Cristopher Sánchez, Continuously Improving

In high school statistics, we got one rule beaten into our heads over and over again: Don’t extrapolate outside the bounds of your data. Then, when we were done learning that one, we got another rule: Outliers tend to revert towards the mean. Gee, thanks a lot, Mr. Gilliam – how am I supposed to explain Cristopher Sánchez using those two rules?
Look at this table of pitching statistics. Clearly, extrapolating past the edge of Sánchez’s performance seems fine:
| Year | K% | BB% | ERA | xERA | FIP | xFIP | SIERA | WAR/200 IP |
|---|---|---|---|---|---|---|---|---|
| 2023 | 24.2% | 4.0% | 3.44 | 3.72 | 3.99 | 3.09 | 3.33 | 3.6 |
| 2024 | 20.3% | 5.8% | 3.32 | 3.60 | 3 | 3.19 | 3.58 | 5.2 |
| 2025 | 26.3% | 5.5% | 2.5 | 3.02 | 2.55 | 2.77 | 3.02 | 6.3 |
| 2026 | 29.9% | 5.2% | 1.82 | 2.74 | 1.91 | 2.27 | 2.48 | 7.8 |
We have all of these ERA estimators – FIP, xFIP, xERA, SIERA – because merely looking at someone’s ERA can be misleading. ERA is noisy. Between inherited runners, sequencing, scoring decisions, and just plain old variance, knowing a player’s ERA in one year doesn’t necessarily mean you know how well they played. The whole alphabet of advanced pitching statistics comes down to trying to solve that noise problem by focusing on indicators with greater stability.
In other words, advanced ERA estimators tend to move around less than actual ERA. It follows naturally that changes in ERA estimators are more predictive of future results than changes in ERA. When all of these markers are moving in tandem – and moving by a lot, to boot – the aggregate picture looks very different than your average pitcher with a shiny ERA early in the season. Read the rest of this entry »







