Charlie's decision to stick with Joe Blanton in Game 4 tonight is proving to be somewhat controversial, just as Girardi's decision to use A.J. Burnett last night in the ALCS was. Evaluating this decision defies easy quantification, but here's a very brief, back-of-the-envelope analysis for why Charlie's decision is not only defensible but probably correct.
Here's a table of each of our four starters' xFIPs for: (1) the full season, (2) the last three months*, and (3) the average of the first two columns. The reason I calculated an average is to try to give greater weight to more recent performance without wholly ignoring what happened earlier in the year. It's a crude method for doing that, but whatever.
| Pitcher | Full Season | Last 3 Mos. | Average xFIP |
| Halladay | 2.92 | 2.86 | 2.89 |
| Oswalt | 3.45 | 3.40 | 3.43 |
| Hamels | 3.43 | 3.12 | 3.28 |
| Blanton | 4.06 | 3.52 | 3.79 |
[* I would have used post-ASB numbers except that Fangraphs doesn't maintain that split. So instead I had to manually calculate a weighted average. These figures might be slightly off, but they should at least be very close.]
Here were Charlie's two options (barring rainouts). "FR" stands for "Full Rest" and "SR" stands for "Short Rest."
| Game | Option 1 | Option 2 |
| 4 | Blanton (FR) | Halladay (SR) |
| 5 | Halladay (FR) | Oswalt (SR) |
| 6 | Oswalt (FR) | Hamels (SR) |
| 7 | Hamels (FR) | Halladay (SR x 2) |
Rearranging these options for ease of analysis gives you this:
Halladay (FR) (Game 5) > Halladay (SR) (Game 4) --> Difference = A
Oswalt (FR) (Game 6) > Oswalt (SR) (Game 5) --> Difference = B
Hamels (FR) (Game 7) > Hamels (SR) (Game 6) --> Difference = C
Blanton (FR) (Game 4) < Halladay (SR x 2) (Game 7) --> Difference = D
Question: Is D greater than or less than the sum of (A + B + C)?
Blanton's weighted xFIP from the last column in the first table is 3.79, while Halladay's is 2.89. So if both were at full rest, you might expect an advantage for Halladay of around 0.90 points. But Halladay will be on short rest times two, plus you're lowering your odds in the other three games by having all of them be on short rest as well. You could consider that to be 5.0 units of short rest.
So in the final analysis, what we need to know is: How much should you expect a pitcher to lose, on average, when he goes on short rest? Is it greater than or less than 0.90 divided by 5.0, which equals 0.18?
I doubt that anyone knows the precise answer to this question, but if you had to guess, I think it would be more than reasonable to guess that a pitcher, on average, will lose more than 0.18 by going on short rest.
I recognize that this is a very simplified analysis with certain non-proven assumptions, and I can foresee what some of the objections will be (some worthy, some not), but I don't have time to write more at the moment. But hopefully we can hash that out in the comments.