THE BOOK--Playing The Percentages In Baseball

I think there should be a noise bucket for non-batted ball PA events and sequencing that credits/debits the pitcher, while the noise on batted ball PA events gets credited to the fielders. Non-PA events are odd, and I guess they can be subject to interpretation. For example, I’ve been accustomed to crediting SB/CS to catchers, but of course the truth is that it should be split between pitchers and catchers.

I think it depends on the intended audience to some extent.  I work in the market research field where the question of what to do with the residual is also asked.  Purists generally say keep the noise bucket, but lay people are confused by that and would like the noise divided out in the same proportion as the non-noise buckets…

This is the discussion we have to have to resolve our feelings as to how we eventually want it handled in WAR.

So, if you want to attach conditions as to how you want to portion out the noise based on the use, that’s fine with me.

Lay out how you want things handled, and why.

I’d distribute the noise and also display a stat without the noise added. The difference gives you the noise (how much random variation impacted that player), and the whole number gives you a more realistic assigning of credit/blame for the season. My guess is that non-analysts would stick to the number with noise.

I’m a bit confused by the terminology used.

When you say “performance of team’s pitchers” it makes it sound like you’re talking about true talent with noise already included, i.e. performance = how they perform & the outcome, not how good they are.

“Impact of these opposing batters relative to all batters” on the other hand is ambiguous, to me. Can you clarify what you mean (for the dense among us)?

And if you meant for the first 5 to refer to actual outcomes, then I don’t get what the noise is for. For example, shouldn’t linear-weighted batting performance + sequencing = runs (no noise needed)?