THE BOOK--Playing The Percentages In Baseball

Cool—I’ve always wondered what these numbers look like. 

Do you have this data linked to team performance in any way (W-L, RA/RS)?  If so, it would be interesting to see if there is any general tendency for winning teams to have a high or low percentage of pitchers.  Maybe look at this within time periods where overall pitcher% was fairly stable.  I wonder if we could say whether a 13th pitcher tends to be more or less valuable today, in general, than a 13th position player.

I can’t work it like that the way I set it up.  I first figured innings for each pitcher-season (meaning merging totals of players who played on multiple teams).  And then counting number of pitchers above that year’s threshold.

To do what you want, I’d have to assign each pitcher to one team.  It’s not hard, but I’m not setup for it (quickly enough anyway).

Maybe someone else can accomodate…

Teams have also been slowly increasing the number of players, as measured under this method, which is interesting in its own right but also suggests that these defintions may be obscuring something.

The issue is roster construction, that is, teams need to find someone to pitch 162 * 9 innings and someone to take 162 * 27 / (1 - lgOBA) PAs.  There are somewhat more of both due to extra innings, but I’ve ignored that for now.  The issue is how many individuals to use to fill these needs.  I’d start by looking at the distributions for all players in terms of PAs and IPs, then try to characterize those distributions, which are not normal, and then track how they’ve changed over time.  Note the number of PAs to be pitched and taken have changed since the number of games changed and the lgOBA has also fluctuated over time, so characterizing the changes needs to be done somewhat carefully.  Then, somehow, that needs to be turned into a description of roster construction.  I think this is more complicated than the initial analysis lays out.  I’m sure this isn’t news, as this is just a first cut at the question. 

I note that the run environment is down this year, but teams are still using the roster construction of the high run environment of the last two decades.  Actually, using an even more pitcher heavy roster construction.  That seems like either an inefficiency to be exploited by someone, or evidence that the past included rosters that had too little pitcher usage, or evidence that in the past teams had players on the roster who went largely unused and this inefficiency has been removed recently.

Anyway, that’s a lot more than I intended to write initially.  Very interesting issue.

I’d love to see this broken down by league, to look for DH effects.

This makes me wonder if there is any way to know how to optimize your roster. Maybe you should just be trying to maximize WAR (within the constraints that LarryInLA mentions)? e.g. figuring out how many Runs Above Replacement you expect out of the last pitcher on the roster vs. the last hitter, given what you perceive their talent level to be, and how much they are expected to play?

I have no data to back it up, but I’ve always felt that the 3rd C that some managers like to keep on the active roster is a waste, much in the same way some NFL teams have eschewed keeping a 3rd QB on their 53-man roster.