It appears that Mark Melancon is going to be a part of the return in the Joel Hanrahan trade with the Boston Red Sox. Last year, he had a 6.20 ERA. This is ugly. He also had a 4.58 FIP. This is less ugly, but still bad. His xFIP, on the other hand, was 3.45. For context, it was 3.14 when he pitched to a 2.78 ERA for the Astros in 2011.
What drives the difference in these numbers is Melancon’s HR/FB rate. The formula for xFIP “is almost exactly the same as the formula for FIP, with the lone difference being how each accounts for home runs:” FIP uses home runs, while xFIP regresses HR/FB rate. Melancon’s poor FIP and poorer ERA were driven by his HR/FB rate, which increased from 11.1% to 22.2%. Melancon still induced 50.0% groundballs, so his 1.60 HR/9 innings is absurdly high as Pat Lackey, of WHYGAVS, pointed out on Twitter.
When Charlie Wilmoth, of Bucs Dugout, and Dejan Kovacevic, of the Pittsburgh Tribune-Review, were…discussing?…Melancon, their disagreement came down to whether they thought that Melancon’s home run rate would return to earth. Essentially, Wilmoth argued that the influx of home runs was the result of “luck” while Kovacevic held that it was not.
Immediately, my mind jumped to a wonderful article written by Dave Cameron at FanGraphs about Tim Lincecum, who had a similar problem. Cameron’s argument is that Lincecum’s home runs aren’t the result of luck, but rather random variation. The sabermetric community has simply fallen into the pattern of calling it luck. (I encourage you to read the whole piece. It’s very well done.)
The difference between luck and random variation is very important. As Cameron points out, it’s not unlucky if you throw a pitch down the middle of the plate and it gets ripped 450 feet for a home run. The number of times you do that for every certain number of pitches is the result of your skill level. The principle of random variation says that it’s very possible for a good pitcher to throw a number of bad pitches with greater frequency in small samples.
For example. Pitcher X throws 500 innings, with a home run rate of 1 home run per nine innings. This would mean that he would have given up roughly 55 home runs in those 500 innings. Random variation says that 8 or 10 of those home runs could very well come in, say, 40 innings. That’s a home run rate of 1.8. Those 40 innings don’t have as much predictive value as the full 500 innings. In fact, they have very little predictive value. As such, it is much better to trust the larger sample size and/or regression to the mean when it comes to HR/FB rate.
Relievers’ seasons are, almost uniformly, small sample sizes. As a result, I put a lot more faith in Melancon’s career HR/FB rate of 13% over 157 innings than his single season rate of 22.2% over 45 innings. If this is the case then he will likely be an effective reliever going forward.
N.B. This is not indicative in any way of my thoughts on the trade as a whole. This is a discussion about Mark Melancon, nothing more. If you saw Twitter tonight, you will understand my disclaimer. This disclaimer is useless because no one read this. This is a matter of form to make me feel better.