I started to do this as a comment to Angelo's post, but I think there's enough here to warrant another full post on the subject.
My starting point is that (as I understand it) H2H differs most from roto in that margin of defeat in a category doesn't matter in H2H. You get 1 point for winning a category, and 0 for losing, so all you're trying to do is win 3 of 5 pitching and 3 of 5 hitting categories per week. Contrast that with the spread of possible outcomes in roto, where you get points based on where you stand against all other teams in the league over the course of the season, not just one team per week. In roto, if you focus too narrowly on a few categories, you tend to fall way behind in the others and only get 1-2 of the 10-12 points available in those categories. The binary nature of H2H scoring makes punting a category or two a much more viable option, since your downside is so much smaller.
This is all by way of a long intro to my thesis that starting pitching might not be the way to go in H2H. You'd be better off with a team of made up largely of good closers and middle relievers, dominating ERA, WHIP, and SV, and hoping to vulture enough W's to steal that category. If you lose K's by 15, who cares? It's just 1 point. But you'll maximize your chances to "win" the pitching side every week by taking those three other categories.
Pitching is of course wildly variable week-to-week, so Angelo might just be running in bad luck. But I think the ability to game the two rate stats while nailing down SV makes relievers proportionally more valuable. Thoughts (other than that we're probably overthinking this)?
Next post will be funny. Honest.