clock menu more-arrow no yes mobile

Filed under:

How Far Back Do You Trust FIP?

Fielding Independent Pitching is a wonderful tool, but how relevant is it to a historical discussion of player value?

Richard Mackson-US PRESSWIRE - Presswire

A couple of things had inspired this post:

  • 1.) I keep hearing rumors that Fangraphs has ultra-secret plans to revamp their WAR calculation into something much more BIP-friendly. This has been whispered at least once in the Composite WAR thread, and again yesterday in the Dickey thread from Neyer.
  • 2.) Beyond the Box Score's Bryan Grosnick has an upcoming amalgamated WAR effort that is sure to fire up this debate very soon, anyway. We might as well be prepared.

But first, a couple of clarifications. I love FIP. I love DIPS. I love Fangraphs. I also recognize like any rational human being, that fielding independent statistics have their limits, just as anything else in life. I also hope Fangraphs pitcher's WAR remains based on fielding independent data because A.) we should have options, and B.) The option Fangraphs gives us is more often a better evaluator of true-talent in single season samples than RA/9, in my opinion.

But, when we begin to examine pitchers historically, I'm not sure FIP is as helpful.

Most readers of Beyond The Box Score are aware that one of FIP's best qualities is that it predicts next year's ERA so well. Typically, a pitcher's FIP in one season correlates to his next season's ERA at about r = .46 while Year 1 ERA correlates to Year 2 ERA at somewhere around r = .38. BTBS alumnus Bill Petti showed us these correlations for seasons 2004-2011 not too long ago. This high correlation is true only for recent seasons, however.

Because as we cruise back through the decades, the correlation between FIP and next year's ERA fades.

2000’s n=538 0.41 0.35 0.56
1990’s n=506 0.50 0.43 0.65
1980’s n=510 0.33 0.21 0.48
1970’s n=618 0.30 0.23 0.39
1960’s n=512 0.35 0.33 0.51
1950’s n=367 0.27 0.26 0.37

min 150IP

Why exactly?

Baseball has grown increasingly dependent on the components of FIP, aka the Three True Outcomes, as it's matured. Strikeouts, Walks, and Home Runs are now monopolizing the outcomes of more Plate Appearances than ever before. If we look at 3TO%, (K+BB+HR)/PA, through the years we can see the trend matches the rise of FIP's predictive powers almost perfectly:


I've always used 1992-1993 as the break point at which I begin to use or weight FIP less, though I've never done the hard analysis to confirm if that is the best time to do so. Does this mean I believe FIP has zero value prior to 1993? Not at all. The entirety of FIP's value is not solely found in it's predictive abilities. It still is a measure of what happened, uninfluenced by the pitcher's defense, and that is always an asset when estimating pitcher value. But I've realized I've chosen to weight it substantially less in seasons prior to 1993 because of how different the game was played prior to that season, and I'm curious if others do this as well.

What about you?

Everyone will like you if you would just follow @JDGentile on twitter.