WAR Updates!
We’ve made some slight changes to the way WAR is calculated. The changes will impact players at most, about 0.5 wins, but the vast majority of individual player WAR will remain pretty unchanged.
1. Position player WAR now includes ground into double play avoidance (or lack thereof) with the stat wGDP. This will impact the very best players about +/- 0.5 wins. All but a few will fall in the +/- 0.2 win range. wGDP is also available as a separate stat in both the leaderboards and the player pages.
2. UBR has been updated to include advances and outs on WP and PB, as well as a few minor changes. Big thanks to Mitchel Lichtman for this and the wGDP inclusion.
3. Pitcher WAR park factors have been changed to FIP specific park factors. You can see all the new park factors under FIP in the FanGraphs Guts! Section. FIP park factors in general are considerably more compressed than runs park factors, so pitchers in extreme parks will be less impacted. Thanks goes out to Noah Baron for the suggestion.
4. We’ve adjusted the pitcher WAR league baseline to better even out the leagues.
David Appelman is the creator of FanGraphs.
A little confused about FIP specific park factors (im a newbie). Doesnt 2/3 of FIP have nothing to do with ballparks? Shouldnt every park be neutral in terms of K and BB? What am I missing? (sincere question)
Ballparks tend to not vary as much with regard to K and BB tendency, but they aren’t completely neutral either. For example, lighting conditions and amounts of foul territory can affect K frequency.
But are those small factors enough to be weighed as heavily as they are in FIP based park factors?
For sake of not repeating myself, I made a reply to John below. Just trying to make sense of it.
I think that Noah’s main (and valid) issue has been that Fangraphs uses a FIP-based WAR, yet used to use a runs-based park factor in calculating it. That short-changed a lot of pitchers on certain teams, such as the Mets, by penalizing their WAR on a runs basis despite the fact that their FIPs were good. Basically, the issue is consistency, and fWAR was relatively useless until tat consistency issue was fixed.
Sticking with Mets, Citi field has been ranked 22 or worse in park factors based on runs every year it has existed( 6 seasons). We can safely assume, regardless of k, bb, or hr rates that Citi field absolutely suppresses runs. 6 years in a row it hasn’t just been below avg, but near the bottom. That’s enough data to trust. I think it makes complete sense that Mets pitcher WARs are hurt despite good FIPs.
Idc if based on fip citi field isn’t a pitcher park when rate of run scored there has been in bottom 8 six years in a row. It is a pitchers park.
Scoko: that only makes sense if you’re going to assess pitchers based on how many runs they give up. You’re arguing that 6 years of run data is better than 6 years of K, BB, and HR data to adjust pitching performance that solely looks at K, BB, and HR
I am absolutely arguing 6 years of run data is better than 6 years of fip data. If 6 years in a row a team is considered a neutral park based on fip yet 6 years in a row the parks run rates rank in bottom 8 to me that just proves that the fip data is missing something and telling the wrong story.
I can’t fathom arguing Citi field is a neutral park bc fip says so when every single year the run rates disagree. The run rates disprove the what fip claims, not the other way around. If Citi was a neutral park than the run rates over 6 years would reflect that.
I think you’re missing the point. 6 years of run data ARE better than 6 years of FIP data for regressing ERA. But that’s not what we’re talking about. fWAR is a FIP based WAR, so whether the park is run-neutral or not is irrelevant, because runs are not part of the fWAR calculation.
If you are a truck driver, you want to know how heavy a brand of cereal is by the box. So you weigh a bunch of boxes and get an average and now you can just multiply the number of boxes you have plus the average and have a very reliable figure for how much weight of cereal boxes you have. Now imagine someone took that average and multiplied it by the average calories per gram of the cereal to come up with the total calorie content in the truck. That is essentially what Fangraphs has been doing, basing their calculation on a very valid measurement, but not the one that actually gets them the information they are looking for. They have now stopped measuring the cardboard the same as the cereal.
Citi field is a pitchers park for runs, but not for FIP. It’s a thing! At least according to Fangraphs, and some people who have looked very, very hard at it (read Noah Baron’s community posts.
scoko, the problem is that fWAR is based on FIP, not runs. In Noah’s article, I think it said that the run factor was 95 and FIP was 102. So think of it this way (ERA and FIP are just made up numbers below).
