# Sabermetric Glossary: Expected FIP

by Ryan Rigato | Guest Writer |

Expected Fielding Independent Pitching, more commonly known as xFIP, is a formula used to stabilize a pitcher’s home run rates.  The rest of the calculation remains exactly the same as FIP. xFIP was created by Dave Studeman from The Hardball Times. The league average home run per fly ball percentage (HR/FB %) usually sits around 10.6 %.

The FIP formula is seen below:

((13 x HR) + (3 x (BB + HBP – IBB)) – (2 x K)) / IP + constant

The xFIP formula is seen here:

{(13 x (.106 x FB)) + (3 x (BB + HBP – IBB)) – (2 x K)} / IP + constant

*As you can see, the only difference is substituting the number of home runs given up with what the league average number would be determined by how many fly balls are hit into the field of play on the pitcher. As always, the constant is used to bring xFIP and FIP on the same scale as ERA.

Is xFIP always effective?

No, however, it is certainly one of the best metrics we have for predicting future ERA.

Can you prove xFIP is the best ERA estimator?

In 2009, Collin Wyers of The Hardball Times wanted to test which ERA estimator best predicted future ERA. To do this, he used a statistical tool known as root mean square error (RMSE). The RMSE allows us to know what the average of the magnitude of error is between two samples. Basically what Wyers did is split every pitcher’s season during the period from 2003-2008, an extremely large sample size, into even-numbered and odd-numbered days. He then tested whether using the data from the even-numbered days could predict the results for the odd-numbered days. Three of the ERA estimators are shown in the chart below:

 IP ERA FIP xFIP 10 3.21 2.73 2.49 20 2.41 2.03 1.87 30 1.89 1.61 1.51 40 1.83 1.46 1.28 50 1.75 1.41 1.35 60 1.18 1.06 1.02 70 1.47 1.28 1.18 80 1.20 1.01 0.94 90 1.07 0.98 0.91 100 1.20 0.85 0.87 110 1.05 0.82 0.76

What this table shows is that ERA is the worst predictor for future ERA. The way to interpret the results for the RMSE is to think of a pitcher having a 5.00 ERA over a 10 IP sample. Based on those results from the RMSE of ERA for 10 IP, the next 10 IP will result in an ERA from anywhere in the range 1.79-8.21, such a huge range is utterly useless. The xFIP estimator would predict an ERA in between 2.51-7.49, which is better but a larger number of IP is crucial for predicting future ERA. As you can see, FIP and xFIP consistently outperform ERA in predicting future ERA. The larger the sample size, the better predictor xFIP is, as the table shows.