Targeting Starting Pitchers Based on xFIP/FIP Differentials
Eno Sarris said not to look at HR rate. He said it and I’m going to listen. However, FIP and xFIP are not HR rates, and I’m going to look at that. Too often we assume that others know, or we actually know, what a statistic represents. We hear it, we think it, we know it. But, take a moment with me to reinvigorate our understanding of these two very important statistics.
FIP gives us an idea of how a pitcher performs regardless of who is playing defense behind him. It accounts for strikeouts, walks, hit by pitches, and home runs allowed. FIP gives us a better understanding of how a pitcher is performing than ERA. xFIP tells us all the same but accounts for the volatility of the HR rate. Quoting from our very own FanGraphs glossary, xFIP is:
calculated in the same way as FIP, except it replaces a pitcher’s home run total with an estimate of how many home runs they should have allowed given the number of fly balls they surrendered while assuming a league average home run to fly ball percentage (between 9 and 10% depending on the year).
If we take the difference between the two (xFIP-FIP) we will be able to see who is underperforming or overperforming. We can see who has fallen prey to the HR rate bad luck that Sarris told us not to look at, and who has been straddling the line between a warning track fly ball and an official dinger. Here are the top 10, most negative xFIP/FIP differentials among all qualified starting pitchers this season:
| Name | K/9 | BB/9 | HR/FB | FIP | xFIP | Diff |
|---|---|---|---|---|---|---|
| David Peterson | 10.73 | 2.96 | 35.7% | 4.51 | 2.85 | -1.66 |
| Kenta Maeda | 8.79 | 2.20 | 25.9% | 5.27 | 3.76 | -1.51 |
| Mike Foltynewicz | 8.37 | 2.16 | 21.4% | 5.57 | 4.28 | -1.29 |
| Luis Castillo | 7.28 | 2.43 | 23.8% | 4.69 | 3.75 | -0.94 |
| Lucas Giolito | 12.03 | 3.82 | 21.4% | 4.25 | 3.32 | -0.93 |
| Bruce Zimmermann | 6.60 | 3.00 | 19.4% | 5.88 | 4.96 | -0.92 |
| Antonio Senzatela | 5.46 | 2.73 | 20.8% | 5.20 | 4.44 | -0.76 |
| Yusei Kikuchi | 7.92 | 3.23 | 20.8% | 4.64 | 3.90 | -0.74 |
| Adam Wainwright | 9.70 | 2.36 | 18.8% | 4.27 | 3.65 | -0.62 |
| Adrian Houser | 6.16 | 3.23 | 21.1% | 4.51 | 3.91 | -0.60 |
What we have here, in a nutshell, are pitchers who have given up home runs and the announcer said, “In any other park that would have been an out.” So, perhaps these pitchers are frustrating current owners and could be trade targets. But certainly, a few of them are on the wire, just waiting for a claim. Here are a few to note:
- #Don’tDropMaeda. His high HR/FB rate has juiced the FIP as he’s been serving up meatballs with freshly grated parmesan. Both his changeup (17.6%) and his slider (32.7%) have a CSW below where he’s finished each season in his career, but Monday night’s showing was very promising that he is, at least, on the track to get it dialed in.
- Bruce Zimmerman (excuse me a second while I start an Ottoneu bid auction) is an arm to keep an eye on along with Antonio Senzatela as interesting options to start when on the road. Though it’s a low sample size, Zimmerman’s HR splits are already favoring the road.
- Don’t sleep on David Peterson. His CSW is up from 27.7% in 2020 to 30.8% in 2021 and he is rostered at only 8.8% in ESPN leagues.