Let’s say you have a completely average pitcher, and the league average ERA is 3.50. If he pitched in Citi Field, he might have a 3.30 ERA because the park was stingy in runs. Since Citi Field inflates FIP though, his FIP might be 3.60. That’s all in line with league average.
However, fWAR would look at his performance and say “you pitched in a great pitcher’s park that suppresses runs by 5%, but your FIP (3.60) was worse than league average (3.50), so you’re a bad pitcher!” We know that’s not right though, because everything is average, not bad.
I think my point is being missed. My point is simply if we have enough data to undeniably say Citi field suppresses runs than who cares what their fip factors are? If fip says a park is neutral isn’t that another way of saying the run rates for that park SHOULD be neutral? So now we judging Met pitchers against a backdrop of a neutral stadium, which is going to make their WAR better than it should be, because, based on the undeniable amount of run data we have since Citi has existed it is absolutely not a neutral park.
citi field suppresses runs overall but doesn’t suppress home runs, which is what fip cares about
Last attempt: fWAR does not look at runs that a pitcher gives up. It does not matter if Citifield suppresses runs. FIP does not look at runs, and fWAR does not look at runs.
Fangraphs used to regress FIP based on the runs factor of the stadium. They were convinced that that was a mistake.
Fangraphs will not be “judging Met pitchers against a backdrop of a neutral stadium, which is going to make their WAR better than it should be.” Rather, they will be judging Mets pitchers ability to strike batters out, avoid walking batters, and avoid home runs based on Citifield’s tendencies to aid pitchers in striking batters out, walking batters, and avoiding home runs, rather than judging their ability to strike batters out, avoid walking batters, and avoid home runs based on Citifield’s tendencies to aid pitchers in stopping runners from crossing home plate by any means.
I understand making the change to make it consistent with WAR, but I think it fails in the purpose of park factors. Park factors is supposed to show how each park fairs in suppressing runs relative to the others. To me fip park factors is another way of saying “using fip to determine how a park should fair in suppressing runs”. If a park consistently suppresses more runs than their FIP park factors indicates then clearly you can toss out the fip park factors for that team. No different than if a given pitcher beats his fip significantly every year of their careers. Going by a form of WAR that views the Mets park as a park that should be neutral in run suppression, to me, is clearly giving you an inaccurate WAR. Even if it is fip compared to fip,I think you can toss that WAR out the window if the fip park factors so clearly tells an invalid story of run suppression.
fWAR is a defense independent pitching stat, so it doesn’t make sense to use park factors that include a load of hits that it doesn’t consider a part of a pitcher’s value
If fip park factors have no consistency with actual run suppression than using it in fWAR is statistically consistent but giving a false representation of the players value, which is the entire purpose of WAR.
It doesn’t necessarily misrepresent a pitcher’s value, but whether defense independent stats are the best way to value a pitcher is a different argument. There’s plenty of good stuff to read on that.
I think I get your point now. You are essentially arguing that fWAR is not a valid measurement of a pitcher’s results, or perhaps even arguing it’s not a valid measurement of a pitcher’s skill because it’s based on FIP, not ERA. Plenty of people agree with you, but it has very little to do with the current improvement in fWAR, since the improvement was one in consistency, and you want a change in philosophy.
We’re basically pretending Citi field isn’t a pitchers park for the sake of consistency. It’s a lie.
A Mets pitcher value is now flat out wrong in fWAR. If a mets pitcher should have a 3.50 era based on fip, then in reality he would have a worse than 3.50 in neutral stadium bc regardless how citi fairs in k bb and hr it suppresses runs. That’s the reality. What fWAR tells us now is not the reality. It’s false value.
Scoko, FIP based WAR does not care about how many runs you allow. It does not care about how many runs your ballpark allows. It only cares about K/9, BB/9, and HR/9. That is how it judges performance. FIP and xFIP are representations of a pitcher’s performance put into and era like form. If you want to include Citi Field ajustments into FIP based WAR you need to adjust for the ballpark’s K, BB, and HR rates.
