Get Involved: ERA, FIP and xFIP

With the off-season mercifully winding down, and with a lot of new(er) faces on the site, it's an opportune time to go into two frequently mentioned advanced metrics: Fielding Independent Pitching (FIP) and expected Fielding Independent Pitching (xFIP).  This isn't tended to be an "us versus them" lecture, nor is it meant to condescend or browbeat. 

Is it rocket science?

Many people on the site seem fine with FIP and xFIP, and its use has come a long way since the first orientation was given on Purple Row in April '09.  It's not particularly new anymore, and the theory of Defense Independent Pitching Statistics (DIPS), conceived and developed by Voros McCracken, has been around for about a decade.  McCracken and DIPS even have their own Wikipedia pages, and McCracken worked for the Red Sox for a time, which is pretty neat. 

Before we go further, I think it's important to share that I'm not a math guy.  I did fine with it in high school, but never much enjoyed it.  I took two math classes in college because it was the minimum to satisfy the core curriculum requirement: intro statistics, and intro calculus.  I scraped together a B in stats, and got a C- in calculus.  (In my defense, it was an 8 AM three times per week during a less responsible era in life.) 

FIP and xFIP rely on the same fundamental numbers that people use in the course of normal baseball discussion: HR, BB, HBP, K.  There's no magic, and the formula (which can be seen in the above link to the Fangraphs glossary) is pretty easy to grasp: HRs, BBs and Ks are appropriately weighted and divided by innings pitched, and a constant is added to give the final product an ERA-esque look.  It's only slightly more complicated than commonly referenced staples like OBP and SLG.  They do some interesting math stuff that wouldn't have occurred to me, but, even with my very limited abilities, I can understand it.

After the jump is a video - humorous, mildly insulting, but to the point - that illustrates how FIP works.  If you haven't seen before, you should watch it.  If you have seen it, you know it's worth watching again.

It was created by Bradley Woodrum, is a cinematic masterpiece, and is plenty safe for work:

FIP: A New ERA (via homebodywithalaptop)

Behind ERA

ERA serves a purpose, but fans of advanced metrics believe it serves a different purpose than what people have assumed for a long time.  The argument is that it's more of a team statistic than a measurement of a pitcher's performance.   Earned runs are the result of a person hitting a home run, or by reaching base (hit, walk, HPB, etc) and advancing to home when another batter(s) has successfully reached base, by stealing home, or by sacrifice.  While it's hard to get past the idea that a pitcher is responsible for any earned run that crosses the plate, but the pitcher is solely responsible for very little within that dynamic.

For years, color commentators and sportswriters have assured us that it's the pitcher's job to prevent runs from scoring.  Yet it's as much the General Manager's job to prevent runs from scoring as the pitcher.   The pitcher can't prevent runs without mound without a competent defense, nor can the pitcher pick his own defense. 

Following the explanation in the video, FIP/xFIP tries to eliminate three big variables that are passively accounted* for in ERA: defense, managerial decisions, and scoring decisions/situational MLB rules.

*"passively accounted for" meaning that they are implicitly built into how a run is created, but are out of the pitcher's control

As indicated in the video, the ability of fielders is particularly influential in creating a run.  Sometimes a hit is a frozen rope single over the second baseman's head; sometimes it's a roller just to the left side of second and Yuniesky Betancourt is your shortstop.  One of those is always going to be a hit, and the other is highly variable depending on the fielder.

Imagine there's two outs and a man on second, and the ball goes past Betancourt for a hit.  A run scores to break a 3-3 tie.  The next man up hits a two-run home run, and the game is 6-3.   Now, go through the same scenario with Tulowitzki as the shortstop.  There's a good chance that Tulo gets to the ball and completes the out.  Even if Tulo doesn't complete the play, but gets to the ball when there's a man, he might stop the lead runner at third.  Then, JDLR gets three quick outs and the game remains tied. 

Sometimes, a good fielder can still make a mistake and be a better friend to the pitcher than the bad fielder.  Say, Tulo ranges to his left, dives for the ball, but throws it away.  It goes as an error.  Betancourt would never have gotten to the ball, and it goes as a hit.  So, Tulo saves the pitcher runs, even though he cost the team runs.

Further, the ways in which a manager creates a conditional outcome by changing pitchers.  How many times, as a real-time fan or a fantasy baseball manager, have you been left speechless when your starting pitcher is pulled and the reliever promptly gives up a game-tying hit, walk, home run, whatever, leaving your team and starting pitcher with a loss?  Throw an intentional walk and then a pull into that mix.  Should the starting pitcher be judged solely by the two additional earned runs and a loss because, with two out, two on, and up one run, he was pulled for a reliever who promptly gave up a home run?  

Also consider how the rules of scoring and scorers can impact a game.  Try to imagine the different ways that a scorer can interpret "slow handling of the ball" or "ordinary effort" as mentioned in Rule 10.12(a)(1).  Once you've deciphered everything in that section, read Rule 10.16, "Earned Runs and Runs Allowed."  Here's an example from that section:

(1) Peter pitches and retires Abel and Baker, the first two batters of an inning. Charlie reaches first base on an error charged to a fielder. Daniel hits a home run. Edward hits a home run. Peter retires Frank to end the inning. Three runs have scored, but no earned runs are charged to Peter, because Charlie should have been the third out of the inning, as reconstructed without the error.

