FanPost

Is Feldman even worse than FIP suggested? // Is FIP Not Really Defense Indendant?

Before going further, I'd like to point out that Feldman is one of my favorite pitchers to watch in a long time, although this year has been painful because he clearly hasn't had the same movement, command, or velocity on his pitches. But going on...

FIP was estimated as:

ERA = A * K/9 + B * BB/9 + C * HR/9 + D

Resulting in A = -2, B = 3, C = 13, D = 3.2


But why were rate stats used to fit ERA (another rate stat) to predict a rate stat? Why not fit to the underlying numbers? Innings Pitched is defense dependant, so lets abuse Feldman as an example of what I'm getting at.

Applicable stats from last year were (From Fangraphs):
IP: 189 2/3
K: 113
BB+HBP: 74
HR: 18 (HR/FB ~ 9.5%)
H – HR: 160
FIP: 4.41 (calculated as [-2*K + 3 * (BB+HBP) + 13 *HR ] / IP + 3.2)
TBF: 791

Now, Feldman’s BABIP as calculated on Fangraphs was .275. I recalculated it as (H-HR / (TBF – IP*3 – BB+HBP – HR) = .273.
Calculating his TBF as H + IP*3 + BB + HBP, he should have faced 821 batters:
178 hits + 74 walks + 569 outs (3*IP) = 821, suggesting he got 30 of his 569 outs from CS, GIDP, other defense related outs. Should Feldman really get credit for stuff like an outfield assist, or every GIDP since it's technically the defense turning it?

Using these rates, I recalculated what would have happened indifferent to defense behind him by assuming no extra outs (CS/GIDP/other defense) with a .300 BABIP (average was .303). He allowed 604 balls to be put in play, so the projected number of hits would be 175.8 (ignoring the 18 HRs he allowed).

This results in the following line:
IP: 175.8
K: 113
BB+HBP: 74
HR: 18
H – HR: 175.8
FIP: 4.52
TBF: 791

As you can see, I’ve had him face the same number of batters (791), and ignored any outs due to defensive-plays (DP, Assist, CS), and calculated an FIP that is .10 higher. Also note that this resulted in him throwing almost 13 fewer innings, which would have been picked up by someone in the bullpen.


The xFIP is also bumped up similarly since he’s penalized for having the 10 fewer innings pitched due to defensive plays resulting in extra outs, and 1.2 extra HRs given an average HR/FB rate of 10%.

Am I missing something? It seems to me that FIP should be calculated with respect to batters faced, and backing into the implied ERA given a league-average BABIP — which leads me back to my original question posted at BTB:

How close would the calculated Runs Allowed / 9 * IP come if you fit the following regression (with an additional variable from the BTB version)
Runs Allowed = A * HR + B * K + C * BB + D * (TBF-TTO)

Would this potentially be a better evaluation of pitcher performance than FIP, given that FIP can be impacted by plays made by the defense?

I would theorize that A would equal the run expectancy of a HR, B would equal the run expectancy of a strikeout, C would equal the run expectancy of a walk, and D would equal the run expectancy of the average ball in play.

Anyone have thoughts on this? Reasons why I'm wrong?