8/26/05

Another thing I think about in bed...

...is the inadequacy of the ERA (earned run average) statistic as a predictor of a pitcher's future performance. The obvious problem is that it doesn't measure baserunners, so if a pitcher strands alot of runners, he keeps his ERA low. But by the law of averages, allowing alot of baserunners is going to catch up with him. It's true that certain guys are good at "pitching out of jams", but in reality, pitching with men on base requires the same basic skills as pitching with the bases empty. In other words, with rare exceptions, a guy who allows a lot of baserunners is sooner or later going to be a guy who allows alot of runs. It's the getting hitters out that's the pure test of the pitcher's skill.

The easiest way to think about this is: Pitcher A gets 2 outs, then gives up a single and then a home run, then gets the third out. So he's charged with 2 earned runs. Pitcher B gets 2 outs, then gives up a home run and a single, then gets the third out. Although he performed identically to Pitcher A in terms of the aspect of his job that he can control (the pitcher-hitter matchup), he's charged with only 1 earned run, because he stranded a runner on base.

So the point is, a guy who, over the first half of the season, had allowed alot of baserunners but a low number of earned runs, because he had frequently pitched his way out of jams, would not be a guy who one would expect to be successful in the second half of the season.

So what you need is a stat that measures hits and walks, and also that further penalizes a pitcher for surrendering extra base hits (which generally result from an objectively worse pitch than do singles, making them a relevant indicator here). The best way to do this, I think, is a Bases Per Inning (BPI) stat, which charges a pitcher 1 point for giving up a single, an unintentional walk, or a hit by pitch, 2 points for a double, 3 for a triple, and 4 for a home run. (Forget about what the runners do once they reach base, whether or not they score, etc. Focus only on the pitcher vs. hitter contest.)

Then, rather than doing it per nine innings, like the ERA, divide that by innings pitched. Nine has become sort of arbitrary because pitchers rarely pitch a complete game any more. So a dominant pitcher might have a BPI of 1 or a little over, meaning that in an average inning he'd give up just one single or walk. A bad pitcher might have a BPI of 3. The average would probably be around 2.5, I'm guessing, although the median would be lower.

The drawback here is that BPI doesn't account for situations that do genuinely test a pitcher's skill but don't necessarily involve hits or walks. So with 1 out and a man on third, a good pitcher will get a strikeout or short pop-up which prevents the runner from scoring. A bad pitcher will more often surrender the sacrifice fly ball that lets the runner score, but isn't counted as a hit. In the "crude" BPI system, he wouldn't be charged for that. But maybe you could fix that by treating a run-scoring flyball or grounder -- or a case where the hitter intends to advance the lead runner and succeeds -- like a single.

It'd be interesting to see how this would change how we rate pitchers. I'd guess that Pedro Martinez, who this year has a very good but not spectacular ERA of 2.86, would fare even better under a BPI system, because frequently this year he has gone thru stretches where he retires 10 or 12 hitters in a row, but then has given up clusters of hits, allowing runs to score. Another Met pitcher, Victor Zambrano, seems to have pitched out of a lot of jams this year, and I'd bet his BPI would be comparatively worse than his (already relatively poor) ERA. But that could be wrong. In a later post, I'll use existing stats on hits and walks and innings pitched to figure out some crude BPIs for a few pitchers (the more sophisticated version would be impossible to tally from existing stats unless you went thru all the box scores, which would take forever), and see how they compare to ERAs in terms of the pitchers' rankings against each other.

By the way, I'm defintely far from the first person to be thinking along these lines. I'd bet that general managers, at least since Bill James, have been much more interested in hits per inning (or some version thereof) than ERA in trying to gauge a pitcher's future performance. Some version of this thinking is probably in Moneyball, which I keep meaning to read.