Understanding Sabermetrics (24 page)

The 2004 season seems to stick out from all the others. Is it possible that a ballpark could depress runs totals by about 10 percent in one year and about 12 percent two years later, but inflate them by almost 8 percent in the year in between? Possible, sure. However, a check of
Green Cathedrals
turns up no significant alterations to Camden Yards for the 2003 season, making the sudden, temporary shift even less likely. Clearly, something is wrong with the formula. So, in an attempt to diminish the effect of one clearly anomalous season, we take a three-year average of PF. In other words, to establish a PF for a particular season, which we will call PF-3 (PF - 3 year average), we add up the PF for that season and the two previous ones, and divide by three, i.e.,

These can also be weighted further by games or innings. Thus, to establish a more accurate PF for a season, say 1999, we take

The park adjustments can shed even more light on the RC/27 statistic (see exercises, chapter 6). If we divide the runs-created total by the number
, we obtain a figure that describes how many runs a team of 9 of a particular player might score per game. In 2003, Melvin Mora created 133 runs, using 382 outs to do it. Recall that outs are calculated by taking AB minus H and adding SH, SF and GIDP (but not CS) to that number. For 2004, Mora’s unadjusted RC yields
while if we use PF the number is 9.02, and using PF-3 his RC/27 is 9.49. If we were to calculate his PF-5, we would see that his RC/27 would equal 9.68. This is an incredible range of difference in runs per game, and an indicator that, just maybe, Mora’s season in 2004 did not receive the degree of recognition that it deserved.

2.

3.