2009 ERA-xFIP Splits
With 2010 nearly upon us, we’re being inundated with retrospectives recapping all that has occurred over the past year. Let’s add one more list to the mix: starting pitchers with the biggest difference between their Earned Run Average (ERA) and Expected Fielding Independent ERA (xFIP).
Why? Because ERA doesn’t give the most accurate assessment of pitching performance. As Colin Wyers at The Hardball Times showed this past summer, ERA does a rather poor job at predicting future ERA.
Wyers found that xFIP, which is based on a pitcher’s strikeouts, walks and a normalized home run per fly ball rate, has more predictive value than any other pitching metric. In other words, knowing a pitcher’s xFIP tells us more about his skill level and what his future performance may be than simply knowing his ERA would.
A pitcher’s ERA is subject to some factors largely outside of his control. The defensive skill of his teammates plays a part in the rate at which balls put in play are converted into outs. BABIP isn’t entirely out of a pitcher’s control: fly balls have a lower BABIP than ground balls, so fly ball-slanted hurlers tend to have a lower BABIP (the fly balls that aren’t caught are far more damaging, though). Plus, pitchers adept at getting infield pop ups generally have lower BABIP figures, as those are near-automatic outs.
However, team defense can have a large bearing on a pitcher’s ERA. Some get the backing of airtight defenses. The 2009 Seattle Mariners, for instance, saved 12 runs more than the average club per 150 defensive games. Other pitchers get a bunch of plodders and iron gloves. Take the Mets, who were over seven runs below average per 150 defensive games.
Maybe a groundball skips into the outfield for a hit because your second baseman has fall-down range. Or perhaps your world-class center fielder shows off his ridiculous D and runs down a sure fire extra-base hit. ERA doesn’t care either way. Franklin Gutierrez is awesome, but we’re trying to measure the skill of the pitcher here.
A pitcher’s ERA may also be distorted by abnormally high or low home run/fly ball percentages and rates of stranding runners on base. HR/FB rates for pitchers tend to stick around 10-12 percent. Some pitchers may be better than others at throwing from the stretch, but it’s usually a good idea to expect some regression if a guy’s strand rate strays too far from the 70-72% league average.
With all of that being said, here are the starting pitchers (100 inning minimum) with the biggest discrepancies between their ERA and xFIP.
First, the guys who outperformed their peripherals. These are the pitchers whose controllable skills weren’t as good as their ERAs would indicate.
Lower ERA than xFIP
Keep in mind, an appearance on this list does not mean that a pitcher is going to spontaneously combust. You’ll note the presence of Cain, Hanson, Carpenter, Kershaw, Santana, Greinke and King Felix, among other well-regarded starters. It just means that we would expect some regression moving forward. Greinke is a monster (3.15 xFIP), he’s just not likely to go all Pedro Martinez circa 1999 on a regular basis.
Here are the starters whose peripherals were better than their ERAs. These guys performed better than their ERAs suggest.
Higher ERA than xFIP
For some of these starters, you’ll notice strong strikeout-to-walk ratios dragged down by lofty BABIP figures and very low strand rates. Nolasco and Hamels are excellent bounce-back candidates.
When preparing for your 2010 draft, lists like these are a good place to start. But it’s also important to look beyond just the 2009 numbers, taking multiple years of performance into account whenever possible. Due diligence can help you avoid wasting a high draft pick on a ticking time bomb and can help uncover a diamond in the rough.
10
“Because ERA doesn’t give the most accurate assessment of pitching performance.”
Neither does xFIP
You really don’t have a clue what you’re talking about. xFIP doesn’t account for so many things. There’s a lot more to pitching than simply looking at one’s FB rate, BB/9 and K/9. Like the amount of subjective “line drives” a pitcher allows. How hard of contact they allow (velocity of batted balls) Etc etc. xFIP simply isn’t good enough. Not even close. Calling it the best pitching metric is ignorant and beyond silly. Oh, and uneducated.
i love how there are a net -23 votes for this comment, when it is factual, but flies in the face of what many ppl on here want to believe.
way to stay classy
…Except that, if you take the time to read Wyers’ article, it’s the best metric we have-it has the strongest predictive relationship, indicating it’s more skill-based. He actually compared it tRA, which does take into account batted ball types, and the correlation was weaker for tRA. If you have a better suggestion, I’m all ears. But calling my reasoning uneducated, while offering nothing more than an insult and an opinion, is sort of the definition of uneducated.