There are two sides to every coin and this side shows us who is getting lucky. The announcers say, “Well, he got lucky there as that ball just barely missed making it out on a windy night here in…” It may be time to sell high or to at least curb your expectations on these qualified pitchers with high differentials:
| Name | K/9 | BB/9 | HR/FB | FIP | xFIP | Diff |
|---|---|---|---|---|---|---|
| Matthew Boyd | 6.06 | 1.77 | 2.1% | 2.97 | 4.98 | 2.01 |
| Kyle Gibson | 7.29 | 2.97 | 0.0% | 2.57 | 3.95 | 1.38 |
| Nathan Eovaldi | 8.31 | 1.82 | 0.0% | 2.13 | 3.50 | 1.37 |
| Gerrit Cole | 14.81 | 0.72 | 2.9% | 0.48 | 1.78 | 1.30 |
| Ryan Yarbrough | 7.02 | 1.62 | 5.4% | 3.14 | 4.32 | 1.18 |
| Carlos Martinez | 4.95 | 2.23 | 5.1% | 3.80 | 4.98 | 1.18 |
| Danny Duffy | 10.20 | 2.70 | 6.1% | 2.61 | 3.69 | 1.08 |
| José Urquidy | 7.15 | 2.12 | 8.0% | 3.84 | 4.91 | 1.07 |
| Matt Harvey | 6.39 | 2.61 | 6.1% | 3.50 | 4.54 | 1.04 |
| Taijuan Walker | 9.00 | 5.00 | 4.5% | 3.37 | 4.33 | 0.96 |
- I will admit to riding the Boyd train and toot, toot, tooting all the way into the station. Every night I see the check next to his name is a night I’m expecting a bad start. But, it hasn’t really happened yet. Four-seamer, changeup, slider with a very low HR/FB rate in cold Detroit means…what? Well, it means this is not sustainable. But, Boyd’s HR/FB rate spiked in 2020 and has come back down to start 2021. If he can find his slider again and sustain his changeup success (CSW currently sits a 34.7%) he may just be a rotation mainstay in deep leagues.
- People have been high on Kyle Gibson, Danny Duffy, and Jose Urquidy, and why not? They have been performing. But, their xFIP would suggest that they have been getting lucky on fly balls so far this season and now could be a great opportunity to use them in a trade to fill a need.
- It is kind of amazing to see Gerrit Cole on this list given how great he’s been so far this season. Carmen Ciardiello recently analyzed his start to 2021 and how it’s been overlooked due to panic in the Big Apple. But, don’t forget about Ben Clemens’ piece last summer, making note of Cole’s increased hard-hit rate in 2020 (37.1%.) He is currently giving up fly balls at a 47.3% rate (4th among qualified starters), along with a hard-hit rate similar to 2020 at 33.3%. #Don’tDropCole, obviously, but don’t be surprised when the long ball starts to bite him a bit.
Expected and actual differentials can be interesting to analyze throughout the season, but don’t let it be the only thing you look at when deciding to drop or claim. One fly ball on a cold and overcast day in April could easily be a nacho-smattering home run in July. Just be sure you’re on the right side of the aisle when hot cheese goes flying through the air.
FIP and xFIP are great descriptive metrics!
gibson is in the top ~20 in xERA last i checked. his ability to induce weak contact via ground balls has been impressive this year
Lucas, are you sure this statement is accurate? “What we have here, in a nutshell, are pitchers who have given up home runs and the announcer said, “In any other park that would have been an out.””
I thought xFIP just meant that if a pitcher gave up… say 70 flyballs, then… say 10% (league average), or 7 of those flyballs will, on average, be HRs. And it doesn’t take into account EV, LA or distance travelled on those 70 flyballs. So, if batters are launching rockets against a certain pitcher, then those extra HRs (above league average) he is giving up are well deserved – and the opposite is true if batters are hitting more lazy (can of corn) flyballs (below league average) against a certain pitcher.
You are correct.
You’re correct. It’s a regression to the mean. I suppose putting it in a nutshell and using the announcer to explain the advanced metric oversimplified it. However, I would argue that random events, such as wind gusts, air temperatures, and vicinity to hot nacho cheese are baked into that 10% league average.
