FIP is a cool, clever, and simple statistic to try and normalize how we work with ERA in a way that simply ignores defensive factors. However, we can ignore, or normalize for, at least one other factor to adjust the comparison.
Expected FIP (xFIP) is a simple tweak to FIP that further reduces noise in pitching statistics by using expected home run rates for the pitcher, rather than the actual number of home runs they gave up. Home run rates for a pitcher are unstable, much in the way that BABIP was for hitters, so looking at how many home runs they “should have” given up based on the number of fly balls hit regresses xFIP a bit more compared to FIP.
To compute it, we just replace the number of home runs with the number of fly balls hit (FB), multiplied by the league percentage of fly balls that are home runs (we’ll call it FBHR%).
\text{xFIP} = \frac{13\times(\text{FB}\times\text{FBHR}\%) + 3\times(\text{BB}+\text{HBP}) - 2\times \text{K}}{\text{IP}} +\text{FIP Const.}This moves us from a measurement of actual performance — FIP is philosophically a predictor of ERA, but actually calculates something concrete — to something meant to predict and compare performance to an idea state. Home runs are dependent on weather and which stadiums the pitcher plays in; while we could come up with some weighting system or league-adjustment system that is more complicated, xFIP is just another rough approximation that tweaks FIP to be closer to what we’re ideally using it for.
xFIP is mainly implemented on FanGraphs, so I don’t have any pieces of trivia handy right now. Again, it will regress FIP towards the average, if just slightly, based on whether a pitcher happened to give up more or less home runs than expected.
Continue to Day 18 – Fixing ERA+ With a Minus