There’s been some talk about the ideal auction budget mix on twitter recently. Perhaps, if you click this link, you can see some of the conversation between Chris Liss, Mike Gianella, Steve Gardner, Peter Kreutzer, and Jeff Erickson upon which I was snooping. It’s a fairly complicated conversation, and far-reaching. Let’s jump in.
The crux of the argument was that your split between pitcher and hitter values matters a lot to your auction values. The aggregate split is around 70/30 right now, but that’s just established market truth — that does not mean that the market is using the most ideal split right now.
You can look backwards and try to derive it empirically. For example, our retro-active auction values that we are going through (you can look backwards position by position by using the positional rankings links on your right). If you sum up the above-replacement hitters, you get $1677.40, and the pitchers give you $1164.60. That’s a 59/41 split! That’s almost definitely because our calculations assumed 13 hitter lineup spots to nine pitcher ones, or a 59/41 split on the roster.
But if you think that pitchers are more volatile than hitters, then you’ll want to push that 59/41 split further towards the hitters. You’d rather draft safer hitters. We know, for example, that pitchers get hurt more often than hitters (by decimal points) and stay hurt longer (by many days). That sort of volatility is easily handled by projections — just project your pitchers for fewer innings to reflect the idea that any starting pitcher that pitched last year is 40% likely to hit the disabled list this year. So that’s a way to handle volatility.
We can try to look at the year-to-year volatility in the position. For example, how many top-100 pre-season hitters showed up in the end-of-season top 100? 33/68 (48.5%) if you use ADP and our end-of-season values. 16/31 (51.6%) pitchers. 14/25 (56%) starting pitchers. Hold on now. Perhaps the top pitchers aren’t as volatile as we assume.
But what if pitching stats themselves are less reliable? We know they are — Dan Szymborski confirmed today — but how much more volatile are they? Much more. Look at this post evaluating the projection systems in 2011 — the correlation on hitters projections to their results was on the order of .62, and not a single projection system did better than .46 for pitchers. So clearly there’s a reason to distrust pitching projections a bit more than hitter projections, as many advancements as we have made.
Perhaps if you think the projections for the best pitchers are more stable from year to year, you can do something more like a 60/40 split at the top of your pitcher values. There’s some evidence for that — part of the problem with projecting pitchers is that innings-pitched numbers are more volatile year to year than plate appearances. The top pitchers may have more stable year-to-year IP totals, perhaps because they can move down the pecking order at their position and still pitch — the second-best second baseman on a team gets a huge hit in PA, but the second-best starting pitcher pitches almost as much as the first-best (provided both are healthy).
But if you look at pitchers as a whole, it does seem to make sense to push the needle towards 70/30, as we have all done. How far, exactly, might take a little more research.
With a phone full of pictures of pitchers' fingers, strange beers, and his two toddler sons, Eno Sarris can be found at the ballpark or a brewery most days. Read him here, writing about the A's or Giants at The Athletic, or about beer at October. Follow him on Twitter @enosarris if you can handle the sandwiches and inanity.