Understanding Sabermetrics (18 page)

Here are Schmidt’s 1986 MVP numbers:
which works out to a BRC of 117.3 and a TECHRC of 121.94.

Continuing the analysis of Mike Schmidt, we see that he contributed 409 outs, a figure determined by taking AB - H, and adding SH and SF, outs not counted in the original total. We also add his 8 GIDP, as these are additional outs he caused. These 409 outs translates to 409 / 27 or 15 games (actually, 15.1). The National League in 1986 averaged 4.18 runs per team per game, slightly lower than the usual historical figure of 4.5. On the other hand, at 15.1 games, Schmidt produced 117.3 / 15.1 = 7.74 runs per game using BRC, and 121.94 / 15.1 = 8.05 runs per game using TECHRC. If we put these figures into the Pythagorean Record, we obtain
for the basic version, and
for the technical version.

Many fans of the New York Mets would dispute the MVP selection of Schmidt in 1986. However, the player on the Mets with the highest runs created numbers was Keith Hernandez. His 1986 statistics are listed here:
which yield BRC = 101.07 and TECHRC = 103.41 (you can verify these numbers).

HDG23, used for both AL and NL data between from 1955 to 1988, has the following formula:

Note how it differs slightly from HDG22, which is used only for the major leagues in 1954:

Since the publication of the 24 HDG runs-created formulas, James has introduced some nuances that bring more situational numbers into the formula, including home runs with the bases empty. However, as always, any element is included because it makes the league estimate more accurate. In one adjustment to the RC formulas, James suggests introducing some coefficients to the A, B and C quantities. Thus, a more accurate runs-created formula, putting the individual player’s numbers into the context of a team, would be

Expanded out, and factoring out the 9, yields
which seems to weight the individual’s numbers as one-ninth of a team’s offense. The other coefficients are unexplained, except that possibly they were arrived at via trial and error using norms throughout baseball history.

In any case, Schmidt’s numbers for HDG23 were A = 246, B = 325.48, and C = 657, so
a value that is about 1 run lower than his BRC estimate. Using these coefficients, Keith Hernandez’s runs estimate drops to 97.7.

1. Here are some numbers for the American League in 1998: