xFIP Seasonal Adjustment
I made a change last night to xFIP that adjusts the major league HR/FB rate on a seasonal basis. Previously, the average was set at 10.5%. In recent years the average HR/FB has been somewhat lower than 10.5%, which is what prompted the change.
2007 – 9.7%
2008 – 10.1%
2009 – 10.1%
2010 – 9.4%
Overall a player’s 2010 xFIP will tend to be a little bit lower, at most around 0.20 runs, but generally more in the 0.05 to 0.10 range.
David Appelman is the creator of FanGraphs.
David, what HR/FB rate are you now using? I am guessing that you are using the league average over the 2007-10 period.
I’m using a 1 year adjustment right now, just like the FIP constant. Given how people generally use xFIP and compare to ERA, I think sticking with the one year average might make the most sense.
How dare you.
Since it has become abundantly clear that some pitchers can have a consistently lower HR/FB, maybe it’s time to re-think the whole equation.
That could be said about a whole lot of things in FIP/xFIP. (BABIP, HR/FB, etc)
Also, lets stop pretending a statistic that has OUTS(via IP) as a variable is defense independant.
That’s why I used expected outs. Unfortunately, however, some outs-producing events are hard to quantify without double counting…
Just because there are a few pitchers on the Giants that have posted abnormally low HR rates recently, doesn’t negate the fact that most of the pitchers in the league fluctuate around the HR/FB average from year to year, and don’t show any ability to control their HR rates. Try not to sound so homer-y and obnoxious next time without data to back it up (and especially since the data disagrees with you).
And even for these very few who show some small amount of control over their HR/FB rate, xFIP is still useful.
The data has been discussed ad-infinitum. It backs me up.
“recently, doesn’t negate the fact that most of the pitchers in the league fluctuate around the HR/FB average from year to year”
Again, the fact that the majority of the league fluctuates around the average does not mean that there is no repeatable skill.
And no, the data actually agrees with DrBGiantsFan.
the data backs you up?
ok, all you guys with a hard-on for matt cain – here is the challenge –
making sure to consider:
– the hr/fb numbers posted above
– home park
– in division offenses (nl west offenses rank 2,10,11,14 in woba for 2006-2010, with 2 being the rockies and their home park)
– facing opposing pitchers
– selection bias due to using pitchers that are both talented and healthy enough to pitch the last 4-5 years
accounting for all of that and anything i forgot, run the data to show how many standard deviations out cain’s performance is over that time compared to all pitchers. if it is at a value that suggests he is an extreme outlier and not a product of chance, continue adding percentage points of hr/fb until he is within a range that suggests it is just random. go ahead and call that added percent his true talent level for hr/fb prevention. i’d be surprised if it came out to be even 1%.
Dutchy,
I believe that calculation was done in one of the earlier articles about Cain’s HR/FB rates. He still come out to be a pretty significant outlier. More than a STD Dev for 5 years in a row.
dutchy that stuff has been done and done and redone, and it still comes out in cain’s favor… get off your horse and admit the stats might not say everything, you wont get hurt
Actually, I don’t think I have seen anyone do quite what Dutchbrowncoat suggested. The only true talent estimates I have seen for Cain’s HR/FB % skill were based simply on regressing his rates to league average according to his fly ball sample. What he said might not matter, but I am not sure we can just hand wave some of those factors away either.
If you have just one season’s worth of stats, you’d pick xFIP (or SIERA, which has a similar approach) over anything else.
http://www.baseballprospectus.com/article.php?articleid=12844
If you have more data, then use a projection, of which plenty of solid choices are available (and even get updated in-season).
How in the world is it “abundantly clear?”
Saberfury,
You might want to go back through the archives of fangraphs.com and read the work that Wolfsberger(I think it’s him, anyway) and Dave Cameron have done looking at Giants pitchers over the last decade or so that Dave Righetti has been the pitching coach. Whether it’s Righetti, or the type of pitcher the Giants tend to acquire or winning-the-lottery-level-good-luck, Giants pitchers have consistently kept their Park Adjusted HR/FB significantly below league average.
