>A statistic that is often given on TV and radio is the batter's average
>with men in scoring position. I think this is as good a ruler as any.
There are two serious problems with BA/SP as an indicator of RBI ability.
The first is sample size: in order to be reasonably sure that you've
captured a player's true ability plus or minus 10 points of BA, you need
to have seen 8000 at bats. If you're willing to settle for +/- 20 points,
2000 at bats will do. Over several seasons, you can get a pretty good idea
of a hitter's overall ability, because he'll get those 2000 at bats. But
it takes twice as long to accumulate PAs with runners on base, and 4 times
as long for runners in scoring position. By the time a player has amassed
2000 PA with RISP, he's not the same player anymore that he was when he
In an individual season, a player like Joe Carter might bat as many as 400
times with men on base. The standard deviation of his batting average over
those 400 PAs is more than 20 points; if he hits .278, you can't be confident
of anything beyond the fact that his true ability is probably somewhere
between .240 and .320. For other players, with fewer PA with men on base,
the problem is even worse. And this model (all PAs identical) will tend to
UNDERestimate the variance, making the true spread of possible "true" values
The second problem with situational batting average as a measure of RBI
ability is that there doesn't seem to be any *ability* to hit well with
men on base (or in scoring position) that is distinct from a general ability
to hit well. The evidence for this is in the distribution across all players
of the *difference* between overall BA and situational BA. In any particular
year, this distribution looks just like you would expect it to look if you
had selected the "situational" PAs by flipping coins, rather than by looking
at a particular kind of situation, with a slight overall shift for the league
as a whole. (This shift, up for RISP and down for LIPS, seems to indicate
that *pitching* is qualitatively different in those situations, but that
Furthermore, the people who are at the top end of the situational batting
lists one season are at the bottom end the next season, and vice versa.
There's no *stability* from year to year, the way there is for overall BA
or HR rate or slugging average or walk rate or any of the other primary
offensive stats. Again, this suggests that any apparent *ability* to hit
better (or worse) than your overall batting stats would suggest in particular
situations is an illusion induced by small samples.
The best approach for quantifying RBI ability is probably to keep track of
a given hitter's % of runners driven in from first (second, third) per PA
with runners on base, and HR rate. Of course, to interpret these numbers,
you'd also have to normalize by number of outs consumed in the process...
Again, though, there are going to be serious small-sample problems here.
A typical hitter might not bat with a runner on third more than 50-100
times in a season, so that the standard error in his RBI% from third base
will be on the order of 4% - 6%, which is large compared to the range of
actual player performances. What good does it do to know that (say) Fred
McGriff drives in (say) 27% of runners from third base, plus or minus 10%?
Professor of Story Problems | And he swatteth one of three.
Dept. of Industrial Engineering | --Darnell Coleridge,
University of Pittsburgh | Rime of the Youthful Griffey