StatLab
← Level 3: Linear weights

xFIP: FIP's what-if cousin

What if his home-run luck were exactly average?

FIP treats home runs as pure skill — but the fraction of fly balls that clear the fence (HR/FB) bounces around wildly year to year, driven by park, weather, and warning-track inches. A pitcher can 'earn' a terrible FIP on ten fly balls that died at the track last year but carried this year.

The substitution

xFIP recomputes FIP replacing the pitcher's actual homers with the number he'd allow at a league-average HR/FB rate on his fly balls. It answers: 'with neutral homer luck, what does his profile look like?' For forecasting, it's often steadier than FIP.

The obvious objection is real: some pitchers genuinely suppress (or serve up) homers as a skill, and xFIP erases that along with the luck. Use the trio as a ladder — ERA (what happened) → FIP (his controllable events) → xFIP (his profile with neutral HR luck) — and note where the rungs disagree.

The formula (optional — skippable)

xFIP = (13·(FB × lgHR/FB) + 3·(BB+HBP) − 2·K) ÷ IP + constant

How this stat lies to you

  • It erases real homer-suppression skill along with the luck — extreme fly-ball pitchers in big parks get misjudged.
  • In tiny parks the 'league average HR/FB' assumption never comes true for anyone.
  • Three ERA-like numbers invite cherry-picking; quoting only the flattering one is a classic bad-argument move.

Check yourself

1. xFIP differs from FIP by…

2. A pitcher's HR/FB was 19% last year (league ~12%). His FIP 4.90, xFIP 3.70. Story?