Catcher ERA: The Sequel
added 3/16/2001 by Steve Cutchen
In the Astros Daily Forum, David in Jackson brought up something that I posted about in my very first Knuckleball... Can a catcher's performance really affect pitcher performance?
I think the effect of a catcher on a pitcher's ERA is the most overrated fallacy in the game. There is no evidence to support it that I know of.
My analysis generated a lot of discussion among the statheads on rec.sport.baseball back in February. I thought I'd take the occasion of David's post to collect my comments about these followup discussions and put them in this sequel.
As far as an analysis of the effect of catchers on pitcher's ERA, David is correct. It certainly is a difficult and imprecise analysis to make.
Craig Wright developed an adjusted catcher ERA in his classic book _The Diamond Appraised_. And he gave a lot of caveats as to how it should be applied...
Wright's work was the basis of my analysis. My assertion is that Mitch has (or had last year... I haven't seen him this year) flaws in his receiving technique that can cause borderline pitches to be balled. And losing close strike calls causes emotional pitchers like Lima or young, less than confident pitchers like Dotel or Holt or Miller, to then press and pitch poorly.
I used Wright's technique to analyze the difference between Meluskey and Eusebio and the difference was huge.
Wright's CERA stat is not the most state-of-the-art statistical technique developed for this type of analysis. Keith Woolner developed another statistic for Baseball Prospectus. His analysis is available online at
with a followup article at
Keith's conclusions are that he cannot extract the effect of a catcher on pitcher performance statistically.
During our discussion on rec.sport.baseball, Keith said that because his league-wide analysis cannot extract a statistical effect, that the physical evidence I perceived combined with the (admittedly imprecise) CERA data are not conclusive. There is no reason to conclude that the difference in CERA is anything but random noise.
I, on the other hand, still believe that the combination of physical evidence and numerical data, applied on an individual and not a league-wide basis, is valid. And that the effect I found might be real.
Keith acknowledged that Meluskey could possibly be abnormal enough that his analysis would lose him in the wash. He said:
The counterargument to what I just said is that there may exist a very small number of catchers who possess a set of skills so poor that they actually do perform poorly year to year, but are so rare that they don't show up in the statistical tests, even when you limit the data set to the very worst catchers. If you want to argue that Meluskey is virtually unique in having poor ball framing abilities that impair his pitchers' success, please go ahead, but I think the burden of proof lies with those making the claim.
This is well said and does indeed describe my position.
And unfortunately the burden of proof will not ever be conclusively made in this case because the framing abilities that are being analyzed are not deterministic. And neither are the hypothesized effects on the pitching staff, split as it might be into steady, non-emotional types and those that live and die on the turn of the umpire's call. Nor other subjective data on Meluskey's performance such as the Houston pitching staff's comments,
"Almost to a man, the pitchers disliked Mitch Meluskey and questioned his decision-making behind the plate."
I did not argue that Meluskey's pitch selection was suspect, just his receiving technique. I'm not privy to the plans for each hitter. But the Astros pitching staff certainly did. And they've repeated this indictment as this Spring has progressed. How much if this is real versus woofing to support Ausmus and the trade? Who knows? Certainly the pitchers say they feel more confident with Ausmus behind the plate, and just ask Hooton how important confidence is in his book...
So let me go on record... The result of my analysis is subjective. No doubt.
But subjective analysis is what makes baseball trades possible. Each side evaluates the trade by their own data, their own needs and their own perceptions.
And that is what makes talking about baseball trades fun.
If we truly did have precise tools we would simply plug all of the data into a Diamond Mind sim and not bother with the games... :-)
And I believe that using Wright's methods to compare Meluskey and Eusebio add rather than detract from the argument over the trade.
But I agree that it is not proof.
And that's fine.
It would be interesting to run Meluskey and Eusebio's 2000 data through Keith's model (as opposed to the Wright CERA model I used) to see how much of an outlier he is on a catcher-catcher by common pitcher basis. In Keith's paper, the data for 95-97 seem to top out with a difference to the mean of about 2.5 to 3 std. deviations. Is Meluskey's data for 2000 within this range of expected outliers? Or is it perhaps WAY out there? Especially for the pitchers I've classified as fragile... If I get around doing this analysis, I'll let you know.
That's my Knuckleball.
Try to hit it.