FIP and xFIP Explained: Why Sharp MLB Bettors Ignore ERA
Available in US
The number every TV graphic shows you that I’ve stopped trusting
I made my biggest single MLB profit of 2023 backing a starter who carried a 5.10 ERA into June. The line had him as a +130 underdog at home. He won outright, struck out nine, walked one, and the bookmaker line on his next start moved a quarter run in his favour. The market caught up. I’d already cashed.
The trade was sitting on my screen for the same reason most edges in baseball sit on screens: ERA was lying about him, and the people pricing him were leaning too hard on the lying number. FIP and xFIP are the two stats I use to call that lie out, and once you understand what they actually measure, the way you read pitching changes for good.
Why earned run average breaks down so badly in small samples
ERA gives a pitcher credit or blame for things he can’t control. A bloop single that bounces in front of a slow left fielder is a hit. A laser one-hopper directly at the shortstop is an out. The pitcher’s input was identical. ERA treats them as opposites.
Across a full career, that noise washes out. Across fifteen starts – the size of sample most early-season MLB bets are based on – it doesn’t wash out at all. A starter can run a 2.40 ERA on a defensively gifted team that converts 73% of balls in play into outs, and a different starter with the same underlying skill set can post 4.80 with a leaky defence behind him. They’re the same pitcher. The market reads them as different.
The break-even threshold against standard -110 juice is 52.38%. Every percentage point you can squeeze out of a single market input matters. If your handicap is built on ERA, you’re paying full price for a number that’s been distorted by inputs the pitcher never controlled.
What FIP actually does, in plain English
Fielding Independent Pitching strips out everything that involves the defence behind the pitcher. The formula isolates the three outcomes a pitcher controls completely: strikeouts, walks and home runs allowed. A small constant gets added so the league-wide average FIP matches the league-wide average ERA, which makes the two numbers directly comparable on the same scale.
Read out loud: FIP equals home runs times 13, plus walks plus hit-by-pitches times 3, minus strikeouts times 2, divided by innings pitched, plus a yearly constant of about 3.10. The constant changes slightly each season to keep the scale aligned with the league ERA.
What that gives you is a number on the same scale as ERA but cleaned of defensive luck and ballpark sequencing. A starter with a 3.20 FIP is performing like a 3.20 ERA pitcher would in a neutral environment, regardless of what his actual ERA reads. League average FIP sits in the high 3s to low 4s depending on the run-scoring environment of the year.
I treat the gap between ERA and FIP as a regression alarm. Anything wider than 0.75 in either direction over more than ten starts is a strong signal the run prevention is about to move toward the FIP. Lower ERA than FIP – luck running hot, expect it to cool. Higher ERA than FIP – defence has been wrecking him, expect his run prevention to improve as the sample grows.
Where xFIP takes the cleaning one step further
Home runs are the loudest single output in the FIP formula, and they’re also the noisiest. A pitcher’s home run rate fluctuates wildly inside a season because a lot of home runs are fly balls that just barely cleared the wall in the right park on the right night. Move the same fly ball five feet shorter or shift it to a deeper park and it’s a fly out.
xFIP – expected FIP – replaces the pitcher’s actual home runs with what his fly-ball rate would have produced at the league average HR-per-fly-ball rate. The result is a stat that asks: assuming this pitcher’s fly balls left the park at a normal rate, what would his FIP be?
That correction matters most early in the season and most for ground-ball specialists who are getting hurt by a small cluster of fluky home runs. xFIP is the better short-sample tool. FIP becomes more meaningful as the home run sample stabilises across a full season. I lean on xFIP through April and May, then trust FIP more from June onward.
Where xFIP is least useful is the genuine outlier – pitchers who really do suppress home runs through stuff and command rather than luck. A handful of starters consistently outperform their xFIP for years. The metric flags them as overrated when they aren’t. The fix is to look at three-year rolling HR-per-fly-ball rates: if the suppression is real, it’ll show up across multiple seasons.
How I actually use these numbers when I’m reading a card
My screen workflow is short. I pull up FanGraphs splits for both starters. I look at the season ERA-FIP gap. I look at the season FIP-xFIP gap. I look at the last 30 days of FIP versus the year-to-date FIP. Three numbers, two pitchers. Six checks. Total time on the laptop: under two minutes.
The bet appears when the line disagrees with what those checks tell me. If both starters look similar by ERA but the FIP gap is half a run in one direction, the moneyline is mispriced. If a starter’s last 30 days of FIP is meaningfully better than his year-to-date, the public is still pricing him on April numbers, and there’s value taking him before the market catches up.
The discipline is to use these as filters, not as standalone bets. FIP doesn’t tell me what the bullpen looks like, what the wind is doing, or whether the opposing lineup hits left-handers. It tells me whether the starter is being mispriced. From there I still have to do the rest of the handicap. The numbers don’t bet for you. They just stop you from betting on a lie.
For the next layer down – actually evaluating the lineup that walks up to the plate – the right tool changes. Pitcher stats clean for defence. Hitter stats need cleaning for park, era and outcome weighting, which is exactly what wOBA and wRC+ do for evaluating MLB hitters.
Is xFIP better than FIP for short-sample bets?
In most cases, yes. xFIP normalises home run rate, which is the most volatile input in the FIP formula. That makes xFIP the more stable predictor across April and May, when home run samples are tiny and prone to flukes. From midsummer onward, FIP becomes the better tool because the home run sample has stabilised and FIP captures any real home run suppression skill the pitcher actually has.
Can a pitcher have a great ERA and a terrible FIP at the same time?
Yes, and it happens every season. The cause is almost always elite defence behind him, a friendly run of weak lineups, or favourable sequencing of base runners. None of those inputs are repeatable. A starter with a 2.50 ERA and a 4.20 FIP across fifteen starts is a regression candidate the market will catch up with – the bet is on the FIP, not the ERA.
Prepared by the how do you bet Baseball editorial staff.
