So we all know that Cole Hamels' 2009 season was basically identical to his 2009 2008 season in nearly every statistical category that matters: K/9, K/BB, HR/9, etc. And we all know that a spike in BABIP (and thus, bad luck) seems like the most likely culprit in explaining why his ERA and W-L record took such a dive from 2008 to 2009.
But I was eyeballing Hamels' stats a little while ago and something else jumped out at me. In 2008, Hamels averaged 103.7 pitches and 6.89 innings per game. In 2009, he averaged 97.4 pitches and 6.05 innings per game. The differences in both categories strikes me as being statistically significant.
Now there are at least two possible explanations for this. The first, obviously, is that managers tend to pull pitchers after they give up a certain X number of runs or a certain number of Y baserunners. If you're unlucky with BABIP, you'll get to X or Y faster than you otherwise would, meaning that your manager will pull you sooner than he otherwise would. So if this theory is correct, it wouldn't affect the analysis at all. The decrease in pitches/game and innings/game is only a result of the spike in BABIP, just like the increase in Hamels' ERA is.
But here's the second possible explanation. Managers also tends to pull pitchers when their expected effectiveness to the next batter (which, by nature, declines from the middle innings onward) passes below a certain threshold Z. Could it be that Hamels lost his effectiveness at earlier stages of games in 2009 than he did in 2008? This wouldn't necessarily be inconsistent with the fact that his overall K/9, K/BB, HR/9, etc. remained constant. Maybe his "peak effectiveness" improved, but his stamina decreased, thus causing his peripheral averages to even out.
(Or maybe he just gave up a lot more foul balls than he did before. As a lot of people have noted, it sure seemed like this was the case. If so, did those foul balls come at the expense of swings-and-misses, balls in play, or something else? But I digress.)
Generally, if a player consistently performs better in one game situation than in another game situation, his averages are still what matters. I know there's no such thing as an ability to hit with runners on base, but let's pretend that there was. Even then, stats with runners on base still wouldn't matter much (except insofar as they're part of your overall stats). If you're able to post a certain overall triple-slash and your stats with runners-on stink, then by definition your stats with bases empty are very good, meaning that although you're disproportionately bad at getting RBIs yourself, you're also creating a disproportionate number of RBI opportunities for your teammates.
However, it seems to me, at least intuitively, that it wouldn't work the same way if a starting pitcher's performance varies because of stamina issues, even if your overall stats remain constant. That is, it seems to me that if you have a tendency to concentrate your "badness" within a group of consecutive at-bats, that will cause you to give up a total number of runs that is disproportionate to what your overall stats might suggest.
I pulled up Hamels' splits on B-R, and there are some noticeable differences in how he performed in 2008 and 2009 when he was between Pitches 75-100. The biggest difference is still BABIP (an incredibly low .200 in 2008 versus .357 in 2009). But there were other differences too. In 2008, he had 4.59 PA/K and 22.44 PA/BB. In 2009, he had 5.13 PA/K and 13.67 PA/BB. (On the other hand, his HR rate actually improved.)
I don't know if I actually believe any of this explains anything, but thought it might be worth throwing out there, if, for no other reason, to send out the MattS Batsignal. (And if he already talked about this in his article, then sorry about that. I'll get around to subscribing to BP someday....)