Saturday, January 31, 2004
IP ERA 1-R Win%
Good Pens 10,088 3.29 .545
Bad Pens 10,301 5.24 .469
He then shows that the teams with "Good Pens" did better, winning 58%, and the teams with "Bad Pens" did worse, winning only 44%, in games decided by two or more runs.
Here's the thing: The over 20,000 inning sample size seems pretty significant, but it's misleading. Only a fraction of those inning actually represent those pitched in one-run games. A better survey would look at every team (not just the exceptionally good or bad), and only in one-run games. Being the nerd I am, my Friday night schedule left plenty of room to begin such an analysis, so here goes:
First, I went through the 2003 March and April American League schedule, and looked at the boxscores for every one-run game in that span (51 games altogether). Then I sorted the pitching stats into two categories: Winning Team Bullpen ERA (WBE), and Losing Team Bullpen ERA (LBE). If carrierd out for the eleven years of Neyer's study, we would have around a 4,000-inning sample to work with. In my limited sample, WBE in one-run games was 3.32, and LBE in one-run contests was 4.54, a difference of 1.22. Bullpens averaged roughly 3 innings on both sides of the one-run games, so bullpens accounted for about 40% of the winning margin in these games.
A better statistic thatn Bullpen ERA would be ERA of a bullpen's top three pitchers. These are the ones that get the outs that count. It is quite irrelevant if a team's mop-up reliever has an ERA of 3.00.
Close game performance, in my view, is the true measure of a bullpen's strength. Pitchers pitch completely differently at the end of these games than they do if the score is, say, 6-3. Batters are fighing for every inch, and pitchers aren't giving in to anybody.
(If you were looking for a strong finish to this post, I'm sorry, but it's past my bedtime, and I'm bushed)