I ran some Buchholz splits in his last game thread. They were so strange that I decided to break them down further. They got much stranger.
This is Buchholz's slash line allowed the first time through the order:
.469 / .553 / .625 (38 PA), bases empty
.563 / .563 / .938 (16 PA), runner on 1B only
.087 / .222 / .261 (27 PA), RISP
And subsequently:
.189 / .294 / .284 (85 PA), bases empty
.094 / .171 / .219 (35 PA), runner on 1B only
.429 / .458 / 1.000 (24 PA), RISP
It doesn't appear to be a stretch mechanics issue, because his results with a runner on 1B match the results with the bases empty. In fact, let's lump those together:
.500 / .556 / .729 (54 PA), 1st time through, non-RISP
.087 / .222 / .261 (27 PA), 1st time through, RISP
.160 / .258 / .264 (120 PA), subsequently, non-RISP
.429 / .458 / 1.000 (24 PA), subsequently, RISP
He's been
absolutely dominant exactly as we know he has usually been in his career, in a well-defined majority of his BFP. You have to take those 120 PA as an indication that he still has the talent. (Unless you're jmei.) We've seen him do it: get into cruise control, then get into a bit of trouble, and then make one terrible pitch. Every time.
We know that pitchers will change their pitch mix and their approach with RISP. His overall numbers the first time through are so terrible that it's clear he has very little that's working for him. That he's been so good with RISP is almost certainly in part a SSS fluke, but the split is so large that it has to have some real component (again, insert jemei's disagreement here, to save him the work).
What you'd do if you were getting killed early and you desperately needed outs is simplify things and just go with whatever one or two things seem to be working best, trading away any element of surprise and gambling you can make up for it with sheer execution. And going with the same one or two pitch / location combinations that seem to be working may actually improve your command of them further, through repetition.
After the first time through, he gets his mechanics down and is himself, yet in an even smaller sample with RISP he's been unimaginably bad. He should have confidence, and it appears as if the terrible results with RISP are the failure to execute pitches he and the catcher thinks he'll be able to. Alas, BrooksBaseball has no filters for base situation. But it wouldn't shock me to learn that once he's rolling, that they try riskier stuff with RISP and it blows up in their faces.
So there is a coherent explanation for those splits, which, again, strike me as way too large to be entirely random.
When he's struggling early and gets into trouble, he gambles on simplification and wins. After all, it's the first time hitters have seen him, so there's no reason not to just to go after guys with the one or two pitch / location combinations that are working at all.
When he's dominating later and gets into trouble, he gambles on diversification and loses. Hitters have faced him already, so he has less confidence in going to his bread and butter because he's sure hitters will be sitting on it. So they try something unexpected that usually works when he's dominating, like backdooring the two-seamer outside to a RHB, misses over the plate, and gets killed.
He has to get his mechanics down to start the game (and when he can't quite, compensate by trying to figure out ASAP what seems to be working best), and he either has to tweak his RISP pitch mix later, or get his head together in those situations, because there's no reason to be missing so consistently when he tries the fancy pitch / location combinations.
But the good news is that there's pretty good evidence that his stuff is as good as always, which matches what we've seen with our eyes. And this is about as late in the season as his bad starts have lasted.