The Remains of Walk-offs

On the surface, Brett Myers seemed to handle his 2012 transition from the rotation to the bullpen pretty well. He notched 19 saves and eight holds between stints with the Astros and White Sox against just two blown saves. He posted a sharp 3.31 ERA, good for an 81 ERA-.

But it wasn’t all happiness for Myers, as he was on the unfortunate end of four walk-offs. Hunter Pence hit a solo home run off Myers with one out on May 15th. The other three times, Myers’s game ended with runners still on base in threatening situations: a runner on third and one out after a Dexter Fowler walk-off triple on May 28th, runners on first and second and nobody out after a Hector Sanchez walk-off single on July 14th, and a runner on second and two outs after a Jamey Caroll sacrifice fly on July 28th.

Those four situations — the runners he stranded (all allowed on by him, although one reached on an error) and the outs at the time — resulted in a combined run expectancy of 3.25. Myers pitched just 65.1 innings in 2012. Even accounting for the nine outs Myers would have needed to record to finish those four innings (and therefore realize the full run expectancy), these 3.25 runs would increase his ERA to 3.59 — a significant difference.

The 3.25 runs Myers left in the remains of a walk-off were the most by any reliever in 2012. There were 48 others to leave at least one run behind. Eighteen of those 48 left at least two runs behind, listed below (click the graph for the full list from 2012):

Two runs probably sounds insignificant, but a pitcher who threw 60 innings — a typical relief season — would see his ERA balloon by 0.30 points with two extra runs charged to his name. For many pitchers, 0.30 runs is the difference between above-average and mediocre, the difference between elite and good, or the difference between serviceable and below replacement level. Myers’s ERA painted a deceiving picture for many reasons — he had a 4.26 FIP, thanks to a fairly low strikeout rate and some home run issues — but getting to strand three runners in scoring position merely by the game ending before he could allow any more runs in was a part of it.

David Carpenter’s 2011 season showcases this concept best. The then-Astro easily topped that season’s list (again, click to see the full list):

Carpenter left a tremendous 4.49 runs on after three walk-off singles in 2011. All three left at least two runners on and less than two outs; two left the bases loaded.

Carpenter pitched just 27.2 innings in 2011. He recorded a 2.93 ERA and a 78 ERA-, especially good numbers for a 25-year-old reliever making his major league debut. But, like Myers, his profile showed more blemishes upon a deeper look — a 4.2 BB/9 and 1.0 HR/9 led to a 4.18 FIP despite a sharp 9.4 K/9. He stranded 88.2 percent of baserunners for the season. He stranded eight baserunners thanks to allowing walk-offs alone (in all three cases, Carpenter entered the game at the beginning of the inning). Adding the 4.49 runs and seven outs he left in the wake of walk-offs would increase his ERA to 4.05 — far more similar to his 4.18 FIP (and 110 FIP-) than his shiny recorded ERA.

Carpenter posted an 8.07 ERA in 2012 between the Astros and Blue Jays. He stranded just 61.3 percent of his baserunners.

The predictive value of this concept of course isn’t as high as this well-chosen example makes it seem; we’re already dealing with small sample ERAs and relief pitchers. There’s other noise involved as well, including intentional walks and inherited runners. But I think it does show yet another limit of ERA in evaluating relievers. There were 435 walk-offs between 2011 and 2012 alone. There were 101 pitcher seasons with at least one run in run expectancy left behind after walk-offs, and 32 of them had at least two runs left behind. In short seasons, these vagaries can make a difference worth considering when evaluating late-inning relievers.