Custom Linear Weights for Pitchers? (August 1, 2003)
I was asked about how Custom Linear Weights works for pitchers. I try to explain in 2 minutes all about Linear Weights and Run Expectancy. There's no question that Pedro's LWTS values are lower across the board (except for the HR).
A run value of an event is based on:
a - chance of scoring due to getting on
b - extra chance of scoring due to moving over
Now, I'll guess that while the typical single will score 27% of the time, a single with Pedro on the mound will score under 20% of the time.
Now, how much run impact would a single add to the runners already on base? To either pitcher, as it turns out, the average single will add about about .32 runs *to the runners already on base*. HOWEVER, there are far fewer runners on base when Pedro is on the mound. So, while an average pitcher might have about .60 to .65 runners on base per PA, Pedro is probably at .50 runners on base per PA.
Working it out:
single (avg pitcher) = .27 + .63 x .32 = .47
single (Pedro) = .20 + .5 x .32 = .36
The interesting one is the HR.
HR (avg pitcher) = 1.00 + .63 x .65 = 1.41
HR (pedro) = 1.00 + .50 x .75 = 1.38
Virtually the same. Why does a HR add .75 runs to the runners already on base with Pedro, but adds only .65 runs to an average pitcher? Because, with an average pitcher, those runners already have a better chance of scoring. Therefore, any extra push from the batter will add very little run potential, relative to Pedro. (The above can be confirmed through a simulation or Markov analysis.)
For this reason, above all others, is why the run value of the HR must be treated separate than all other events. And, for this reason, every run estimation formula out there is inherently wrong, and inferior to BaseRuns.
--posted by TangoTiger at 11:00 AM EDT