2010 FIP Challenge Results Part I
In 2009, I did a column at the All-Star break to see if FIP or xFIP would be more helpful for fantasy players in making trades. This year I did the same thing. You can see the results below
Should Fantasy Owners Use FIP?
FIP Challenge Results Part I
FIP Challenge Results Part II
FIP Challenge 2010
Here is the table from the 2010 article, along with an additional column, the pitcher’s ERA in the second half of the season.
| Name | HR/FB | ERA | FIP | xFIP | 2nd Half ERA |
|---|---|---|---|---|---|
| Paulino | 1.9 | 4.40 | 3.25 | 4.60 | 15.88 |
| Liriano | 2.5 | 3.86 | 2.18 | 2.97 | 3.31 |
| A. Sanchez | 3.4 | 3.66 | 3.46 | 4.52 | 3.44 |
| Buchholz | 3.6 | 2.45 | 3.45 | 4.26 | 2.20 |
| J. Johnson | 3.8 | 1.70 | 2.31 | 3.06 | 3.50 |
| Matsuzaka | 4.1 | 4.56 | 3.83 | 4.98 | 4.79 |
| J. Santana | 4.5 | 2.98 | 3.62 | 4.69 | 3.00 |
| Vargas | 4.7 | 3.09 | 3.62 | 4.84 | 4.66 |
| Verlander | 5.3 | 3.82 | 3.11 | 3.89 | 2.89 |
| Zito | 5.3 | 3.76 | 3.91 | 4.79 | 2.89 |
| Jimenez | 5.4 | 2.20 | 3.13 | 3.71 | 3.80 |
| Gorzelanny | 5.4 | 3.16 | 3.26 | 3.92 | 5.20 |
| Danks | 5.6 | 3.29 | 3.41 | 4.13 | 4.19 |
| Hanson | 5.6 | 4.13 | 3.26 | 4.02 | 2.51 |
| Cain | 5.7 | 3.34 | 3.82 | 4.72 | 2.91 |
| Kershaw | 5.7 | 2.96 | 3.11 | 3.79 | 2.84 |
| C. Lee | 5.8 | 2.64 | 2.58 | 3.34 | 3.79 |
| Wilson | 6.1 | 3.35 | 4.14 | 4.71 | 3.36 |
| L. Hernandez | 6.2 | 3.37 | 4.02 | 4.71 | 4.02 |
| Fister | 6.3 | 3.09 | 3.75 | 4.38 | 5.06 |
| Carmona | 6.3 | 3.64 | 4.08 | 4.61 | 3.93 |
| Floyd | 6.5 | 4.20 | 3.28 | 3.78 | 3.91 |
| Buehrle | 6.6 | 4.24 | 4.16 | 4.85 | 4.32 |
| Morrow | 6.7 | 4.86 | 3.42 | 3.93 | 3.69 |
| Cueto | 6.9 | 3.42 | 3.91 | 4.45 | 3.96 |
| Lackey | 6.9 | 4.78 | 4.39 | 4.98 | 3.97 |
| Correia | 15.7 | 5.26 | 4.82 | 4.22 | 5.64 |
| Hamels | 15.2 | 3.78 | 4.53 | 3.85 | 2.23 |
| Blackburn | 14.8 | 6.40 | 5.89 | 5.14 | 3.93 |
| Millwood | 14.8 | 5.77 | 5.03 | 4.32 | 4.23 |
| Karstens | 14.5 | 5.42 | 4.88 | 5.50 | 5.00 |
| Duke | 14.5 | 5.49 | 4.89 | 4.36 | 7.00 |
| Shields | 14.3 | 4.87 | 4.11 | 3.55 | 5.59 |
| Bannister | 14.0 | 5.56 | 5.26 | 4.69 | 9.47 |
| Wolf | 13.9 | 4.56 | 5.81 | 5.24 | 4.11 |
| Davis | 13.7 | 4.69 | 5.69 | 5.10 | DNP |
| Nolasco | 13.7 | 4.55 | 4.39 | 3.84 | 4.40 |
| Kennedy | 13.7 | 4.12 | 4.83 | 4.31 | 3.38 |
There are 38 pitchers in the above chart. On a raw scale, the two systems were almost identical, with the FIP metric did a better job of predicting 2nd half ERA in 20 cases while xFIP did better 17 times. Furthermore, FIP did a better job of forecasting 14 of the 26 players with low HR/FB rates and both systems got five of the 11 pitchers (Davis being a wash) with high HR/FB rates.
In 2009, xFIP did better on a raw scale, as it did a better job predicting 20 of the 34 pitchers in the sample.
Similar to 2009, most of the pitchers fell outside the range predicted by FIP and xFIP. For example, Liriano’s FIP was 2.18 while his xFIP was 2.97. But he had a 3.31 second half ERA. Only seven of the 37 players listed above had a second half ERA between their first half FIP and xFIP. The two systems were close on this, too, with FIP being better on four of these seven pitchers. In 2009, 28 of the 34 pitchers had second half ERAs outside the range of their first half FIP and xFIP numbers.
Later today I will post a breakdown of all 38 pitchers in this survey.
I have a hunch that something that’s somewhere between xFIP and FIP is probably the most accurate predictor (or something else entirely). It would be interesting to see what the statistics would look like comparing the dead middle between xFIP, and FIP. I’d wager this number would be a better predictor then either of the two individually. While xFIP ignores park/league factors, FIP takes this into account, but ignores lucky/unlucky HR/FB rates.
So funny, I wrote this and didn’t see your post. I took the numbers from this year and last from these posts and took out Paulino and Davis. I found that the average is about .1 runs better than either FIP or XFIP (which were very similar). Thats significant over the course of a whole season, as it can be the difference between 1st and 8th in ERA.
I was doing some quick run on these numbers, and I found that if you average FIP and XFIP, you will find that this will be somewhat significantly closer on average. I know mathematically speaking there is no basis for this, but it’s true here and I use this mechanism in my fantasy decisions and it has done me well.
Any thoughts?
The sample size is too small to crown one of the three methods a winner. If I lived in my grandmother’s wall-papered guest room, I would have her make some sandwiches so me and my friends, Retrosheet and Excel, could extrapolate the $#*! out of this $#*!.
What about just comparing first half era to second half era? Does that do a better or worse job predicting?
ERA was interesting (Remember, small sample). ERA was actually only slightly behind FIP in distance from the mean. However, the standard deviation was significant compared to FIP, xFIP, and AVG of the two.
ERA had the largest grouping of pitchers within .15, outpacing xFIP 3 to 1, but it got much worse from there.
xFIP and FIP are based on linear regression models, and their inputs lack “interaction terms.” Basically, take two players, one that allows an OBP of about .300 because of good walk rates, and one that allows an OBP of .350 due to poor walk rates. Obviously homeruns will hurt the second pitcher more overall, since there are likely to be more runners on base. So an equal increase in homeruns should not logically produce equal changes in ERA/FIP. An interaction term is some stat that combines OBP and HR/9, for instance, so that the formula will expect an increase in Roy Halladay’s HR/FB to affect him less than, say, Barry Zito.
On that note, pitchers that don’t allow a lot of balls into play (high strikeout guys) and allow more groundballs should also be affected less by expected increases in HR/FB. I think this probably explains why FIP is more accurate for elite pitchers, but xFIP predicts better for “soft tossers.”