xERA: Fixing ERA (kinda, sorta)

The statistic ERA has significant limitations in measuring how a pitcher performed. Pitchers have no control over their teammates’ defense, the official scorers, the relief pitchers who deal with their inherited runners, the sequencing of the opposing lineup, park effects and maybe most of all, dumb luck. Yet, using ERA as a metric assigns full responsibility for all those variables to an individual pitcher.

Despite that fatal framework, ERA remains one of the most public-facing statistics. It dominates broadcasts and reporting. Analysts further compound these problems when they break down a pitcher’s ERA into month-by-month splits and try to spin narratives off those arbitrary, limited samples of a tragically bad stat.

In a way, this is somewhat understandable. For decades, ERA was the best measure we had. It even seemed precise compared to pitcher Wins. But better alternatives are available now, a click or two away.

An array of “fielding independent” pitching stats mitigate many of the problems with ERA. FIP (Fielding Independent Pitching) uses strikeouts, walks and home runs to evaluate pitchers. xFIP tweaks FIP a bit, substituting fly ball percentage for home runs. SIERA is a more complex version of xFIP. These stats are often referred to as “ERA estimators” because they are scaled to look like ERA for familiarity. They are based on the assumption that the pitcher has limited control over batted balls once they are put in play. Research has shown they are more accurate in predicting a pitcher’s future ERA than is a pitcher’s past ERA.

The folks at MLB’s Statcast have entered the fix-ERA fray with a new ERA estimator: xERA (expected ERA). It’s on the Statcast leaderboard, so expect xERA to become popular as a mentioned alternative.

The concept underlying xERA is simple. It starts with measuring the quality of contact the pitcher actually gave up, plus strikeouts and walks. But instead of assigning the pitcher the specific scorebook result of those balls, xERA credits the pitcher with the average scorebook result for similarly-hit balls. That subtle difference factors out defense, scorers, relievers, sequencing, park effects and (to an extent) luck. In other words, variables the pitchers don’t control.

There’s nothing artificial or imaginary about a pitcher’s xERA. It’s based on the balls he actually gave up, his actual strikeouts and walks.

If you were paying close attention, I said xERA takes out some of the luck factor. Pitchers are lucky or unlucky based on where the bloops, line drives and fly balls happen to land. xERA normalizes those outcomes, so it corrects for that kind of good or bad luck.

But there is another kind of luck that can help or afflict pitchers and that deals with control over quality of contact. Pitchers do have a bit of control over the quality of contact, largely through velocity. And they can somewhat influence whether a ball is hit on the ground or in the air. But research shows that — overall — pitchers have limited control over what happens after the ball is put into play.

Unlike the Fielding Independent ERA estimators, xERA assigns the pitcher full responsibility for quality of contact.

Let’s put xERA to use and take a look at the 2020 Reds pitching staff to see how they rated. We’ll look at the six starters: Luis Castillo, Sonny Gray, Anthony DeSclafani, Trevor Bauer, Wade Miley and Tyler Mahle. This table shows their ERA, xFIP and xERA from 2019, along with the MLB average for starting pitchers.

What do these numbers tell us?

Luis Castillo – His xERA and xFIP are about the same. Castillo did a little bit better with balls hit in play. Both of his ERA estimators were in line with his ERA.

Sonny Gray – His xERA and xFIP are close. Both were considerably higher than his ERA. Regardless of measure, he pitched much better than league average. But the estimators show, based on the actual hit balls he gave up and the variables over which he had most control, that his 2019 ERA was lucky by just under a run.

Anthony DeSclafani – His xERA is in line with his ERA. That means he wasn’t lucky or unlucky with respect to balls in play, defense, etc. His ERA accurately reflected the balls he gave up. But his xFIP is about a third of a run higher, indicating the variables he most controls came in a bit worse.

Trevor Bauer – His xERA and xFIP are off by a fifth of a run, which isn’t much. They’re also both in the neighborhood of his ERA. That’s a consistent report on how he pitched in 2019, which was just a bit better than league average.

Wade Miley – His xFIP is about a half run higher than his xERA or ERA. That means he got by on balls that were hit softly, not on strikeouts/walks. Whether you expect him to be league average or a half-run better depends on whether you think a pitcher can consistently induce soft contact or whether that’s mostly luck.

Tyler Mahle – Mahle’s stats say a lot. Start with his xERA. It’s more than a half run better than his ERA. That means he was unlucky with the outcome of balls in play. Based on quality of contact, strikeouts and walks, Mahle was right around league average. But his xFIP was another half run below his xERA, which indicates his strikeouts/walks and fly ball rate supported an even lower run rate. If you buy what xFIP is selling, Tyler Mahle was the third best starter on the Reds staff.

We’ll take a similar look at the Reds 2020 relievers in an upcoming post.

You can find xFIP at FanGraphs and xERA at Baseball Savant. Both represent massive improvements over ERA in understanding how a pitcher performed.

Which is better? Your conclusion depends on how you come down on the underlying assumptions. If you think pitchers control contact quality, you’d favor xERA. If you think that’s largely out of a pitcher’s control, you’d prefer xFIP.

The good news is that it’s not important to choose. If you understand what each measures, you know more about that pitcher.