clock menu more-arrow no yes mobile

Filed under:

When a problem comes along you must FIP it

Does Fielding Independent Pitching work when analyzing the Colorado Rockies?

Mike McGinnis

On August 12 Rany Jazayerli wrote an interesting piece documenting the Chicago Cubs' rebuilding strategy. The gist: the Cubs' brass have focused almost exclusively on collecting--via draft or trade--phenomenal young offensive talent. They passed on Jon Gray to go after Kris Bryant, they drafted Javier Baez, they traded Jeff Samardzija for Addison Russell, and they gave a whole heap of money to sign Cuban Jorge Soler. They are developing a fearsome core of hitters; but who is going to pitch for them?

This is where it gets interesting, at least from a Rockies perspective. Jazayerli praised the Cubs for targeting via free agency or trade pitchers that have underperformed their FIP. FIP is an old concept in sabermetrics, but still durable. It stands for Fielding Independent Pitching, and its basic premise is that pitchers have little to no control over whether balls put in play by batters will result in outs; in other words, it is in his defense's hands. The components of FIP are what the pitcher does control: strikeouts, walks, and home runs. It is then scaled to ERA and voila, you have a metric that evaluates pitchers while leaving the skill (or lack thereof) of their defenders out of the equation.

Pull quote from the article:

There are other stats that measure the same thing — xFIP, SIERA, Nate Silver’s Quick ERA — but for simplicity’s sake we’ll stick with FIP. The key point with all of these stats is this: They predict what a pitcher’s ERA will be next year far better than his actual ERA does. If a team is considering two pitchers, one of whom has a 3.00 ERA but a 4.50 FIP, and one of whom has a 4.50 ERA and a 3.00 FIP, it should bet on the latter pitcher every time.

So supposedly FIP is the secret sauce of pitcher evaluation. Regressions bear that out. Fangraphs uses FIP, not ERA, in their WAR calculation.

Earlier this week there was a minor hubbub about the Marlins designating former top prospect Jacob Turner for assignment. Indeed, it was probably the impetus for Jazayerli's article. He's young, has a prospect pedigree, and hasn't experienced a velocity drop; but his ERA was pushing 6, and the Marlins were evidently moving on. His FIP was a much more intriguing 4.00, so the Cubs put in a waiver claim and now have him for basically nothing.

Relevant to Rockies fans: why didn't the Rockies claim Turner? They had priority on the waiver wire due to their worse record. The Rockies are sending Yohan freakin' Flande to the hill every five days and they decided Jacob Turner wasn't a target worth acquiring for, again, basically nothing. Jazayerli:

Really, the only reason to think Turner would not wind up in Chicago was that the Cubs didn’t have first crack at him. They were second on the waiver wire behind the Colorado Rockies, and all the Rockies had to do was put in a waiver claim. But they didn’t, possibly because the Rockies are so rich in pitching that they just don’t have the room for him, or possibly because they’re a completely clueless organization that doesn’t know what the hell it’s doing.


But that got me thinking: What do we know about FIP in Colorado? In a sport where most ballparks play in the same rational plane of existence, Coors Field is a gibbering lunatic that inflates ERAs, batting averages, and managers' blood pressures. If the ballpark is unique in the sport, then maybe the evaluative tools we need are unique as well.

Premise 1: FIP is a more accurate way to predict next year's ERA than ERA is.

2014 5.01 4.63 4.14
2013 4.44 3.96 4.02
2012 5.22 4.59 4.28
2011 4.44 4.24 3.98
2010 4.14 3.83 3.8
2009 4.24 3.96 3.99
2008 4.77 4.3 4.32
2007 4.32 4.52 4.48
2006 4.66 4.51 4.69
2005 5.14 4.7 4.59
2004 5.54 5.13 4.98
2003 5.24 4.99 4.74
2002 5.21 5.08 4.62
avg 4.8 4.5 4.36

As you can see, since the installation of the humidor the Rockies regularly post higher ERAs than their FIPs (and even more so than their xFIPs, which normalizes home run rates. Coors Field inflates home runs as well. For simplicity's sake I'm sticking with FIP). This makes sense. FIP ignores balls put in play, but Coors Field is the very worst place to allow balls in play. The dimensions are so spacious that more hits will fall in Coors than anywhere else.