I would add that there are other pitchers scattered around baseball who have done it, enough so it becomes a stat that is unreliable to include as a regression target in a universal assessment of pitching performance.
i really wish my comrade would give it a rest. yes, you have a crush on matt cain, we get it.
Nominated as my favorite post of 2011.
Sorry, it’s not just Matt Cain. It’s pretty much the entire Giants pitching staff for almost a decade! It’s not just the Giants either. There are other pitchers scattered around baseball that consistently beat their xFIP. Zambrano comes immediately to mind, but there are others. Kershaw may be another. Most of them fit the profile of hard-throwing flyball pitchers.
You don’t like xFIP, fine, don’t use it.
Leave the poor horse alone, it’s long past dead.
it seems my associate must have got his PhD in thread hijacking.
He didn’t hijack it, the entire post was obviously constructed as a veiled shot at Cain from the start.
Fangraphs hates the Giants.
SSG,
No, fangraphs.com does not hate the Giants. Many of the writers do, however, consistently underrated Giants pitchers by clinging to a flawed stat.
Zambrano’s not a flyball pitcher and he used to be an extreme groundballer
Jeffrey,
I believe I said MOSTLY hard throwing flyball pitchers. As for Zambrano, I believe that for the last several years when he’s been beating his xFIP, his GO/AO has been only slightly above 1.
Well, regardless of the correctness of the system, this is why I’ve updated the xWHIP calcuations to allow you to modify HR/OFFB rate….
Verlander fits that bill(hard throwing flyball pitcher) and he too has kept his HR rate down for his career. Yeah he does pitch in a friendly HR park but his splits show that he has kept it down on the road as well(7.9 career HR/FB% at home vs. 7.4 on the road)
This had been bugging me for some time. Thanks for correcting it. Just wondering but what constant is used in calculation of FIP/xfip to add to the 3*bb/2*k/13*hr mumbo jumbo? I know it’s usually 3.2 but does that vary too?
Yep, that varies too. In 2010 it was set at 3.08-ish.
How do you calculate the constant?
It’s just whatever makes the league average xFIP equal to the league average ERA.
Jeffrey,
Multiply the coefficients by the frequency of the event, divide by innings pitched and solve for whatever constant you need to match league ERA.
For 2010 it would be (13*4613 + 3*1611 – 2*34306) = 39690; 39690/43305.1 = .92; 4.08 – .92 = 3.16.
The thing I still don’t get about the whole Matt Cain thing is even if you give him full credit for the HR prevention he’s still just a slightly above-average pitcher. Is that what you want us to admit? Fine. He’s okay. Congrats.
No, he’s quite a bit better than slightly above average. Probably one of the 20-25 best starters in the game.
WAR says more like 30-35.
Dial down the homerism.
Last 5 years: 14th in WAR
Last 4 years: 20th
Last 3 years: 20th
Last 2 years: 21st
Last year: t-27th
has made >30 starts all 5 years. how does that equate to 30-35th in WAR?
“has made >30 starts all 5 years. how does that equate to 30-35th in WAR?”
livan hernandez has done that for 13 straight years. perhaps we should consider him among the best pitchers in the game? he is a solid pitcher in a pitcher’s park who has had the good fortune of being healthy for five years straight. just being healthy that long will move him up a war leaderboard. he is a good pitcher, but let’s not get too far ahead of ourselves here.
Htpp,
Do you really need the enormous flaw in your logic to get pointed out?
Hes putting up 4WAR a year using FIP/xFIP. If you use actual performance, its closer to 5 WAR a year. He’s an elite player.
using baseball reference’s war, based off of era (“actual performance”), he was 26th for pitchers with 3.9 war.
is he an elite player? eh…
is he is an elite #2 pitcher? yes.
