Why Run Support Can Make Win-Loss Records Misleading

Nick AllenCorrespondent IJune 28, 2009

ST. PETERSBURG, FL - APRIL 13: Pitcher Scott Kazmir #19 of the Tampa Bay Rays starts against the New York Yankees on April 13, 2009 at Tropicana Field  in St. Petersburg, Florida.  (Photo by Al Messerschmidt/Getty Images)

A pitchers' win-loss record can be very misleading.

Many factors contribute to the win-loss statistic, one of the biggest factors is run support. With no run support, it's going to be impossible to win, obviously.

So according to the win-loss statistic, a pitcher that goes nine innings gives up one run, and gets no run support at all, isn't as good as a pitcher that goes five innings, gives up nine runs, but gets 10 runs of run support.

That's what makes this stat flawed.

For example:

Scott Kazmir has a 7.69 ERA, but a 4-4 record, he can thank his offense that provides a Major-League-Leading 7.7 runs of run support per game.

Matt Palmer is 6-1 with a 4.47 ERA, but, he also has the 5th best run support among qualifiers this year.

Jason Vargas has a 3.79 ERA which is well below league average, yet he sports a 3-3 win-loss record due to his run support of 2.8 runs per game, fifth-worst in the majors.

In 1992, Jim Abbot had a 2.77 ERA the fifth lowest among AL qualifiers that year, but he went 7-15 that year due to the 2.5 runs of run support he recieved while pitching.

In 2008, Jake Peavy, had a 2.85 ERA, good for sixth best in the majors but a sub .500 winning percentage, with a 10-11 record, because his run support of 3.5 runs per game was sixth-worst in the Majors.

In 2004, Shawn Estes had a 5.84 ERA and went 15-8 just because his offense produced 5.6 runs per game.

I asked Zach Fein to run some numbers for me on run support, ERA, and win-loss percentage and he came up with this:

Using the 88 pitchers that qualified in 2008, the correlation between win-loss percentage and ERA was -0.65; the correlation between W-L% and run support per game was 0.55. The difference between the two is not that large and suggests that run support plays a large role in W-L%.

As well, a regression between the three variables (W-L%, ERA, and run support, the latter two standardized) shows that every point of ERA affects W-L% by -0.08, while every run of run support affects W-L% by 0.07. There's minimal difference between the effects of ERA and run support on W-L%.


A pitchers win-loss record shouldn't be used to justify that one pitcher is better than another, due to the variables that go into a win or a loss.