Sunday, July 11, 2010

FIP and xFIP in Your Fantasy Toolbox

We assume a certain level of fantasy baseball experience and fantasy baseball understanding here at Roto Think Tank.  As such, we do toss around statistics and terminology with the thought in mind that everyone knows what we're talking about and can process that information.  A friend of mine, however, reminded me privately that there are newcomers to fantasy baseball who take a look at this site, or, at the very least, are new to the more formulaic/sabermetric approach to the game that we look at here. 

Since we are talking quite a bit about trading the last day or two, and since this is a key time before most leagues' trading deadlines, I thought I would take a look at a set of my favorite stats, FIP and xFIP, and talk about how to use these stats in evaluating your own team and in evaluating potential trades.  One caveat: for purposes of this article, I am going to ignore some very meaty discussions that come along with FIP/xFIP, including what a pitcher is and is not responsible for; whether FIP should be scaled with runs against versus earned runs against; the relationship between FIP and wins above replacement; and other similar articles.  Those debates are for another time, but feel free to Google the topic if you are interested in reading more after this article.

FIP stands for Fielding Independent Pitching.  It essentially tries to account for everything that a pitcher is responsible for during the course of a game.  The formula itself is: (HR*13)+((BB+IBB-HBP)*3)-(K*2)/IP + C, where C is a league-wide constant (HardBallTimes puts that constant at around .92).  So what does FIP tell us?  It essentially tries to tell us how well a pitcher pitched inside a vacuum that does not include the fielders behind him.

xFIP is still in its advanced experimental stage, but the reasoning behind it is sound.  xFIP is Expected Fielding Independent Pitching.  It normalizes home runs by using the average number of home runs per fly ball, under the theory that home runs are the product of fly balls and home runs.  In theory, this is the better predictor of a pitcher's future performance.

Alright, now that we know what FIP and xFIP are, how so we use them correctly?  Let's use John Lester as an example.  Lester is currently sporting an ERA of 2.76.  His FIP is 2.92 and his xFIP is 3.38.  So what do these numbers tell us?  Well, FIP tells you that he is pitching somewhere nearly as good as his ERA would reflect.  Why, then, is the xFIP so much higher (relatively, at least)?  Well, remember that xFIP normalizes home runs.  So while FIP is looking at Lester's 5.8% FB/HR rate, xFIP is normalizing that rate and predicting regression.  Of course, looking at Lester's home run rate in previous years, xFIP does appear to be the potentially more accurate indicator of Lester's future performance; assuming an increase in FB/HR rate, Lester's ERA should rise.

Of course, I picked Lester because I wanted a complicated example.  The HR/9 and FB/HR rates suggest that Lester is in for an upward correction, as seen in xFIP.  If you are looking at Lester on your team and say, okay, his ERA might rise to 3.38, which means that he's going to put up a 3.8+ the rest of the way, you probably shrug and say "okay, I can live with that."  But Lester poses an interesting analytical issue.  In one of his two prior full seasons, Lester outperformed both his FIP and xFIP, mostly based on his low HR/9 and FB/HR rates.  So it would not be shocking to see Lester outperform his FIP and xFIP and continue at his current pace, in light of the low HR/9 and FB/HR rates.

I am never an advocate of one set of numbers, so, as an aside, Lester's increased ground ball rate, increased swings at pitches out of the strike zone, and decreased swings at pitches in the strike zone suggest that the ERA can continue to outperform the xFIP.

Francisco Liriano is an example on the other end of FIP and xFIP.  Liriano is sporting a 3.32 ERA, but his FIP is a ridiculous 2.09.  Why?  The miniscule 2.5% FB/HR rate.  The xFIP accounts for this some, but his xFIP is still a nice 2.88.  Now you might look at the strand rate, ground ball rate, and FB/HR rate, and think that the ERA is in for a moderate correction.  On the other hand, if you take a look at Liriano's 2006 season, you will see this is not the first time that these numbers have made their way to Liriano's stat sheet.  So you might want to hang on to Liriano if you have him, because the numbers suggest that those numbers could be better in the second half.

The obvious advice is to take a look at xFIP before looking at a starter in trade (not so much for relievers, because of the smaller amount of innings; the sample size is too small to make it an accurate indicator).  As a guide, here are some pitchers that appear to be good targets to buy and sell (with the caveat that some of these pitchers, such as Baker, do give up a lot of home runs, so xFIP is much kinder to them than FIP):

Pitcher ERA-xFIP

James Shields 1.31
Scott Baker 1.14
Dan Haren 0.91
Randy Wells 0.9
F. Liriano 0.88
B. Morrow 0.85
Gavin Floyd 0.69

Clay Buchholz -1.82
Tim Hudson -1.72
Jeff Nieman -1.66
Jaime Garcia -1.6
L. Hernandez -1.59
David Price -1.54
C.J. Wilson -1.44


Derek said...

Hey Toz,
I had a couple thoughts/points/questions.

1) You say that you don't like using xFIP for relievers. But why is the small sample size of a reliever any different than the small sample size of a starter? And what do you use instead?

2) I'd disagree that we should be unsurprised if Lester beats his xFIP simply because he had a low HR/FB in 2008. HR/FB is such an unstable stat that one abnormal year doesn't mean a whole lot. It takes 4 or 5 years for us to have enough data to predict even half of the movement in it. Nearly every pitcher in baseball has had at least one year with an abnormal HR/FB.

3) I'm not sure if you're familiar with it, but if not, I thought I'd point out a few articles that I've written over the past year about the shortcomings of FIP:

Derek Carty
The Hardball Times Fantasy

Toz said...

Derek - I didn't go back and look at the links, but I know I recognize the titles, so I'm sure I've read them at some point in the past.

For this post, I wanted to lay out some basic principles and thoughts on the topic, and I hope to get into some of the deeper issues related to FIP and xFIP. I am admittedly fascinated by these measurements.

As for Lester, I agree that everyone has an abnormal HR/FB year, and I also agree that the sample size is small. On the flip side, so many of his numbers look like 2008 that I think it is fair to say this could be a repeat. Of course, if we all agreed, well, what fun would that be?!

Thanks for the comments Derek, and I hope we can further explore the topic!