A Closer Look at September Numbers

My impression is that the issue is overblown because 1) Most September callup pitchers don’t actually get all that many innings. 2) September callups generally aren’t taking innings away from elite pitchers, they’re taking innings away from guys who are marginally better than they are.

For instance, last year the Pirates promoted 4 pitchers after the rosters expanded on 1 Sept: Brad Lincoln, Steven Jackson, Brian Bass and Justin Thomas. Here’s what they did:

Lincoln: 2 IP, 9.00 ERA
Bass: 1 IP, 9.00 ERA
Jackson: 4.1 IP, 4.15 ERA
Thomas: 2.1 IP, 3.86 ERA

So on the 2010 Pirates, a team out of contention, September callups accounted for a little less than 3% of the teams’s innings pitched in Sep/Oct. Less than one complete game’s worth. Some other teams might use their callups a bit more, but I doubt that anyone is using September callups for more than 10% of the team’s innings. And again, a callup reliever like Steven Jackson isn’t going to take inings away from a really good reliever like Hanrahan or Meek, he’s most likely going to take innings away a guy who’s just a bit better than he is, like Karstens or someone like that.

There’s a probably an exhaustive study on the “September effect” in the THT archives or someplace like that if anyone is inclined to look.

I think I added this to the other thread at about the time Woz put this up. May as well include it in this part of the discussion:

Look at all Qualified hitters, look at their September numbers.

I don’t have time to do any spreadsheetery, but here are the wOBA for each quartile, full season first:
          Full Season September
99 %ile .447 .468
75 %ile .366 .375
50 %ile .340 .332
25 %ile .316 .293
1 %ile .248° .214°°

So what do you see? Exactly what you’d expect between any larger and smaller sample size – with a larger n, you get regression towards the mean, such that 50% of qualifying MLB hitters hit between .316 and .366 for the season, while the middle 50% in September ranged from .293 to .375 – an extra .033 of spread. But, if anything, the hitters were worse in September as a group. Maybe that’s fatigue, maybe that’s because I only looked at one year, but it’s really, really hard to argue that MLB regulars see a reliable and significant bump to their hitting numbers from facing AAAA pitching in September.

[Note: the samples are not identical; there are 11 more qualifying hitters for Sept than for the season as a whole, which reflects injuries, rested vets, and mid- and late-season callups. But we’re talking 149 guys in the full-season sample, the vast majority of whom would have had to qualify in Sept. to qualify for 2010 as a whole. I’m confident that these are largely the same players]

° Iszturis, worst in the league by .020
°° there’s actually a guy at .165, but he’s a really big outlier in a SSS, so hell with him°°°
°°° Aha! I’ve finally figured out a method for using asterisked notes that the software doesn’t translate into bullets.

 about just being the end of the season in general. I don’t think teams way out of it have much to play for. I don’t know how this changes things since there is no way to prove it even does but I’d have to say it affects the mental/psychological approaches of players/managers.

Good post. I think the what I take from this is that September stats should be taken in contect when that is all you have to go on. For instance, they need to be contexualized when you are looking at a player called up in September when there is no other track record to go on. Since Pedro was not a Sept call up looking at just his Sept stats is less meaningful since there is a larger body of work to analyze.

If there’s a rigorous analysis out there, I’m certainly open to changing my conclusions. But this smells to me like a nugget of old-fashioned baseball wisdom that has somehow slipped into sabermetric convention without close examination. It makes intuitive sense, but that should really be a red flag for the statistically-minded, not a recommendation.