ERA is half a pitching stat
Earned run average looks like the cleanest individual stat in baseball. It is denominated in runs per nine innings, it appears next to the pitcher's name on the broadcast graphic, and it has been the headline pitching number since the 1910s. The cleanliness is an illusion. ERA is a measurement of runs scored against a pitcher's team while he is on the mound, lightly adjusted to exclude the runs the league has agreed to forgive. That measurement depends on the seven position players catching baseballs behind him, on the park he is pitching in, and on the sequencing draws of the innings he happened to throw. The pitcher's name sits on top of a number whose biggest single component is often somebody else's work.
The defensive dependency
Every batted ball that becomes an out instead of a hit lowers the pitcher's ERA. Every ball that becomes a hit instead of an out raises it. The pitcher influences the type of contact — ground ball vs fly ball, hard hit vs soft — and on a longer time horizon the type of contact does statistically translate. On any given start, the defense behind him determines whether the soft fly to right-center falls in or not, whether the slow grounder up the middle becomes an out, whether the rope down the line is run down or rolls into the corner for a double.
Defensive efficiency for a full lineup ranges from roughly 67% to 73% of balls in play converted to outs across modern teams. That seven-point spread, applied to a single starter across a season, is worth roughly half a run of ERA. A league-average pitcher who happens to throw in front of a top-five defense will post an ERA half a run lower than the same pitcher would post in front of a bottom-five defense. He pitched identically. The number on his ledger looks like a different season.
The park effect
Park effects are the other large dependency that does not belong to the pitcher. Coors Field has been suppressing the ERAs of pitchers who leave it and inflating the ERAs of pitchers who arrive at it for thirty years. The 2025 park factors place Colorado about 18% above league average for run scoring, Cincinnati about 11% above, Tampa about 8% below, and Seattle about 9% below. Those factors are compounded across roughly 16 starts of home pitching per season for a typical rotation arm.
The implication is that a Coors pitcher needs to be roughly a run of pitcher-quality better than a Seattle pitcher to post the same ERA. The free-agent market knows this. Front offices model it. Hall of Fame voters mostly do not, which is why the historical ERA leaderboard is heavy with pitchers from suppressive environments and why pitchers who spent their careers in launching pads remain underrepresented relative to their actual run-prevention contribution.
The sequencing draw
The most underappreciated structural problem with ERA is sequencing. A pitcher who allows three singles in the same inning gives up multiple runs. A pitcher who allows three singles in three different innings, with the right outs in between, gives up zero. The pitcher's underlying contact-allowed profile is identical. The runs scored line is very different. ERA captures the runs scored line, which depends in part on the order the hits arrived in, which is not a property of the pitcher's pitches.
Strand rate — the percentage of baserunners who do not score — has a league average around 72% and a normal range of plus or minus four points season to season for individual pitchers. Pitchers whose strand rate runs hot will post ERAs well below their underlying contact profile. Pitchers whose strand rate runs cold will post ERAs well above it. The strand rate regresses almost completely toward the league mean across multi-year samples. The ERA gap that looked like ace-vs-back-end one year will frequently flip the next, with nothing in the pitcher's process changing.
What FIP and SIERA were built to fix
Fielding Independent Pitching strips ERA down to the three outcomes a pitcher controls almost entirely: strikeouts, walks, and home runs allowed. It throws out everything else and rescales the number to look like ERA so it reads on the same axis. FIP correlates with next-year ERA significantly better than current-year ERA does, which is the cleanest empirical demonstration that ERA contains a substantial chunk of stuff that is not the pitcher.
SIERA, the next-generation version, adds batted-ball type and base-out context and outperforms FIP slightly. xERA, introduced when Statcast launched, uses exit velocity and launch angle on every batted ball to estimate the run environment the pitcher actually generated. All three of these stats post smaller year-to-year swings than ERA, which is the way to recognize that they are isolating signal from noise. They are not perfect. They are closer to the question the ERA column is pretending to answer.
Why ERA still gets the headline
ERA persists because it is intuitive and because the alternatives require explanation. A 3.20 ERA sounds like a good year. A 3.20 FIP also sounds like a good year, but the FIP requires someone to first explain that it is not the same as ERA and does not move with it in a given season. Broadcast convention prefers numbers that do not require disclaimers. The consequence is that the season's ERA leaderboard usually contains two or three names whose underlying pitching was much closer to average and whose defense or strand rate or park did the heavy lifting.
Awards voting has slowly caught up. Cy Young ballots since roughly 2015 have shown clear weight on FIP and on innings-adjusted versions of ERA estimators. Hall of Fame voting has not. Both pitcher contracts and pitching coach evaluations are now built primarily on FIP-style metrics, which is a quiet vote of no confidence in ERA from the people whose money is on the line.
The honest read
The right way to read an ERA is as a noisy estimate of a pitcher's run-prevention contribution, with somewhere around 40% to 50% of the variance in any given season driven by defense, park, and sequencing rather than by the pitching itself. The number is not useless. It is what actually happened. The mistake is treating it as a clean readout of skill when it is a partial readout filtered through a stack of dependencies the pitcher does not control. The pitcher whose ERA was a half run lower than yours may have been a worse pitcher with a better infield. The ERA column will not tell you. The estimator columns will, if you let them.