Surprised you even took the bait, though it is hard at times I bet!
This was a solid read, good analysis.
Nice article Dave. Don’t you just love guys like FG who sign up anonymously and bash your work with no backup to support it?
FG-what exactly is “etc.etc.” Oh, and what is better? and please tell us where we can find it with an explanation as to why it is better than xFIP. Not too much to ask when you have all the answers.
My snarky side is enjoying a little game of “what might FG be short for…”
David Bush is on that list every year thanks to LOB% issues. He fools someone every year.
Whenever you look at FIP minus ERA or xFIP minus ERA to see who should improve/regress it almost always boils down to inflated and deflated BABIP. That, and strand rates. Nevertheless, these lists are fun and useful.
Liriano looks like a good bet next year.
How much of a difference between ERA and xFIP can we generally accept as a result of defensive skills? It seems foolish to expect a significant regression towards xFIP for a pitcher who will largely be pitching in front of the same 8 fielders the next season. For instance, should we expect Mike Pelfrey to end up on this same list next year with more and more batted balls going past Castillo? Or is a .5 run difference too much to attribute to shotty defense?
Charlie,
You’re definitely right in saying that we should still consider the quality of a pitcher’s defense in trying to rank him for the upcoming season. An average, pitch-to-contact starter could turn in drastically different seasons based upon the range of the players behind him (see Millwood, Kevin).
It’s just hard to know how good or bad those defenses will be in advance, given the amount of roster turnover. I think it helps to get a feel for a pitcher’s defensive-independent skills. That way, we have a good idea of that guy’s own talents, should the circumstances around him change.
I’m no expert, so I have a question … doesn’t FIP because of it’s equation favorise strikeout pitchers over pithers with less strikeouts … + thing about pitchers having no influence on BABIP … ”don’t better pitchers tend to throw better pitches that are harder to convert into hits as opposed to worse pitchers”?
I’m not criticising, just want to hear your explanation.
Great article & analysis!
It would be very interesting to go back over several years to see the E – xFIP diffs for the top 25 MLB pitchers (based on ERA) to see the trends.
A sort of compare the greats by decade to find out if modern day pitchers are better or worse than earlier periods -or- if one can compare across periods.
Thanks again & keep up the great work you guys do to make baseball analysis really fun & illuminating.
On the lower era side we have two dodger, two cardinal and two braves pitchers in the top twenty-roughly twice per team what a random distribution would predict. Similarly, on the lower xFIP side we have two Marlins, two Brewers (the same Brewer pitcher who are waiting for an infield defense every year), two Astros and two Rockies.
I used xFIP last year and believe it is a highly predictive metric. What I am wondering is that if you could factor in team UZR ratings in the xFIP calculation, would guys like Parra and Bush (both of whom were also on the list year) disappear?
Similarly, for pitchers whose BABIP and HR9 are (or will be, a murky area indeed) consistently low year over year (generally, the higher velocity, more dominant pitchers) how good is the regressor indicator? I am prepared to believe Millwood is not a good pick for this year, but guys like Kershaw, King Felix and Carpenter, I might take them anyway.
Chris,
xFIP really isn’t affected by team defense-that’s why it’s expected fielding independent ERA. Pitchers aren’t rewarded or penalized based upon the quality of defense behind them. Guys like Bush and Parra, with decent peripherals but lousy BABIPs, aren’t dinged more than another guy with similar K/9, BB/9 and HR/FB numbers but a lower BABIP.
Pitcher BABIP tends to show a very weak correlation from year to year:
http://www.baseballprospectus.com/article.php?articleid=8932
As for home runs, xFIP doesn’t penalize a guy with a low HR/9, per se. It simply normalizes a pitcher’s HR allowed per fly ball. A ground ball pitcher, who gives up fewer fly balls overall, will still get credit here. For instance, let’s just say that Pitcher A gives up 100 fly balls and pitcher B gives up 200 fly balls. If they both have a HR/FB% of 10, then Pitcher A gives up 10 homers and Pitcher B gives up 20 homers.
I would definitely bet on some regression for all three guys you mentioned. That’s not to say they’ll be bad by ANY means-those are three of the better pitchers in the game. But ERAs in the two’s (and low-two’s in the case of Carp) just aren’t sustainable on a year-in, year-out basis.