Lineup Analysis Can Get Weird

Just for fun I thought I'd try punching the lineup for the All-Time All-Star National League Team, Integration Era Only, into the statistical lineup generator tool. That team is: Johnny Bench (C), Albert Pujols (1B), Jackie Robinson (2B), Ernie Banks (SS), Mike Schmidt (3B), Barry Bonds (LF), Willie Mays (CF), Hank Aaron (RF), and Stan Musial (DH). My impulse for that lineup is Jackie/Stan/Willie/Barry/Hank/Albert/Mike/Ernie/Johnny. They project 6.964 runs per game for that lineup, i.e. 1128 runs per 162 (holy shit!). Their favorite arrangement of those players, based on career stats, is Bonds/Pujols/Schmidt/Mays/Musial/Banks/Aaron/Bench/Robinson. That's, uh, a weird lineup. Yeah, you want OBP from your leadoff hitter, but Barry Bonds is Barry frickin' Bonds, he's gotta be your clean-up hitter. Anyway, they project that lineup for 7.079 runs per game, i.e. 1146 runs, a difference of just under 2 wins (out of, like, 140 or whatever).

Just for fun, using neutralized batting statistics instead, I get my preferred lineup at 7.085 runs per game, i.e. 1148 runs per year (I guess these guys mostly played in low-run environments?). Meanwhile, they prefer a batting order of Bonds/Pujols/Schmidt/Aaron/Musial/Banks/Mays/Bench/Robinson, for some reason, and project 7.2 runs per game, i.e. 1166 runs per 162. Again, just under two runs of difference from "optimization." Barry Bonds, leadoff hitter... yeesh.