This rule was one of my favorite discoveries as a fantasy baseball player.  If an error was made after the second out was recorded, a pitcher can give up five home runs and have none of them counted as earned runs.  It would be easy to see where the pitcher would be given a pass on the next batter after the error, but it's entirely arbitrary to absolve the pitcher for the rest of the inning. 

This is where the distinction between earned and unearned runs starts to erode the legitimacy of ERA as a measurement of pitchers: 1) it acknowledges that some runs team runs and others are pitcher runs, and 2) if the error is dependent on a good defender getting a bad break (getting to a tough ball but mishandling it) or a defender with bad range not getting near a tough ball allowing it to go for a hit, those 3-4 extra runs will be assigned differently.

What makes FIP and xFIP different?

FIP/xFIP make it easier to determine a pitcher's skill by eliminating a lot of the ambiguity that defines ERA.*  There's no worrying about a scorer impacting a pitcher's final line.  It doesn't matter what deserved to be a hit and what deserved to be an error, and whether that maybe hit/maybe error that allowed a run to score should count against the pitcher.  There's no worrying about whether Rafael Betancourt deserves to get punished because Matt Belisle came on and allowed his runner to score.  If, for example, Betancourt walked the guy, he still gets punished.  There's no worrying about what Zack Greinke would look like with a stellar defense. 

*As the video notes, FIP doesn't eliminate all of the variables; the umpire can still influence strikeouts and walks, some parks will still have more/less foul space or shorter/longer outfield fences, and there will still be some luck involved (though less than in ERA). 

All the pitcher is judged on are things reasonably and reliably within his control: 1) whether he can find the strike zone (BB and HBP); 2) whether he can keep the ball away from the batter's wheelhouse (HR); and 3) how many outs he can get without relying on his defense (K). 

Within those factors, three variables still impact the pitcher: 1) an umpire still has an impact on the strike zone, which could impact walks and/or the opportunities for a hitter within an at bat; 2) the dimensions of a park may make it easier/harder to hit home runs, and may impact the length of an at bat by giving more or less foul territory; and 3) good or bad luck may still impact how many home runs a pitcher gives up (unlucky placement), or whether he can get a strikeout (maybe the batter makes weak contact).

xFIP goes a step further, good or bad depending on the extent to which you think HR are dependent on good or bad luck, by using an expected number of HR allowed in the formula instead of the actual number of HR allowed.  The expected number of HR allowed is obtained by multiplying the number of FB by the average HR/FB rate (~10.6%), and weighting that number in the same way as FIP (this is what makes xFIP park adjusted).  HR/FB tends to greatly fluctuate from year to year, so there is some value.  However, it is possible for a pitcher to consistently underperform his HR/FB rate. 

How does it apply to the Rockies?

Fangraphs provides CC Sabathia as an example, but for instance, Ubaldo Jimenez's 5.1% HR/FB rate explains why his xFIP (3.73) was significantly higher than his ERA (2.88) and FIP (3.10).  xFIP would have expected double the home runs (21 vs. 10) in 2010.  For three straight years, Ubaldo has had an xFIP more than .19 above his ERA or FIP.  The spread has been as wide as wide as .85.   It's a significant disparity, and a lot of the disparity is tied to his low, low HR/FB in recent years, although the improving, but still rough BB/9 will also generally lead to a gap between ERA and FIP/xFIP.

FIP and xFIP aren't as kind to Ubaldo as ERA, but he may not deserve the kind treatment that ERA provides.  In 2010, his FIP and ERA were within the top 10% of all pitchers.  That's pretty impressive, especially considering his BB/9 was not the greatest, but FIP pretty much backed up ERA.  On the other hand, his xFIP was only within the top 25%.  This is mostly because his high GB% leaves few opportunities for FBs, and within that, relatively few FBs turn into HRs.  As was frequently discussed last summer when some arbitrarily dismissed Ubaldo's amazing start based on a huge ERA/xFIP disparity, this may be a bit unfair.  Ubaldo has had three straight years with a HR/FB rate significantly lower than the average.  xFIP can't take into account the type of contact made against Ubaldo, and that contact (especially in the first half of last season) may drive down his HR allowed.

Going Forward

I don't think of FIP and xFIP as needing to demonstrate their value relative to ERA, because they're based on an entirely separate philosophy.  However, as noted in the video, one of the supposedly superior aspects of FIP/xFIP is that they have a better predictive value going forward.  Next time, I'd like to demonstrate that value using Rockies pitchers.

In the meantime, please let me know if this exercise has been a complete waste, or if any of the illustrations are confusing.  Initially, it's difficult to believe in a new statistic that purports to measure a player in a similar and better fashion than a well known metric, but it really is a philosophical difference.  Once you consider what each metric really measures, I don't think you'll never look at FIP/xFIP and ERA in the same way.