I enjoyed reading Jeff Sackmann’s recent columns on the relative strength of rotation spots (here and here). Sackmann divides up each team’s 2006 starting pitching output into 5 slots and reports the results, which show that the prevailing concept of a #1 starter, #2 starter, and so forth is strongly distorted. While this probably doesn’t come as a surprise to many sabertypes (I am reminded of debate in the Rich Lederer-Kevin Goldstein thread over whether “#1 starter” refers to a Top 30 starter or only to a group of pitchers numbering not more than a dozen) or anybody who has noticed that “#4” is oft-used to refer to someone with a league average ERA (and thus an above-average starter), it was nice to see an attempt to quantify it.
However, I think that Jeff fails to note some of the problems with the way he presents his findings, even if I suspect he is well aware of them. First, there is a semantic question: while in practice injuries typically force replacement-level (or sometimes worse) pitchers into the rotation, I think it’s still somewhat fair to refer to a 1-5 schema for rotation slots that assumes total health. Jeff’s model necessarily is disconnected from the reality of rotation slot discourse because he only goes from 1 to 5, when in reality fans, managers, and front offices often refer to “sixth starters” and so forth. The rotation slot talk is linked to talent, not performance, and so the need of teams to fortify their rotations with sub-5’s should be accounted for. I think the rotation slots discourse would, if it weren’t so artificially inflated, refer to pitchers in the top 30, top 60, top 90, etc. *prior to the talent pool being diluted by injury*. Obviously, some degree of injury needs to be included in such a model, but it should not be all, as Sackmann uses, or even the majority, I would say.
Second, these ERA’s are selectively sampled, plain and simple. If you went by DIPS numbers, for instance, you would find true 3’s labeled as both 1’s and 5’s, for instance, in Jeff’s sample. That does not mean that what Jeff did was not worthwhile, but since the talk of a “#1 starter” is used, as far as I can tell, mostly as an index of talent rather than performance, it is misleading in this context. If you want to gauge the talent level of a #1, the best way to do so is probably to take the average projection of the top 30 starters, then repeat from 31-60 for #2, 61-90 for #3, and so forth. Alternately, you can label the top 160 pitchers as 1-5 based on their projection tier before the season and then follow up at the end with their actual numbers. The downside to doing it otherwise is that you end up being blown away by the wretched numbers put up by the back end, as Jeff does. Well, since Jeff is taking the worst starters on the team to compose that #5, that will occur by definition as long as ERA has a fairly wide distribution, especially since Jeff is taking blocks composed of different pitchers to make his sample. While a reminder here and there that lots of “bad” pitching occurs in any given season isn’t a bad thing, it’s only a revelation when fans and media have drenched their rhetoric in
While I haven’t endeavored to do an alternate study, I would suspect that a model to match my concept of 1-5 would set the tiers closer to 3.90, 4.30, 4.70, and 5.10 than Jeff's 3.87, 4.36, 4.84, and 5.67, and there would also be a cap on #5's at 5.50. Those are rough guesses, obviously, but I hope it communicates the general point that there can be a happy medium between the inflated concept of what a #2 etc. is on the one hand and including injury replacements and Joe Mays in the definitions on the other.