As cool as Fangraphs is, this is a bunch of crap. Sorry, author. xFIP = the K strikeout stat. It’s so obvious, and then it gets all dressed up.
Nope.
Ever wonder why there aren’t any soft tossers anymore? I don’t.
Kyle Hendricks would like to have a word with you in his office…
This article feels so….2012. Unless you can show the the diff. between xFIP and FIP somehow adds accuracy to the RoS Steamer projections, I’m not sure what the point is.
So, I’m not going to complain about xFIP like everyone else here without at least giving something to back it up. What I will do is point you to the pitcher pop up leaderboard on StatCast: https://tinyurl.com/2021PopUps
Ryan Yarbrough is #2, with 15 pop ups. Matthew Boyd, #3 with 14. Jose Urquidy, tied for #4 with 13. Kyle Gibson, tied for #10 with 10. Gerrit Cole and Carlos Martinez, tied for #17 with 9. Eovaldi has 7, Duffy/Taijuan have 6, not fabulous, but still pretty good. The point is, the xFIP calculation includes pop ups in the HR/FB calculation, even though precisely zero of the 50,085 pop ups in the StatCast era have gone for HR. In fact, only 1,094 of them (2.18%) have gone for hits. So, for every pop up, you can take 0.06 off of xFIP for someone who has thrown 30 innings, and 0.045 xFIP for someone who has thrown 40 innings, just accounting for the fact that they don’t ever go for home runs, let alone that they’re basically automatic outs. Factoring out Yarbrough’s pop ups, for example, gives him an xFIP of 3.60. John Means is the biggest benefactor, his xFIP would go from 3.48 to 2.78!
Pop ups are a repeatable skill, generally speaking. Marco Estrada has 4 of the top 5 seasons on the StatCast leaderboard when you look at pop ups/pitch (min 2000 pitches, 570 qualifying seasons). Matthew Boyd has two of the top 20. I am in agreement that xFIP is an antiquated statistic, because it does include pop ups and doesn’t evaluate quality of contact. But I don’t want to just come on here and say “duh, xFIP bad” without having an explanation.
Excellent, thanks, because even the fact that there can be two different metrics FIP and xFIP is kind of weird. The author is saying that the difference is pretty much “fielding-independent” vs. “field-independent” but that would be kind of silly to look at even if it were true.
John Means has a change-up that seems to defy gravity. A pitch like that is a spanner in the works for advanced metrics.
No…that’s not what I’m saying. But, I can see how my announcer comments made it seem that way. There aren’t necessarily park adjustments in the calculation. xFIP simply strips out unstable and volatile home run rates by:
“replac[ing] a pitcher’s home run total with an estimate of how many home runs they should have allowed given the number of fly balls they surrendered while assuming a league average home run to fly ball percentage (between 9 and 10% depending on the year).”
““replac[ing] a pitcher’s home run total with an estimate of how many home runs they should have allowed given the number of fly balls they surrendered while assuming a league average home run to fly ball percentage (between 9 and 10% depending on the year).”
John Means currently has a HR/FB of 10.2%, a FIP of 2.97 and an xFIP of 3.45. Just trying to understand what is going on here.
Yea, that’s a good one. Hmm….I’m going to investigate.
This was fun to work out by hand. Means’ FIP (2.97) calculation:
((13 * 5 HR + (3 * 10 BB) – (2 * 50 Ks) / 46 IPs )) + cFIP
Means’ xFIP (3.45) calculation:
((13 * (49 fly balls * 13.7% league average HR/FB%) + (3 * 10 BB) – (2 * 50 Ks) / 46 IPs )) + cFIP
The xFIP calculation is showing us that when we use league average HR/FB% (13.7%) against his 49 fly balls, he would have been expected to give up almost 2 more home runs. Means’ fly ball rate is 46.2% compared to the league average 35.2%. It also should be noted that the league HR/FB% is higher than what I quoted earlier, resulting in a higher xFIP as well.