The Collmenter Conundrum: Underutilized as Reliever, Overutilized as Full Starter

On Wednesday, Josh Collmenter had his longest outing of the season to date: four full innings, four strikeouts, and just three hits and one walk. Good, right? But that doesn’t mean he should be moved to the starting rotation, at least not as a normal starter. We’ve learned some lessons from Collmenter’s 35 career starts that stand for that proposition.

I’m not here to bash the D-backs pitching up and down, and the truth is that even if Collmenter-as-starter was a good idea, he can’t be a tonic for the entire staff. If he’s moved to the rotation, he’s replacing one guy. And after he lasted just 3.1 innings while giving up 3 runs, 3 walks and 6 hits, it’s safe to say that that guy would be Randall Delgado.

I’m not sure I agree with that — he probably deserves more of a chance, and that chance will only ever come this year, or for another team. And there’s some evidence to suggest that part of the problem is pitch calling. Delgado isn’t just a fastball/changeup guy. He’s got two fastballs (four seam, sinker) and a curveball as well — it’s not like there’s no room for pitch sequencing. Yet, a study by Robert Arthur of Baseball Prospectus showed that Randall Delgado was among the five worst starters in 2013 in terms of whether his individual pitches carried information about the next pitch. If it’s easier than it should be to guess what’s coming next, that’s not good. Completely randomizing Delgado’s pitch sequencing has at least a chance of helping.

Still, let’s take as a given that Delgado should be removed, if possible. Should his replacement be Collmenter? For my own part, my answer is yes, but a qualified yes. If the team is planning to have Collmenter throw 90-100 pitches as a starter and go a few times through opposing batting orders, it would be a terrible use of Collmenter and unlikely to work.

Why? The Times Through the Order Penalty (TTOP). Master of most things baseball Mitchel Lichtman (“MGL”) thoroughly examined the TTOP this last fall. He published his findings in terms of wOBA (which is hard for me to translate from my own data source), but the results are pretty simple: there is a smallish but significant penalty for pitchers as they face batters the second or (more so) third times on a particular day, and because MGL controlled for a number of factors, it seems pretty clear that the TTOP has to do with batters seeing more pitches and getting used to the pitcher.

One of MGL’s findings: “a pitcher’s past deviations from the league average, in terms of their TTO penalties from the first to the fourth times through the lineup, are not very predictive, much like BABIP.” Far be it for me to disagree with MGL, especially without an extensive study to back me up, but I think we should at least consider the possibility that some pitchers may tend toward higher than normal TTOPs. I hate to rely on anecdotal evidence, but when it matches theory, I feel a little more confident.

And here’s the theory: pitchers who succeed, in part, because they have a different look or pitch may have a higher than normal TTOP. A pitcher like Josh Collmenter may succeed, in part, because of the contrast to other pitchers. It stands to reason that, for batters, getting used to Collmenter’s unique look would matter more than it would against other pitchers. Said differently: maybe deception pitchers have higher than normal TTOPs.

And here’s the anecdotal evidence: Collmenter got absolutely shelled by opposing batters after they’d seen him once, back in his days as a starter. Here are his combined starter/reliever stats for times through the order:

And here’s Randall Delgado’s, in part just for comparison’s sake:

I’m not saying that Delgado’s stuff couldn’t play up in the bullpen — he’s got the repertoire to keep batters guessing, and he could use it all if he didn’t need to save something for later. But Collmenter? He’s pretty much a two-pitch pitcher: cutter and change. He uses both at the outset of outings. To continue having success at the beginning of outings, he’d have to continue to do that. Doesn’t that leave him exposed the second or (especially) third times through the order?

I think the Collmenter table above supports that proposition (or, at least, doesn’t disprove it). Collmenter is good once through the order. After that? Decent, but not so decent that he’d be a marked improvement over Delgado. So I’m just not sure it’s worth replacing some mediocre Delgado innings with more innings from Collmenter that are also likely to be mediocre.

Collmenter pitched 92 innings last year, good for second in baseball behind Anthony Swarzak (96 IP) and in front of Alfredo Simon (87.2 IP). That is very good usage of Collmenter. For 75.1 of those innings, Collmenter was facing batters the first time through the order. Just 14.2 innings were against the second time through an order. That’s really good deployment for Collmenter.

The only improvement over last season’s usage of Collmenter that I can think of would be using him 2-3 innings every third game. That would bring his innings total to around 125, and they should all be “good” innings. But I’m a big believer in a starter-by-committee approach (still working on my manifesto about that, it’s kind of a hobby within a hobby), so take that with a grain of salt.

Using Collmenter every third day would not really be possible, although the idea of a tandem of pitchers taking two spots in a rotation sounds really good to me. But given a likely TTOP issue, it would not be a good plan to use Collmenter as a 6-7 inning starter.

Instead, if the D-backs replace Delgado with Collmenter, they absolutely should be willing to let him be a 4-5 inning pitcher, and they should stand at the ready with a quick hook and a dedicated long reliever (probably Delgado). Sticking to that kind of plan religiously should also mean that Collmenter would be available halfway around the rotation; so if Collmenter was starting Game 1, he might be available to pitch an inning or so during Games 3 or 4. As I plan someday to explain, a staff of relievers could thrive even if the staff was just 10 pitchers. Having a single reliever-starter actually stresses the roster out a bit more, but still within acceptable limits.