clock menu more-arrow no yes mobile

Filed under:

FIP and xFIP and the rejection of 'in-between results'


The art of pitching has a clear goal: prevent the opponent from scoring runs. While it may not be possible to stop the opponent from scoring runs entirely, the pitcher has the clear objective of keeping the damage done to a minimum. Traditionally, the runs that a pitcher has given up have been measured by ERA, or Earned Run Average, which counts the number of runs scored from baserunners that the pitcher has allowed, excluding those that have gotten on base because of an error made in the field. The idea behind this stat, then, is that a pitcher takes full responsibility for the runs that are scored while he is on the mound (or later, in the case of inherited runners scoring).


Sabermetricians and their followers reject this "philosophy" of pitcher responsibility, because they argue that the defense behind pitchers plays an important role as well. To filter out the influence of the defense, they assume that all "in-between results" are not predictive of future performance. The FIP, or fielding independent pitching, statistic therefore only counts the most well-hit balls (home runs) as a measure of hittability, walks as a measure of bad command, and strikeouts as a measure of the ability to miss bats (well, duh). So all types of contact a pitcher allows beyond home runs are considered irrelevant, because a pitcher, it is believed, doesn't have any control over how well batters make contact on non-home runs.


Now, xFIP (expected FIP) goes a step back, in eliminating home runs from the equation. Instead, xFIP uses flyball% as a measure of the pitcher's ability to induce weak contact. More flyballs means a pitcher is less skilled at inducing weak contact. Now, consider this graph: Vla_vs_sob_plot_medium


Here you can see that those balls that are called flyballs are indeed usually hit hard (high Speed Off the Bat or SOB), while groundballs are usually hit sofly and at a low angle (VLA). Although flyballs (and also home runs, in the extreme) and groundballs are on opposite sides of this spectrum, both line drives and popups form an important part of the spectrum as well. Popups are a sign of weak contact (but are included in FB%, so seen as bad), while line drives are a sign of hard contact, but they're not counted as flyballs, so they actually improve a pitcher's xFIP. Popular theory, well sabermetric theory anyway, holds that pitchers have no influence over line drives hit against them, and that the influence of popups is insignificant. But is that belief justified?

Line Drives

Batted ball types have been measured since 2002, so naturally I'm looking at pitchers who have pitched a significant number of innings in the period 2002-2011. In this case, I'm using 700 innings pitched as my minimum. That's kind of arbitrary, but I picked it without any prior knowledge as to how this would affect the results of whatever I may research. Looking at line drives, the percentages range from 23.5% (Glendon Rush) to 15.5% (Fausto Carmona). The ones who allowed a lot of line drives were often the less illustrious names on the list (Esteban Loaiza, Tomo Ohka for example) or aging veterans (Tom Glavine), while those who didn't allow line drives were groundballers (Tim Hudson, Felix Hernandez) or other guys who have some decent stuff (Scott Kazmir, Jered Weaver), and there was also the one knuckleballer almost at the bottom (of course, Tim Wakefield). Let's see if these groups are/were defying the expectations that the xFIP/FIP theory had/would've had, assuming that this is a large enough sample size to use ERA quite confidently:

Name LD% ERA- FIP- xFIP-
Glendon Rush 23.5 117 99 99
Jason Thomson 22.8 99 92 97
Russ Ortiz 22.7 112 110 115
Kevin Millwood 22.1 95 92 95
Livan Hernandez 22.0 104 102 106
Kenny Rogers 22.0 94 99 104
Casey Fossum 21.9 123 110 104
Cory Lidle 21.9 107 96 92
Brian Moehler 21.8 122 111 105
Esteban Loaiza 21.6 101 92 94
Jason Jennings 21.6 104 99 107
Tomo Ohka 21.6 96 105 105
Tom Glavine 21.5 92 103 105
Mike Hampton 21.5 107 103 107
Jeff Weaver 21.5 111 103 103
Jered Weaver 18.4 77 86 94
Jeremy Guthrie 18.3 95 107 108
Jonathan Sanchez 18.3 103 100 100
Adam Wainwright 18.3 72 81 85
Kelvim Escobar 18.2 84 82 90
John Danks 18.2 91 93 96
Scott Kazmir 18.1 97 97 99
Ubaldo Jimenez 18.1 84 80 92
Tim Hudson 18.0 78 89 87
Chien-Ming Wang 17.8 94 93 96
Brandon Webb 17.7 72 76 75
Felix Hernandez 17.2 77 82 77
Tim Wakefield 16.9 97 102 109
Derek Lowe 16.8 94 90 85
Fausto Carmona 15.5 108 106 101