Amusingly, 2007 is the only year where the Rockies outperformed their FIP. That 4.52 mark is only sixth best on the list, behind such wonderful years as 2013 and 2011. Man, I just want to grind 2007 up and snort it.

Anyway, the point is that you can't just look at a Rockies pitcher and say, "His FIP is better than his ERA, so positive regression is in store." Colorado pitchers generally have to add a third of a run to their FIP to get their ERA.

That said, FIP still seems to be a better predictor of next year's performance than ERA (though I'd guess that it's much less strong than the correlation for other teams). In six of the 12 years since the humidor's installation, the Rockies' previous year's FIP, as opposed to their ERA, was closer to the subsequent year's ERA. So despite the fact that Rockies ERAs are routinely higher than their FIPs, in this small sample FIP was, at least half the time, a better predictor.

Differential in year-to-year ERA vs. year-to-year FIP to ERA

Year (Year) ERA - (Year-1 ERA) (Year) ERA - (Year-1) FIP FIP closer
2014 0.57 1.05
2013 -0.78 -0.15 x
2012 0.78 0.98
2011 0.3 0.61
2010 -0.1 0.18
2009 -0.53 -0.06 x
2008 0.45 0.25 x
2007 -0.34 -0.19 x
2006 -0.48 -0.04 x
2005 -0.4 0.01 x
2004 0.3 0.55
2003 0.03 0.16

Additionally, year-to-year ERA's for the Rockies are typically more volatile than FIP. Each season, the average change in ERA--whether positive or negative--is .42 runs. Meanwhile, the average differential between (year) ERA and (year-1) FIP was .35.

This may sound like gibberish, and I'll admit that my last stats class was many years ago, but it appears that the previous year's FIP is a better estimator of the next year's ERA than previous year's ERA.

Now that's obviously too simple. The sample size is 12 units. There is significant turnover in pitching staffs year to year. Injuries, especially to Rockies, happen constantly. So apply grains of salt. But at least we can breathe a little bit easier in knowing that the Rockies seem to operate on the same plane of existence as other teams, and that FIP is still a useful tool to observe.

Circling back to the Jacob Turner debate, there are two ways to approach the Rockies' decision to not claim him. You can defend the decision by citing the fact that Coors Field exacerbates the penalties of balls in play, and that pitchers who allow hard contact (looking at you Jeremy Guthrie) are going to get victimized. If the Rockies brass decided Turner was that kind of guy, then you can understand why they passed on him. I mean, do we really need Christian Friedrich 2.0?

On the other hand, if FIP is a better predictor than ERA, shouldn't the Rockies, who are doing nothing this year, still have taken the chance? Let me reiterate: Yohan freakin' Flande.

I don't have answers. An interesting study would be to aggregate pitchers whose ERA's habitually underperform their FIP's and see how they do at Coors Field. That would be a good starting point, but my database and programming kung fu is somewhat limited.

We know there are some ways to limit the damage of balls in play. Namely: assemble a pitching staff of ground ball pitchers and a crack infield defense; or try to accrue pitchers that allow weak contact via missing the bat barrel. As of now, the Rockies seem to have hitched their wagon to the former approach, stressing pitching to contact and ground balls. To put it mildly, that approach has been only fitfully effective.

Assembling a successful pitching staff at 5,000 feet is a monumental undertaking, and the Rockies have never really brought one together. The normal rules that teams operate under, rules that say that peripheral numbers like FIP are important, don't work in the same way at altitude. Is there a secret sauce for the Rockies? Who knows. But in 20 years of searching, answers are still elusive.