Aim For The Head: Dropping the Worst Starts

One common defense you hear of a pitcher–usually one of the speaker’s favorites–whose season statistics don’t quite measure up
is something like: "Yeah, but if you take away the two starts when he got pounded and the manager left him in, his ERA is
really good!"

There’s some merit to the argument that a few starts can skew a pitcher’s cumulative line, and there have been attempts, such as
Michael Wolverton’s Support-Neutral statistics to better model the maximum impact a single game can have on a pitcher’s value.

Separate from these more sophisticated analyses lies a more fundamental problem: taking away X starts from a single pitcher,
without doing so for every pitcher, will overrate the pitcher in question by giving him a benefit the others do not have. If we
really want to go down this path of eliminating disastrous starts, we should give every starting pitcher the same benefit.

I looked at all pitchers from 2001 who had at least 100 total innings pitched in games started. I arbitrarily selected the three
worst starts as the number we would drop from season totals. Then, I compared the rankings and values from the whole season to
the trimmed season to see who was most affected. Pitchers who started for more than one team have their tenure with each team
measured separately, and are not combined.

I looked at two measures, one of which was simply Run Average (RA, which is Runs Allowed/Innings Pitched * 9). The other is a
simplified version of VORP, or Runs above Replacement Level, which I call Value for this article and define as IP/9* (6.00 –
R/IP*9). Note that RA and Value are not park-adjusted for this comparison, nor are we strictly comparing to a measured (or
modeled) replacement level, but instead using a 6.00 RA as replacement level. My intent was to identify the pitchers who are
helped/hurt the most by excluding three starts, rather than to accurately model the value of the top N-3 starts in a pitcher’s
season.

Top/Bottom 20 Pitchers by Value Lost by Trimming Three Starts:

Pitchers who completely bombed in three starts are helped most by this method. Scott Schoeneweis got blasted for 28 runs
in 9 1/3 innings in his worst three starts, dropping his season’s value by almost 22 runs. Overall his rank in value would go
from #78 to #58 just by dropping every pitcher’s three worst starts. Jimmy Anderson is similarly affected. There are some
good pitchers on the list, too. Matt Morris jumps into the top ten if we disregard three starts.

Pitchers who are very consistent are penalized by this method, since their worst starts aren’t too far behind their average
starts. Roy Halladay‘s worst three starts were barely below replacement level, and so he gains less than two runs by
removing them. His rank in Value drops 26 points as flakier pitchers pass him. Again there’s a mix of great pitchers (Randy
Johnson, Curt Schilling, Greg Maddux) and mediocre pitchers (David Wells, Jeff Suppan, Dave
Williams, Brian Tollberg) in the list. Consistency, by itself, doesn’t mean quality–you can be consistently good or
consistently bad.

Top/Bottom 20 Pitchers by RA Decline by Trimming Three Starts