You keep on harping on how Citi Field surpresses runs, however if that run suppresion does not show up in K/9, BB/9, or HR/9 then it must be due to BABIP affects. FIP based WAR does not care about BABIP.
“in reality he would have a worse than 3.50 in neutral stadium bc regardless how citi fairs in k bb and hr it suppresses runs.”
FIP based WAR is not factoring in Citi Field’s run suppression, it is not factoring in pitcher era.
Skoco, I don’t think you know what FIP is. You can read about it in the glossary here on Fangraphs: http://www.fangraphs.com/library/pitching/fip/
Scoko:
“To me fip park factors is another way of saying ‘using fip to determine how a park should fair in suppressing runs’.”
No one is saying that FIP calculates how a pitcher SHOULD perform. FIP simply calculates how well a pitcher performs in Ks, BBs, and HRs, and scales this number so that it looks like an ERA. It does correlate with ERA, and it is more of a repeatable skill than ERA, but we know that it’s not a perfect measure of true talent for everybody.
You are bringing in a much larger debate that has been going on all over this website for a long time, but has nothing to do with this simple improvement being made to fWAR. Right now, fWAR does not judge pitchers based on balls in play at all, and as long as that is the case it makes no sense to adjust their fWAR with an overall park factor that includes park effects on balls in play. The FIP park factor MUST be used as long as fWAR is calculated based on FIP.
The reason fWAR goes by FIP is because our current analytical tools simply don’t have the ability right now to accurately divide credit between pitchers and fielders for what happens to balls in play. Someday I think we will be able to do this with more granular data. For now, you can use FIP-WAR, or RA/9-WAR, or some mix of the two. All of these solutions are wrong, since FIP-WAR assumes pitchers have zero control over balls in play (which is wrong), and RA/9-WAR assumes pitchers have 100% control over balls in play (which is wrong). That said, to say that FIP-WAR is a “lie” is ridiculous. It’s simply one of the only answers we can calculate right now with our analytical tools.
It is an absolutely lie!!!!
Look at the Mets!!!!
Nothing within fWAR is taking into acct whatever factors make Citi a pitchers park. Because fip park factors are favorable for hitters in Citi, yet run park factors prove it is a pitchers park, all the fWARs of Mets pitchers are higher than what I think their actual value in terms of wins is to their team.
I dont see how you can argue that fWAR of Mets starters is now fairly representing the value of the players since there is no correction for the fact that it is a pitchers park.
I think a combination of FIP, and park factors based on runs would give a much more accurate version of value, or WAR. I dont see why they cant mix. Its taking the trustworthy information in terms of how well a pitcher or how a park will perform and combining them. In so many stats are unrelated but trustworthy statistics combined and weighted and transformed into a stat representing some form of value.
Run park factors accurately shows a parks ability to suppress runs.
Fip shows a close enough value of a pitchers ability.
Why cant there be a pitcher WAR that incorporates each? Didnt we just have that?
Scoko, here is why you are wrong: The Mets’ FIPs are not helped by Citi Field. The end.
BTW, UZR is park-adjusted. So, all of the Mets position players WARs are adjusted down via their UZR’s being adjusted down, since it’s easier to prevent singles, doubles, and triples in Citi Field. The effect you are looking for shows up in the evaluation of the Mets fielders.
” The Mets’ FIPs are not helped by Citi Field. The end.”
” The Mets’ FIPs are not helped by Citi Field. The end.”
” The Mets’ FIPs are not helped by Citi Field. The end.”
Why isn’t Derek Jeter the most searched for player on Fangraphs right now?
Noooooooooooo Kershaw passed #CoreyKluberSociety
I imagine Old Boss Radbourne would say his ERA hasn’t changed in 100 years.
Double play avoidance factored in? Rob Deer’s 1990 season just became epic…
No, not at all. There are many things that parks can do that affect strikeout and walk rates. For starters, picking up the ball against the backdrop behind the pitcher (the “batter’s eye”) is different everywhere. For another, the amount of foul territory makes a difference; places with smaller foul territory tend to allow batters to work deeper into counts, and increases both K and BB totals. Atmospheric conditions also matter; it’s harder to strike people out in Denver because breaking balls don’t break as much in the thin air.
