when he polls the fans, is it before or after he releases his metrics? there’s some arguments for bias if he’s more likely to poll fans who pay attention to defensive stats than not.

If I’m understanding you correctly, Tango’s poll basically asks fans to rate and/or rank defensive players. Since any given defender is only involved in a handful of non-routine plays per game, there’s not that much to keep track of – the main question is whether fans are observant enough not to be fooled by flashy, but actually mediocre, defenders. In contrast, every game involves a couple hundred pitches in myriad combinations – an observant fan might notice something useful about high-profile PAs or count-situations (like the efficacy of Lincecum’s changeup as an out pitch), but to glean marginal effects over every one of those pitches? It just doesn’t seem plausible, especially because the fan then has to do the same thing 161 more times just to have a sense of how 1/15 of the MLB season went.

It’s a good argument that there aren’t gross imbalances, but no one says there are – even the original claim is on the order of a few percent – maybe 2 pitches per team per game

Anyway, why wouldn’t this argument also work to say that the sac bunt has always been used optimally? After all, it’s easy for even a casual fan to keep track of whether a player on first is more likely to score if he’s bunted over than if the subsequent batters swing away.

Now I understand what they were getting at with that part of the thread.

That said, I’m still not sure that the claim is prima facie dubious (well, maybe the scale is), because of the way pitch selection is done (at least the way announcers assure me it’s done; what do I know?). The following aspects strike me as red flags of a situation in which the parties may not be optimizing:

- generally catchers make pitch selections, with input from pitchers. Neither one has access to data as they make selections, so they will be combining their recollection of scouting reports plus their own experiences with heuristics

- both pitchers and catchers are prone to be overly influenced by anecdote – “this guy beat me on a FB last time, I won’t let him do it again.” Retired pitchers talk about remembering the pitch for every HR they gave up – which is impressive, but ultimately useless, as the sequence largely determines outcome (outside of hanging curves and the like).

- competitive psychology pushes in predictable directions – “if he’s going to beat me, it’s going to be on my best pitch.” That’s not to say this is a terrible idea – everyone in the park knows Mariano’s bringing the cutter, yet he still usually succeeds – but that it’s an emotional, not a rational, decision

- near as I can tell, the basics of how to sequence pitches are unchanged since the end of the Deadball Era. As I said above, baseball is prone to being hidebound, and it strikes me as more likely that players and coaches are reluctant to change than that pitching strategy was perfected 90 years ago

I agree that existing practices are probably quite close to optimal; I just see reasons to doubt that they’re exactly optimal, and wouldn’t be surprised to see an analysis similar to this one reveal inefficiencies.