I spend a lot of my time looking through the various ways people get to this site. While this usually results in a bunch of Google searches and people randomly clicking the SBN sidebar, I occasionally find and outside source that links here. Twice now, I've been lead to forum posts where people have been complaining about my recent use of ERA to evaluate past pitcher performance.
If you've been reading this blog for any length of time, you'll know that when I want to evaluate how a pitcher will perform in the future, I will use things like the peripheral stats, ground ball to fly ball ratio, park factor, and so on. These are far greater indicators of future success than ERA. However, in looking at all these stats, sometimes we miss the forest for the trees. Pitching is not a contest to see who can compile the best strikeout to walk ratio, all that matters his how may runs you allow.
For example, Jorge Sosa had a 2.55 ERA in 2005 with terrible peripheral stats. If the Braves wanted to improve production out of that part of the rotation in 2006, they wouldn't need to do the simple task of finding someone with better peripheral stats than Sosa, they would need to find someone that was capable of posting a better than a 2.55 ERA. Of course, to find that player, you'd find the guy with the best peripheral stats, and you would realize that 2006 Jorge Sosa isn't any kind of substitution for 2005 Jorge Sosa.
Even if the 2007 Dodger rotation compiles much better peripherals than the 2006 Dodger staff, if they allow more runs, it's going to be a downgrade. In the end all that matters his putting runs on the board, and keeping your opponent from scoring, no matter how unlikely it is you'll be able to repeat that in the future.