August 13, 2010
Those Who Need Support
On August 5, Felix Hernandez toed the rubber against the division-rival Texas Rangers. Hernandez entered the game with a fantastic 2.79 ERA, and peripherals almost exactly in line with those posted a year ago, when he finished second to Zack Greinke in the American League Cy Young Award voting. The King lasted 6
Hernandez was no stranger to a lack of support. In his previous start, he scattered three runs over seven innings against a very solid Twins team and received, again, zero runs of support, recording another loss. On the season, Hernandez has a 2.71 ERA and just a 7-8 W-L record. Wins and losses are in no way the best measures of pitcher performance, but it still looks weird in spite of this knowledge to see such a low ERA alongside such an uninspiring record. But the situation with Hernandez lends perfect credence to the specific reason that wins and losses are not good proxies for success: run support. A pitcher can go five innings, give up seven runs, but be lucky enough that his team scores eight and walk away with a win. Or, Hernandez can pitch wonderfully, receive zero runs of support, and lose, simply because a pitcher decision can’t be handed to an offense.
In his eight losses, Hernandez has received—ready for this—a whopping seven runs of support. Yes, seven, and in four of the losses the Mariners have failed to score for him. Now, as I mentioned, a pitcher is not defined by his record, but it becomes very frustrating when dynamite pitchers end up with substandard won-lost marks when one considers how much this measure plays into the end-of-season awards. As Tom Tango recently showed, Hernandez's seasonal numbers are eerily similar to those posted a year ago; and when I say eerily similar, I mean they are really, really, really similar. The big difference is that last year he finished the season 19-5, while he seems more in line for something like an 11-13 record this season.
The major difference here is his run support—it’s impossible for someone to pitch equally well in two consecutive years and finish eight games “worse” if not for a lack of support. Overall, Hernandez has made 25 starts and received a whopping 68 runs of support, which comes out to be 3.36 per nine innings when stacked up against his league-leading 182
Hernandez’s strife got me thinking about run support in general and how infrequently it is mentioned. With that in mind, I was able to compute the run support for every starting pitcher in each start from 1995-2009, tallying only the runs of support received while they were in the game. For instance, if Roy Halladay leaves with a 2-1 lead and the Phillies win 6-2, Halladay received two runs of support, not six. From there, I calculated the run support per nine for each pitcher in each year, and compared the marks to their W-L records for the year.
First things first, though, here are the lowest rates in the wild-card era (minimum: 175 innings):
As we should have expected, the run support in the “leaderboard” above correlates strongly to the W-L record. Clemens is certainly an outlier as he managed to prevent runs at an incredible pace that season. Had he received some support, it stands to reason that a shinier record would have led to another Cy Young Award on his mantle. Does run support per nine correlate strongly to W-L record on an overall basis? Running a correlation across all of the qualifying pitchers produces an r of .48, suggesting that about 23 percent of the variance in winning percentage for pitchers can be explained by their run support, which seems a bit low given my expectations.
Next, my mind wandered toward which pitchers with next to no run support from a relative standpoint managed to produce fairly high winning percentages? These would likely be aces, but as Clemens’ 2005 season tabled above shows, even with tremendous run-prevention skills, getting no support produced just a 13-8 record. Here are the highest winning percentages in the wild-card era for pitchers who received <= 3.75 RS/9:
It isn’t impossible to produce a great winning percentage with such little run support, but the odds are definitely stacked against the possibility. Then again, certain winning percentages do not necessarily look as good or bad as the respective records. For instance, a .636 winning percentage is great, but does Cain’s 14-8 mark last year make anyone salivate? Didn’t think so. The only 20-game winner in this span to receive fewer than 3.75 runs of support per nine was Oswalt, but there are several future Hall of Famers on the list if we extend it beyond the elite eight.
Hernandez is unlikely to end the season with a shiny record and, because of that, he is very unlikely to receive any support when it comes to the Cy Young Award. In fact, I would go so far as to say that even if Hernandez finishes the season with a sub-2.50 ERA and great peripherals, that he won’t even receive a single first-place vote, simply because of the possibility that his record falls too close to, or under, the .500 line. This is really a shame, because run support seems to be an aspect of the game that is mentioned in passing but which doesn’t have any real pull.
Forgive me for thinking that certain voters may acknowledge a lack of run support for a candidate but ultimately let the ugly looking W-L record sway their vote in another direction. Don’t misinterpret this as my plea to give the award to Hernandez but rather one so that he is not overlooked simply because of a much less meaningful statistic. This leads into a macro issue in that it does not really seem that W-L records are going away anytime soon, so the way I see it is that perhaps metrics that are on the same scale need to be touted much more.
I think it would be much easier for someone less familiar with the more commonly discussed value stats to grasp that a pitcher’s record is affected quite a bit by his run support, and if the goal is to use W-L as a measure of how many good games a pitcher logged relative to the amount of clunkers, allowing for the possibility of cheap wins and tough losses is as suboptimal as it gets. I am all for keeping the chain moving in the areas of determining value, evaluating defense, and deriving the best run estimators, but I also think analysts and hardcore fans can sometimes lose the forest from the trees, forgetting that some of what we might consider the more basic tenets are the most important as far as advancing the reach of analysis and understanding.
At BP, we have an array of statistics in the support-neutral field that all attempt to measure performance while adjusting for run support. Statistics like Support Neutral Wins and Support Neutral Losses, as well as the corresponding Support Neutral Winning Percentage are all important metrics that too often get lost in the fold. Hernandez may be just 7-8 right now, but with league average run support our statistics reports suggest he has more likely been a 15-9 pitcher this year. Using run support to help shape a W-L record is an important tool, but without tangible numbers that make adjustments, claims are more likely to fall on deaf ears.