The argument against this is that with varying off-days and rainouts, plus other reasons for disturbing the rotation, that this would not hold true, that it would vary enough that my theory would turn out to be wrong. I understand these factors, but from my fantasy league experience, I find that varying off-days might change one or another, but then they fall back into sync again. Rainouts are more problematic, but the Giants typically don't have a lot of those to deal with.
What I present is not definitive, but certainly indicative that my supposition has some validity. Here are the 2008 Giants starters run support, which I'm ordering them from 1 (ace) to 5 (back of the rotation), based on who went 1 to 5 to start the season:
- Zito: 3.60 runs per 9 innings support
- Cain: 3.12 runs
- Lincecum: 4.57 runs
- Sanchez: 4.79 runs
- Correia: 3.96 runs
For 2007 Giants, from 1 to 5:
- Zito: 4.21 runs
- Cain: 3.12 runs
- Morris: 4.99 runs
- Lowry: 4.70 runs
- Ortiz: 5.45 runs
- Lincecum: 4.54 runs (took over Ortiz's spot in rotation; average of the two is 4.67 runs)
For 2006 Giants, from 1 to 5:
- Schmidt: 4.38 runs
- Morris: 4.11 runs
- Lowry: 5.40 runs
- Cain: 4.67 runs
- Wright: 4.67 runs
For the 2006-2008 Giants, from 1 to 5, average of 3 years:
- #1: 4.06 runs support
- #2: 3.46 runs
- #3: 4.99 runs
- #4: 4.72 runs
- #5: 4.44 runs
I should note here that the setup of the rotation for the Giants rarely did not change when it came time to readjust the rotation after the All-Star break. Still, as one can see, generally, the front of the rotation gets much less support from the offense because they are facing the front of the rotation for other teams. And the mid to back of the rotation gets much more, over 1.5 runs more from #2 to #3 and, at the closest, over 0.38 runs from #1 to #5.
The odd thing, obviously, is that runs scored is greater in the middle than at the back end. I attribute part of that to the fact that teams' off days are not always in sync, thus pushing the ace to pitch against the #5 sometimes, bringing down the average run support for the #5 starter. Likewise, sometimes the #5 would get pushed to face the other team's #4.
Still, it is odd that the highest run support was usually in the #3 spot (Lincecum fell short in 2008, but remember, he got moved to the #1 spot in the second half, which reduced his run support). And that the #4 spot was still lower than the #3 spot.
A wild guess would be that the truly worse starter is typically #3 or #4 because teams like to reward the vets senority and pitch them higher in the rotation than they should be, had expected performance been used to drive the selection instead. Then age decline or expected performance happens and more runs are given up in the middle, while the new young starters typically get thrown at the back of the rotation, where they sometimes kick butt, pushing the runs scored lower. But if they do poorly, then they are like the other pitchers in the back of the rotation. Thus the upside potential at the back end of the rotation is greater than the expected averageness (but low chance of upside) that is normally in the middle of the rotation.
Still, not conclusive, but the general sizing is that the back of the rotation gets more runs support than the front, and the only plausible explanation of that is that rotations generally keep the same general order over at least the first half of the season, which is demarked by the All-Star game, which generally happens around the 95th game of the season, not the 81st, or roughly 19 starts in the first half, 13 starts in the second half, which is when rotations often change, but typically the ace is still up top, and generally the #2 is still #2.
I think my idea not only has legs, but good support by the data, albeit only one team over only 3 seasons, but the general shape held well across the time span, each starting position generally held the same rank: #1 got more runs than #2, but much less than #3, #4, #5; and #3 was greater than #4, which was greater than #5, for the most part.