This hasn’t always been well understood, but some teams have. The proclivity of the Dodgers for stockpiling power pitchers since the 1960s is no coincidence–Dodger Stadium has a poor batter’s eye, in addition to other factors that favor a pitcher. The Mets in the 1980s were the same way; picking up the ball in Shea was tough for a hitter, and tougher when the pitcher was a fireballer like Doc Gooden or a guy with a herky-jerky motion like Sid Fernandez.
I assume you are replying to me…
I doubt the factors that influence Ks are meaningful. Dimensions, atmospheric conditions, batters eye seem like factors that would influence runs much more than Ks. Because of those things I see WAR based park factors to be more meaningful. I dont like the idea of weighing K and BB so heavily when analyzing the effect of ballparks on pitchers bc I simply cant believe that any factors of 1 park drastically influence K or BB output over another. The depth of the outfield wall or the a jet stream to right center arent exactly factors that influence a K or a BB.
Remember that FIP includes dingers, so the aforementioned outfield depth and jet stream are included. But the K/BB totals are not insubstantial; in 2015, the K factor varied from 95 (Detroit, Pittsburgh, and Colorado) to 103 (Cincinnati, Houston, and Atlanta), and the BB factor varied from 96 (Pittsburgh) to 107 (Chicago AL). Now, as single season results, the observed spread has a decent chunk of variance, but some of that difference is constant across years. Are you suggesting a change of 3-5% in a pitcher’s strikeouts or walks wouldn’t be important?
Not important enough to completely overshadow the effects on pArk factors can have on run totals, IMO. Not denying that there are factors that influence K and BB, just think fip based park factors put too much stock into it.
Keeping competition the same, I can’t imagine if u took a staff and switched its park the staffs k rate wod be drastically better or worse from year to year.
I think fip based park factors ignores dimensions direct connection to runs.
K and BB park factors are easily calculable and usually stable from season to season. They might not be as easily observable but they still exist
Why would you want to even out the leagues when there are obvious differences due to the DH rule and talent differences observable through interleague records?
That should be a step backwards IMO.
Also, will there be a wGDP primer?
When were these changes implemented? I pulled the WAR projections for all players last week, to compare to throughout the season — I don’t want to identify a “change” that’s just due to a different WAR calculation.
These change were implemented because they improve WAR. Any improvements that could be made should be made, even if it means a lack of convenience.
That isn’t the furthest thing imaginable from an answer to my question, but it’s pretty close. I guess you thought I wrote “Why”, instead of “When”?
Yep, my bad.
Plus or minus FIVE wins for GDP? Should that say runs instead of wins? (Or 0.5 wins)
Can you go into #4 more? Does it help the AL or NO more?
Whoops, that’s supposed to be NL obviously.
As long as these changes make WAR a little bit more reliable, than change is always welcome! I’m particularly excited to learn more about wGDP.
Casey McGehee hates this move! -4.9wGDP last season has resulted in his WAR dropping from 2.0 to 1.4.
The Mets pitchers are pleased.
So are Padres and Angels pitchers. Rockies pitchers are mad.
It’d be interesting to see which players were the most impacted by the wGDP change. My first thought for current players was Pujols. I know he was in the 90’s for WAR already, and now he’s coming in at 88.5. Eclipsing the 100 WAR mark seemed like a certainty for him a few years ago, and now 11.5 more seems like it’s going to be very difficult to achieve.
This is very much appreciated. Some people may be angry about sudden changes, but these improvements legitimize FanGraphs (and their WAR) in many ways.
I’m excited to work to improve the WAR model even more in the future.
As a Mets fan, I’m whatever the opposite of angry is. Is it happy? No, it can’t be.
It’s joy. I’m pretty sure it’s joy.
I remember joy. It felt really, really nice. Appreciate the feeling while you’re can. It can be fleeting 🙁
Noah, would love for you to shed some light for me based on my comments above, especially the one about the Mets. I don’t really see what I am missing here, but clearly something if this change was made.