Dutchy,
There are 30 teams in baseball. If Cainer is #26, how does that make him a #2? Even if he is an “elite #2”, whatever that is, that’s quite a bit better than “slightly above average” which is what some idiot said earlier in the thread.
MATT CAIN
MATT CAIN
MATT CAIN
John Lannan is another guy. I know, the boo birds will look to last year, but he was decidedly better after he came back from Double-A and he has been beating his xFIP the other two years. I would bet that he’ll beat the xFIP projection this year too.
I wouldn’t be surprised to see some extreme control pitchers get past the flawed xFIP statistic. Look up Greg Maddux for example….
What is Maddux’s career xFIP, out of curiosity.
Surely you aren’t comparing his career ERA to just the last seven seasons of xFIP listed here?
You mean the Greg Maddux who had an xFIP of 3.73 from 2002-8 but whose “extreme control” allowed him to have a 3.92 ERA over the same period?
The Greg Maddux who maintained insanely low BABIP’s throughout his peak years. In fact, his CAREER BABIP was .280! I’m pretty sure if he was churning out BABIP’s of .240-.260 now, there would be writers on this site proclaiming he was bound to regress because his BABIP was so low.
I’m talking about the statistically undervaluing bias of pitchers who have consistently have had low BABIP’s and are clowned on this site with desiring dismissals about eventually regressing to the mean. If you take a look at the players, some ‘seem’ to be able to control where the ball goes, or more exactly, the type of swing the batter has on the ball. The FB issue for Cain is another example.
Maddux has a career ERA of 3.16 while the “flawed (x)FIP statistic” gives him an equally great, and very close, 3.26. And that’s not even accounting for the fact that he pitched in the NL, which has a lower BABIP due to the pitcher hitting, and that he pitched in front of a great defense during his career.
http://www.baseball-reference.com/players/m/maddugr01.shtml#pitching_value::none
Notice he had 64 runs of defensive help in his career.
Not sure why anyone would ever use Maddux as an example of FIP’s flaws since it does a very good job of capturing his value yet for some reason commenters on ESPN and now seemingly Fangraphs constantly use him to “disprove” FIP. Hmph.
Toffer,
Just looking at Maddux’ career numbers doesn’t really give an accurate picture since he was a far different pitcher in the last 6-7 years of his career. During his peak years, his ERA outperformed his FIP on a very consistent basis and by a significant margin. I don’t see HR/FB numbers for those years, but he put up BABIP’s that were consistently in the .240-.260 range. In the waning years of his career, his BABIP’s were conistently higher and his ERA tended to underperform his FIP and xFIP.
It is highly likely that Greg Maddux possessed a skill during his peak years that allowed him to consistently hold hitters to insanely low BABIP’s. No way is it all due to defense/league factors.
One consequence of updating the numbers (which seems long overdue, but thank you for doing it) is that lots of old articles will now have incorrect xFIP numbers in them. Some conclusions will probably be affected a little bit; we’ll have to watch for that whenever we want to cite archived articles.
14th in WAR last five years is pretty close to elite, and far from “slightly above average”. Also, Cain would like to thank Dave for this adjustment
Randy Johnson is 1st in total WAR over the last 20 years. He must be the best pitcher in the league, no?
SSG,
RJ might have actually been the best pitcher of the last 20 years over the course of his career. Difference is he’s now retired and Cainer is still pitching so that was a pretty dumb response there.
No, using cumulative totals over a span of years to “prove” that someone ranks better than where they should be is the “dumb” response.
If someone ranks from the mid 20’s to the mid 30’s each year, how does that make them the 14th best?
“If someone ranks from the mid 20?s to the mid 30?s each year, how does that make them the 14th best?”
By that logic I guess Roy Halladay can’t be consider the best pitcher in baseball because each year he usually ranks around 3rd or so in WAR.
SSG,
It’s called consistency, my friend, and durability. A pitcher who is top 25 WAR for 5 years in a row is way more valuable than one who is top 10 1 year and stinks the other 4. Not that top 25 is a bad thing in itself. It’s good enough to be a #1 pitcher on a lot of teams.
