By Bill James

    It has become increasingly common in recent years to hear that a relief pitcher's ERA is unnaturally low, by about 50 points or a full run. A relief pitcher undoubtedly has an ERA advantage over a starting pitcher, created by the fact that he often begins his work with one or two men out in the inning. If three pitchers pitch one inning, with each pitcher giving up two walks and then getting an out, the first (starting) pitcher will be charged with two runs, the second with one run, the third with none. A relief pitcher has an advantage because his share of the inning most often comes at the end of it, while the men who score the runs come at the beginning.

    But how large is the advantage? Who says it is a run or a half a run? Since every announcer in baseball seems determined to say that, shouldn't somebody check and see if it is true? This study, then, is an attempt to measure precisely the relief pitcher's ERA advantage.

    I began by compiling a list of every pitcher who pitched at least 40 games or at least 100 innings for any major league team between 1956 and 1970, and who made at least 90% of his appearances either as a starter or a reliever. There are 1006 such pitchers, 512 starters and 494 relievers. For each of these pitchers, three items of information were computed and/or recorded: the number of hits allowed per nine innings (HPG), the number of walks allowed per nine innings (WPG), and the ERA.

    The pitchers were then coded to indicate the number of HPG and WPG they allowed. A pitcher who allowed 5.50 to 5.99 HPG was coded "A", 6.00 to 6.49 "B", 6.50 to 6.99 "C", etc. through the letter "J", representing 10.00 to 10.49 HPG. A pitcher who allowed 1.00 to 1.49 WPG was coded "Q", 1.50 to 1.99 "R", etc. through "Z", which indicates 5.50 to 5.99 WPG. There were 40 pitchers who had such extremely high or low ratios of HPG or WPG that they were excluded, but the remaining 956 were thus divided into 100 "cells" of approximately equal pitchers. For example, the code "FT" indicates that the pitcher allowed 8.00 to 8.49 HPG and 2.50 to 2.99 WPG.  There are 35 pitchers, 20 starters and 15 relievers, in cell FT. Thus for every starter, we have a field of comparable relief pitchers, and for every reliever, a field of comparable starters.

    The Earned Run Averages of the starters and relievers in each cell were then compiled. The complete contents of one of the smaller cells, cell GW, are given for illustrations:

    Once this was done, the information in each cell was weighted and grouped with the information from the nearby cells to improve the reliability of the data. Thus the data listed for cell GW is not identical to the content of that cell alone. An explanation of precisely how the data were grouped, as well as detailed notes of the experiment, can be obtained from the author.

    Thus for every range of HPG and WPG, we devise an estimate of the ERA which is typical of starters and relievers in that range, and so have a basis for comparison of essentially equivalent pitchers.

    The information from the "border" cells, cells coded A, J, Q, and Z and representing the highest and lowest ranges of HPG and WPG, was not sufficient to establish reliable comparisons in those ranges. The typical ERA's derived from all other cells are presented below (asterisks indicate unreliable data):

STARTERS

RELIEVERS

    By subtracting the relief ERA's from the starting ERA's (using reliable data only), we then have 45 estimates of the relief pitcher's ERA advantage. These are:

    1. There is no doubt whatsoever that a relief pitcher does enjoy an advantage in compiling his ERA. There are 45 cells giving reliable information, and in all 45 the relief pitchers have better ERA's than starters of comparable ability, as measured by statistics not affected by pitching in relief.

    2. The data derived is quite consistent in suggesting an ERA advantage to the relief pitcher in the range of .15 to .25. We may with very little fear of contradiction say that a relief pitcher's ERA should be adjusted upward by .20 for accurate comparison to the ERA's of starting pitchers.