinking, always a dangerous thing.

Somebody on the air was saying that Pettitte had been "remarkably consistent," so I looked to see how remarkable he in fact had been.

In eight starts, which amounted to two wins, 4 losses and 2 no-decisions, Pettitte has given up three earned runs in a start four times, two earned runs in a start three times, and one earned run in a start one time. So Pettitte has given up an average of 2-3/8 runs per start with a standard deviation of .70. 87.5% of his starts fall within one standard deviation of the mean, and in a normal distribution, 68% do, so one must come to the conclusion that Pettitte in fact has been extraordinarily consistent when it comes to giving up runs per start.

But of course, ignored in the above is the length of his starts--as well as his strikeouts, walks, hits, and just about everything else but the number of men to touch home plate. Fortunately, Bill James came up with his game score metric as a convenient way to kind of roll all those messy numbers up into one easy-tto-use überstat.

Running through Pettitte's game scores, you see a little more variance. His average game score over his eight starts has been 53.5, and this is with a standard deviation of 7.45.

No doubt about it, Pettitte's game scores are much more inconsistent. Only 62-½% of his game scores lie within one standard deviation, which is actually less than what is predicted in a normal distribution.

Which means that Pettitte overall, when you figure in all his numbers, has been a little on the inconsistent side.

Just goes to show you.

UPDATE: As this game has gotten more and more out of hand, I've played with the numbers concerning Pettitte a little more:

Runs Hits BB GSC IP K
Avg 2.375 6.375 1.500 53.500 6.500 3.750
SD 0.696 1.555 1.118 7.450 0.727 1.785
% 87.5% 37.5% 50% 62% 75% 75%
% - Percentage of datapoints within 1 Standard Deviation

This all looks very good for data that are now rendered hopelessly obsolete by Pettitte's crappy start, one that no-one could possibly call "consistent."