At SaberSeminar a few years ago, Farrell made the point that middle relievers show enough variation year-to-year that it makes sense to over-pay mid-season for a reliever having a good year.
The flip-side of that would seem to be not paying for a player at a "position" that shows such high year-to-year variation in the off-season before you know what kind of season the guy is having.
I dunno if this is actually a factor in the decision making this off-season, but it's definitely something I would expect them to be considering. Hope, even.
I would be interested to see the analysis behind this. If there is so little correlation in things pitchers control from year to year, I’d be shocked if there’s more meaningful correlation from first half to second half.
Put another way, variation for a relief pitcher can be broken down into : variation in health, variation in stuff not related to health, and variation in luck. The part attributable to luck would be even greater in a smaller sample size. On the other hand, health would be less volatile the shorter you have to worry about staying healthy. Maybe those somewhat offset, but I’d bet luck is way more variable than health in half season samples.
So, how correlated is stuff and could that push the net toward half seasons? I would think stuff would account for most of the year to year correlation that exists, and maybe a little more correlation within the first and second half of a season than year to year. So it seems quite ambiguous. Would love to see the numbers.
But, also, there’s certainly more correlation in closers than in middle relievers too. That suggests, if the Farrell hypothesis of middle relief is correct, that correlation in stuff increases in the quality of that stuff. And that suggests the “3 closers” approach to building a bullpen is the correct one.
It also, I think, suggests that you shouldn’t skimp on the high end of your bullpen (assuming you don’t have a cheap relief ace in the first place) while overinvesting in your 4th and 5th starters. But I need to work that logic out out more when I have time.