“If someone ranks from the mid 20?s to the mid 30?s each year, how does that make them the 14th best?”
Because very few people are able to maintain that level of performance. You have guys who may be 10th one year, but can’t crack the top 30 any other year.
Being valuable every year is better than being valuable some years.
that’s a really silly argument, small sample goodness.
cain isn’t an ace but he is an elite #2… what in god’s name is wrong with that. he’d be the ace on a bad team.
The argument isn’t mine, I simply pushed the “silliness” to its fullest.
Well, you either made a dumb response or you pushed the satire envelope a bit too far for most of the readers here to catch on.
By most you mean one or two Giants fan with a reason-clouding bias?
SSG,
I think you used the key word here yourself….”silliness.”
As much as its “silly” to count total WAR, isn’t it also a little ludicrous to completely discount lack of injury when looking at a players overall worth and “elite” status. I would much prefer a player who stayed health for 5 years and posted a solid WAR every year, than one who interspersed great WAR with consistent time on the DL. Especially as it may only take 1 DL stint to bring that WAR down.
i.e. I think playing for 5 years without injury shouldn’t be discounted as irrelevant to his caliber as a pitcher.
The rule never should have been interpreted as “all pitchers have the same underlying HR/FB rate, subject to ballpark adjustments”, which is how a lot of people seem to interpret it.
That would be a remarkable result – no matter whether a pitcher threw at 80mph or 100, controlled or wild, over-arm or side-arm, sinker balls or rising fastballs, he’d have the same rate. Not impossible, but it would be remarkable. As a hypothesis it would need strong statistical evidence to show it applies to the vast majority of pitchers – which doesn’t exist to that extent.
Yet somehow, we’ve got into a stage where the burden of proof lies on the pitcher to escape the rule, rather on the statistician to show the rule is wide ranging enough to apply indiscriminately.
People seem to want to put stats in two baskets – either stats that are unreliable and therefore borderline useless (ERA, RBIs, Wins, Batting average over smallish samples, etc) and fantastic stats that are wonderful and can’t be contradicted by eyeballs (FIP/xFIP, WAR, incremental hot dog sales, etc). xFIP lies in the second category.
Likewise, when somebody like Cain appears to break the “rule”, his supporters always want to place him entirely outside the scope of the rule – i.e. they are just willing to accept his historic performance is his true level, with no regression required at all. This is just as silly as expecting him to regress all the way back to league average.
I’m not a statistician, though I do have some elemetary understanding of probability, so please correct me if I’m wrong. Cain’s variations from the rule don’t strike me as significant enough to disprove the rule.
A player consistently being 4-6 standard deviations off the mean doesn’t make you think the assumption of randomness (and thats all it is, an assumption), might possibly be wrong?
Cain’s variance from the norm is far enough that is is much more likely due to a skill he possesses than to chance. I don’t know if you can ever “disprove the rule.” People do win the lottery after all, although I doubt that anyone has won it 5 years in a row.
Is anyone here going to suggest that Gavin Floyd is one of the 20 best starting pitchers in the league? Or that he’s an ace? Because he’s a pretty similar pitcher to Matt Cain; in fact, the numbers suggest he’s actually been a win and a half better.
The idea that Matt Cain is an ace is absurd. As has been said, he’s a great #2, just like Gavin Floyd.
That should say a win and a half better over the past two seasons
Read this: http://www.baseballprospectus.com/article.php?articleid=14603
The gist of it is that career HR/contact correlates better with HR rate than career FB/contact. The reason why xFIP is slightly more predictive of ERA than FIP over a small sample is the fact that normalizing HR/FB regresses a pitcher’s HR rate to the mean in a somewhat roundabout way. The best way to predict a pitcher’s FIP and hence ERA using a small sample of data is not xFIP but rather a projection system like RoS ZIPS.