Explaining Analytics: A friend in FIP

Game No. 162

Credit: Getty Images

Yu Darvish #11 of the Texas Rangers throws against the Los Angeles Angels in the second inning at Rangers Ballpark in Arlington on September 29, 2013. (Photo by Ronald Martinez/Getty Images)

Print
Email
|

by JOSEPH URSERY

WFAA Sports

Posted on January 7, 2014 at 9:34 AM

Updated Tuesday, Jan 7 at 9:38 AM

Kevin Millwood once led the American League in Earned Run Average.

 
That's not a joke. It actually happened. You can look it up and everything.
 
You know what that means, don't you? Yes, you do. It means if you want to look at a pitcher's contributions, you'd best find a metric better than ERA.
 
You need FIP, friend.  You need Fielding Independent Pitching.
 
What is FIP?
 
First, a few broad strokes before we get down to the details (and the devil who lives in them): FIP is built off of a philosophy that runs counter to the mainstream- that is, once the batter makes contact with a pitch, there's very little that happens that the pitcher can control. Obviously that's not a complete truism; if it were, we wouldn't be so concerned with ground ball rates as we are. But, largely, it works; by focusing on the Three True Outcomes (Strikeouts, Walks, and Homers) we can divine a more accurate picture of the pitcher's quality than ERA does.
 
The other thing to note is that FIP doesn't work so well on its own. There's nothing intrinsically wrong with just citing FIP, but it's best used as a piece of the pie. Think of it this way; ERA is like your weight, FIP is your body fat percentage, and xFIP (we'll get to that later) is your cholesterol. Any one carries information, but the best practice is to look at them all together (right, doctors?).
 
Ideally, FIP gives you an idea of how repeatable a pitcher's ERA is. For its faults, ERA does a decent chance of communicating what actually happened; at its heart, it tells how how many runs a pitcher gave up per innings he pitched. It includes a lot of other noise, though, which can lead to situations like that one time Kevin Millwood lead the AL in ERA. That magic season, Millwood had a 2.86 ERA. That's good. However, his FIP was 3.73. That's still pretty good, but it's not leading-the-AL good. 
 
That FIP, being a almost a full run higher than his ERA, tells us that we should expect his ERA to rise over the larger sample to get closer to the FIP mark. And his next season, he put up a 4.52 ERA with the Rangers. 
 
Here's the real kicker: he was pretty much the exact same pitcher those two seasons. He struck out the same amount of guys (6.84 per nine innings in '05, 6.57 in '06), walked the same (2.44 to 2.22) and gave up pretty much the same rate of home runs (.94 per nine in the magic '05 season, .96 in '06). Where did that extra run-and-a-half-plus of ERA come from, then? Well, his BABiP (Batting Average on Balls in Play) in '05 was .281, and in '06 it was .306. More balls turned into hits than outs in '06, so he gave up more runs.
 
That's why his FIP was 3.87 in '06, basically the same as his 3.73 in '05. In both cases, FIP said his raw results (ERA) didn't match his real contribution (FIP), just one year went to his happy side and one year went to his sad side.
 
There's a lot of ways we could try to describe that difference, but the shorthand that's common is 'luck'. Surely, luck is just one factor, and it's a bit demeaning to pitchers to sum everything up to good luck or bad luck, but for allt he research that's been done, it's still really the best overall explanation we have. Pitcher's BABiP's tend to fluctuate pretty heavy year-to-year; if putting up low BABiP was a skill, wouldn't pitchers want to do it every year? 
 
So how do I use FIP?
 
FIP is ERA's good buddy, rather than his ruthless replacement. Anytime you look at ERA, take a look at FIP and see how they interact. If FIP is a lot lower than ERA, expect ERA to rise (maybe the pitcher has been 'lucky' and the baseball gods are going to smite him). If it's lower, expect it drop (maybe the pitcher's luck has been bad and the baseball gods will look favorably on him in the future). If they're close, then, he probably neither owes nor is the owner of a debt against the baseball gods.
 
Generally, FIP will run just a little bit higher than ERA, so don't fret over a half-run or so difference. If it's that close, then he's probably not going to go in either direction pretty crazily.
 
What does FIP tells us about the 2013 Rangers?
 
Here's the good news: Yu Darvish is really good at baseball. His 2.83 ERA was ninth in baseball last year. His FIP was 3.28, which was good for twentieth. Remember what I said about the half-run difference? There you go, and there Yu goes. File this away, we'll come back to it.
 
Tanner Scheppers was pretty good last year. He had a 1.88 ERA, which is pretty phenomenal. Except FIP sees him as 3.74. That's just about twice as much. FIP is pretty bullish on Scheppers maintaining his elite-ness in 2014.
 
So what does FIP not do?
 
FIP is a metric that needs a lot of data fed into it, which means it's not great for small samples. It's not a great metric at all to break down an individual game, or parts of a season.  Partially, that's because it's actually possible to have a negative FIP, which means I guess that you pitch so well it actually adds runs to your offense.  It's extraordinarily difficult to maintain that over more than a few innings, but it's fun when it happens (to your team's pitcher- it's maddening when another pitcher does it to you)!
 
Also, FIP allows for a little bit of what we call luck to bleed in, in that if a pitcher gets a little unlucky with his home runs per fly balls allowed (HR/FB), it will still show up in FIP. Luckily, we can fix that- enter xFIP (Expected Fielding Independent Pitching).
 
What differentiates FIP and xFIP is xFIP subs a league average HR/FB rate in for the pitcher's individual rate. Ideally, this helps normalize against flukey fluctuations (hey, have you heard of my new band, The Flukey Fluctuations?) and differences in ballparks/environments (a pitch that's ten rows up in Yankees Stadium is dead on the track in San Diego). IT does help sharpen the predictive ability of FIP, a little bit.
 
Remember when I said we'd come back to Yu Darvish? How I said he was twentieth in FIP in the league, and his FIP was a little higher than his ERA? His xFIP was fifth in all of baseball, 2.84.
 
A difference of .001 in ERA to xFIP (remember, 2.83 ERA for Darvish in '13) tells us that it's actually likely Darvish's ERA would look a little better moving forward than it did. That should excite you, friends. I'm pretty sure it excites Yu.
 
So, next time you fire up your ERA calculators, pay attention to his buddies FIP and xFIP and think about what they say as a whole. Remember, weight, body fat percentage, cholesterol- and we want to keep them all in control with diet, exercise, and wise statistic usage.

Joseph Ursery is a part-time nerd, full-time dad and clocks regular overtime for his Rangers fandom. It's a busy life. If you want to hear more about it, follow him on Twitter at @thejoeursery.
 

Print
Email
|