Chad Qualls, Kevin Millwood, and Sample Size

The San Diego Padres have reportedly signed Chad Qualls to a one-year contract for about two million dollars in an effort to restore some of the bullpen depth they’ve traded away this winter. In signing Qualls, they are making a pretty big bet against the predictive power of ERA, considering he posted an atrocious 7.32 mark last year.

However, the Padres understand how to build a bullpen on the cheap, and they also realize that ERA is a pretty lousy way to evaluate relief pitchers. To illustrate the point, let’s compare Qualls’ season to that of another free agent, Kevin Millwood.

Qualls had the following miserable line last year: 59 innings, 85 hits, 56 runs, seven home runs, 21 walks, 49 strikeouts. That’s just not good, no matter how you slice it.

Millwood had the following miserable line last year: 51 innings, 86 hits, 54 runs, 11 home runs, 25 walks, 36 strikeouts. That’s even worse than Qualls’ performance.

The difference? That awful stretch of pitching represents just a portion of Millwood’s 2010 season – June 3rd to August 1st, specifically. For two months, he was every bit as bad as Qualls, but because he’s a starting pitcher, those 51 innings didn’t represent his entire season’s body of work. While he was brutal for that stretch, he got a chance at redemption through a larger sample size, and he took advantage of it.

Here’s Millwood’s line from immediately after August 2nd through the end of the season: 66 innings, 60 hits, 25 runs, 7 home runs, 23 walks, 38 strikeouts. After posting a 9.18 ERA in the middle two months of the season, he finished the final two months with a 3.26 ERA, despite not improving his peripherals much at all. The only real difference? His BABIP during the first stretch was .410, and his BABIP in the second stretch was .262.

I know its tempting to look at Qualls’ 2010 line and think that there’s no way he just experienced lousy defense and bad luck all year. A .399 BABIP is really terrible, after all. However, we have to keep in mind that it was just 60 innings, and starting pitchers have lousy strethces for 60 innings at a time too. However, because that only represents 30 percent of their season, we don’t notice these awful stretches as much, and when regression to the mean inevitably kicks in, it gets washed away entirely.

Since relief pitchers throw just a fraction of the innings that starting pitchers accumulate, they are subject to much larger seasonal variations. They don’t have time to regress to the mean in-season, and as such, you’ll see things like a pretty solid reliever in Qualls posting an ERA over 7.00. The Padres are willing to be that one 60-inning stretch doesn’t mean that Qualls is finished as an effective major leaguer, and given how Millwood finished the year for the Orioles, they have some pretty recent evidence to support the theory.