Noah wrote an article about it. Essentially, while Citi Field suppresses runs, it actually inflated FIP (mainly because it inflates HR). I don’t know *why* it does that, but it does. And since fWAR is based on FIP, it makes sense to use park factors based on FIP.
http://www.fangraphs.com/community/trying-to-improve-fwar-part-1/
How far retroactive are these changes?
wGDP appears to start in 2002.
Are the pitcher WAR adjustments limited as well?
Also, who are the players with the biggest changes in career WAR due to this update?
The Mets overall pitching staff just went from 30th in baseball to 18th, and the rotation just went from 25th to 10th.
deGrom – from 3.0 to 3.5
Colon – from 2.1 to 2.8
Wheeler – from 1.8 to 2.5
Niese – from 1.6 to 2.3
Gee – from -0.1 to 0.4
I believe the Royals staff was comparably affected.
Interesting changes, but next time it would be good to know ahead of time so we could export the tables and compare the difference.
How does pitch framing affect the FIP park factors? This seems like it would introduce a varible (catchers) in a stat that is desired to have long term stability (park factors).
Theoretically, I would think it doesn’t impact park factors the same way that having a great power-hitting team doesn’t impact park factors. The catcher should help his team on the road, as well as at home. And the opposing team shouldn’t get the benefit of the pitch framing. So it should be normalized for the talents of a team.
The Padres’ rotation ranking on the Depth Charts projections went from #15 up to #4 overall, while their relief corps projections went from bottom third up to #4 overall. That’s crazy.
I took a look at the archive…I believe Billy Butler just lost 30% of his career WAR.
I am surprised to see GDP coming into play. I thought all research generally suggested that GDP was not a skill, due to it generally having no year-to-year correlation with itself?
Lack of correlation doesn’t mean what you think it does.
This I am pretty certain of, as my skills in stats are poor at best. A quick google brought me here, which I’m pretty certain is the source of my thinking.
https://books.google.com/books?id=VsmnfVUKJskC&pg=PA119&lpg=PA119&dq=jonah+keri+gidp&source=bl&ots=t7-O0fwh-b&sig=LoFTcze_54SumekyHtXBvecRj5M&hl=en&sa=X&ei=plsQVb-ZCe_IsASr84G4Cg&ved=0CDYQ6AEwBQ#v=onepage&q=jonah%20keri%20gidp&f=false
Not that I expect you to take the time to educate me, but, if you’re feeling generous, what about Keri’s suggestion that avoiding GIDP is luck is wrong?
Thanks.
BABIP does not have very high correlation year-to-year but is included in WAR. The goal of WAR is not to evaluate true talent, its to value what a player did some of which is necessarily luck. A projected WAR, on the other hand, should regress a low correlation stat appropriately.
1 yr park factors? 3 yrs? 5 yrs? Hopefully not 1 yr.
Why different PF for pitchers and hitters? That makes no sense. Also makes no sense not to consider handedness for PF.
1 yr UZR still, unregressed as the author of UZR recommended (50% suggested)
No wonder WAR is off
I’m a bit confused on what the FIP park factors mean for WAR? Is this new update trying to tell us that different parks can influence strikeouts and walks? I know they can influence homers but do these FIP Park factors also involve strikeouts and walks? It doesn’t make much sense to me if it does.
“do these FIP Park factors also involve strikeouts and walks?” Yes, they do.
Why? The strikeouts and walks across all the parks can’t vary that much and if they did it would have to be just a one year thing, it’s not a thing that correlates well I assume?
As Noah said above, if you run the numbers, the effects are there and apparently easy to calculate and consistent. I am speculating, but here are some things that may cause them:
-different batter’s eyes in CF
-different light levels, especially shadows that keep on showing up between the mound and home
-psychological or strategic effect of walls, which can cause pitchers to be more or less aggressive towards hitters in certain parks
-amount of foul ground, which can effect how long at bats get extended by pop fouls
-different humidity or temperature, which may affect how pitchers can grip the ball or how much the ball breaks
-how delicious the club house food is, affecting the focus of pitchers coming in mid game.
etc.
Noticed as some pitchers had their career WAR increase, their LOB-Wins dropped about the same amount. What was the reason for this?