(ERA-, FIP- and xFIP- are all corrected for league and ballpark)

Averages for the High LD% group: ERA- 105.6, FIP- 101.1, xFIP- 102.5

Averages for the Low LD% group: ERA- 88.2, FIP- 90.9, xFIP- 92.9

While nothing conclusive can be drawn from such a small sample, it does seem as though line drives do have some correlation with allowing hard contact. Other factors, of course, do play a role. Derek Lowe and Fausto Carmona have both been bad at keeping runners on the bases, which means they have underperformed their (x)FIP predictions despite allowing fewer line drives than other pitchers.


Name Popup% ERA- FIP- xFIP-
Chris Young 8.26 90 102 106
Jered Weaver 6.60 77 86 94
Ted Lilly 6.58 92 100 99
Eric Milton 6.45 117 111 109
Tim Wakefield 6.35 97 102 109
Jason Schmidt 5.55 82 78 88
Scott Baker 5.54 97 93 94
Barry Zito 5.54 93 103 108
Matt Cain 5.53 80 88 101
Jaime Moyer 5.31 98 109 106
Johan Santana 5.29 67 77 80
Chris Capuano 5.23 105 103 97
Oliver Perez 5.21 111 116 108
Randy Wolf 5.16 97 105 103
Woody Williams 5.12 100 106 106
Chien-Ming Wang 1.55 94 93 96
Jason Marquis 2.33 108 111 107
Zach Duke 2.20 109 102 102
Paul Maholm 2.17 105 100 100
Felix Hernandez 2.15 77 82 77
John Lannan 2.14 98 112 108
Greg Maddux 2.12 92 90 85
Fausto Carmona 2.08 108 106 101
Tim Hudson 1.96 78 89 87
Mark Mulder 1.81 94 98 90
Aaron Cook 1.68 97 93 102
Mike Hampton 1.62 107 103 107
Jake Westbrook 1.59 100 99 93
Derek Lowe 1.40 94 90 85
Brandon Webb 1.19 72 76 75

High Popup% group averages: ERA- 93.5, FIP- 98.6, xFIP- 100.5

Low Popup% group averages: ERA- 95.5, FIP- 96.3, xFIP- 94.3

Interestingly, though perhaps not surprisingly, we see a few familiar names on the bottom half of the list. Plenty of pitchers who avoided giving up line drives did so by making batters pound the ball into the ground. While this is a solid strategy, the weak outs generated by this strategy may be cancelled out in most cases by not generating many popups, which are easy outs as well. The group that does get a lot of popups seems to do very well at generating weak outs, and those pitchers are generally underrated by FIP and even more so by xFIP, which was off by as much as 7% (0.30 points of ERA if the league average ERA is 4.30).


Although both xFIP and FIP are not perfect, their method of using only extreme outcomes and nothing in between seems to work pretty well. Even if a pitcher somehow has control over the line drives he gives up, the effect of that on his ERA seems minimal. Flyball pitchers, who often generate a good amount of popups, do seem quite a bit underrated by this system, because their type of weak contact, the infield fly ball, is actually seen as a bad thing by xFIP, which punishes pitchers for giving up fly balls. FIP underrates these pitchers because they trade some extra home runs for more weakly hit balls in play, which FIP is blind to, since it only counts home runs, strikeouts and walks.

Rookies Jeremy Hellickson and Josh Collmenter specialized in using popups to get outs in 2011, while on the Jays Carlos Villanueva (6.42%, .271 BABIP) impressed, at least for a while, using the same "strategy" (we don't know if it was intentional). Can they successfully emulate the Jered Weavers of this league for an extended period of time? Or will those popups start turning into harder hit flyballs that sometimes leave the park?