Who are the best starting pitchers of the past 60+ years? One way to answer that question is using RE24, the measure of how much a pitcher reduces his opponent’s’ run expectancy with each batter faced.
Starting from each of the 24 base-out states (ranging from nobody on, nobody out to to bases loaded, two out), there is an expected number of runs a team will score in the remainder of that inning, based on average hitters facing average pitchers. With the result of each plate appearance, a pitcher is credited with the resulting change in run expectancy (which can be positive or negative) less any runs allowed.
RE24, then, tells you how many runs a pitcher saved or cost his team relative to the average pitcher in the same base-out situations. Over the course of a career, the batters each pitcher faces will collectively approximate an average batter, allowing some reasonable basis for comparing different pitchers (with the possibly large caveat that RE24 does not adjust for park factors, team defense or other factors).
After the jump, the top 50 since 1950.
I’m using the metric RE24/9 to show the number of runs per 9 innings that a pitcher was better than the average pitcher in the base-out states that the pitcher faced. To qualify, a pitcher must have compiled 2000 IP since 1950. Only seasons since 1950 are counted (the data back to 1950 are mostly complete, with some data back to 1945, and scant data prior to that). Of the 220 qualifying pitchers, these are the top 50 in RE24/9.
[table id=162 /]
The column labeled boLI stands for base-out leverage index, which is a measure of how much leverage was associated with each plate appearance. What is meant by leverage? Essentially, this is a measure of how much variability there is in the run expectancies that could result from a given PA. Without getting too technical, from a base-out state with m baserunners and n outs, a PA can obviously only result in base-out states with a maximum of m+1 baserunners and a minimum of n outs. A weighted average of the differences (absolute values) in run expectancies between the current base-out state and each of the possible subsequent states (with the weights corresponding to the empirical probability for each transition) yields the boLI for that plate appearance.
Intuitively, boLIs are higher with more runners on base and lower if fewer. Thus, the pitchers with lower boLI numbers are those better at keeping men off base, while pitchers with higher numbers are less adept at this. By dividing boLI into RE24, RE24 is normalized (or “de-leveraged”) by showing how well pitchers did relative only to the range of outcomes possible from their base-out states. If I’ve confused the heck out of you, this explanation may be easier to grasp.
Ranking by RE24boLI/9 (boLI divided into RE24/9), those same fifty pitchers look like this.
[table id=167 /]
And, combining the two measures, here are our 50 pitchers, ordered by the sum of their rankings in the previous two lists.
[table id=168 /]
Finally, here are some other notable pitchers and their rankings (out of 220) in both RE24 and RE24boLI.
|Player||RE24/9||RE24 Rk||RE24 boLI/9||RE24 boLI Rk||IP|