PDA

View Full Version : Fangraphs WAR vs. Baseball Reference's WAR



xnick5757
09-27-2011, 10:09 AM
which do you prefer? And can someone explain to me why they are different?

quiksilver2491
09-27-2011, 11:01 AM
Well there are slight differences in linear weights, Baseball Reference uses Sean Rally's WAR (baseballprojection) and Fangraphs uses their own method.

The main difference comes in how defense is rated, BR uses range factor while Fangraphs uses UZR, which is the better defensive metric of the two.

Also, Fangraphs ratings are a bit more peripheral based then BRs so they project to more of how a player should of performed rather then how he actually did. I definitely prefer Fangraphs, mainly for the UZR, but I know some like BR.

Milnertime
09-29-2011, 10:18 AM
They honestly just measure different things. IIRC, BR's linear weights are set to be weighted against the single season a player is competing in and FG's linear weights uses a more "for all time" kind of method.

Both tell meaningful things depending on what you're looking for. It's good to use both in tandem and not rely on just the "WAR" portion and look at the component parts of each and compare. If UZR really likes a player but BR's defensive metric doesn't, there's probably a reason.

Just as an observation, BR's batting run values are almost always less than those at FG.

bagwell368
11-19-2011, 10:32 PM
Well there are slight differences in linear weights, Baseball Reference uses Sean Rally's WAR (baseballprojection) and Fangraphs uses their own method.

The main difference comes in how defense is rated, BR uses range factor while Fangraphs uses UZR, which is the better defensive metric of the two.

Also, Fangraphs ratings are a bit more peripheral based then BRs so they project to more of how a player should of performed rather then how he actually did. I definitely prefer Fangraphs, mainly for the UZR, but I know some like BR.

I find when taking my memories of players - Kaline, Clemente, Evans, Fregosi, Nomar, etc. that BR's defense metrics seem more accurate and less apt to jump all over the place - call it more damped. UZR seems to jump around quite a bit, and at times in ways that does not seem correct year over year.

Since UZR is only 10 years old, and requires a solid 3 year period before one can make judgments with it, RF seems distinctly more useful going backwards into time, and less twitchy to use.

On of the pitching side I find FIP to be useful in projecting future seasons, but the notion that Sabathia was slightly better then Verlander in 2011 is simply bizarre.

Sign me BR forever.

BTW who is Sean Rally? Isn't the Sean Smith the guy behind rWAR and TZR?

Kirel
11-20-2011, 02:04 PM
I find when taking my memories of players - Kaline, Clemente, Evans, Fregosi, Nomar, etc. that BR's defense metrics seem more accurate and less apt to jump all over the place - call it more damped. UZR seems to jump around quite a bit, and at times in ways that does not seem correct year over year.

Since UZR is only 10 years old, and requires a solid 3 year period before one can make judgments with it, RF seems distinctly more useful going backwards into time, and less twitchy to use.

On of the pitching side I find FIP to be useful in projecting future seasons, but the notion that Sabathia was slightly better then Verlander in 2011 is simply bizarre.

Sign me BR forever.

BTW who is Sean Rally? Isn't the Sean Smith the guy behind rWAR and TZR?
The question though is if range factor is even legitimate, or if it's just an arbitrary measure that "feels" better.

And being bizarre(which, unless you look at ERA alone, it really isn't) does not make it untrue. The two are very, very close to being identical in all peripherals except BABIP, and neither, on average, looks to have an edge on reducing BABIP vs the other. Verlander just had a very good "lucky" year while Sabathia had a rather "unlucky" one. Normalize luck and defense and you probably cannot reasonably claim Verlander definitively superior. Sure, Verlander threw more innings, but he faced fewer batters than Sabathia, so the workload argument isn't there. The difference there likely lies entirely in BABIP and the resultant difference in hit rate.

There is virtual zero statistical basis that Verlander was better than Sabathia. The results certainly lean in Verlanders favor, but the performance...

bagwell368
11-20-2011, 05:07 PM
The question though is if range factor is even legitimate, or if it's just an arbitrary measure that "feels" better.

And being bizarre(which, unless you look at ERA alone, it really isn't) does not make it untrue. The two are very, very close to being identical in all peripherals except BABIP, and neither, on average, looks to have an edge on reducing BABIP vs the other. Verlander just had a very good "lucky" year while Sabathia had a rather "unlucky" one. Normalize luck and defense and you probably cannot reasonably claim Verlander definitively superior. Sure, Verlander threw more innings, but he faced fewer batters than Sabathia, so the workload argument isn't there. The difference there likely lies entirely in BABIP and the resultant difference in hit rate.

There is virtual zero statistical basis that Verlander was better than Sabathia. The results certainly lean in Verlanders favor, but the performance...

I know Verlander had some good D behind him, but let's do the rounds...

Prospectus says:

JV FRA 2.95 pVORP 58.3 WARP 6.7
CC FRA 3.53 PVORP 44.1 WARP 4.4

Internet voting says (Verlander with ease):

http://www.baseballprospectus.com/article.php?articleid=15554&mode=print&nocache=1321670338

FanGraphs says:

FanGraphs tERA and SIERA were lower then CC's

FG also didn't raise a fuss when Verlander won, which they usually do when they don't like a vote:

http://www.fangraphs.com/blogs/index.php/its-unanimous-justin-verlander-wins-2011-al-cy-young/

Oh yeah, my quick and dirty look at the performance vs divisions, says that JV did better vs the "tough" AL East then CC did vs the AL Central. I've heard that critique as well as the defense and the "lucky" BAbip. CC and Weaver are worthy 2nd and 3rd to Verlander, but I am afraid the voters got it right and FG (which part of it?) and ESPN Insider got it wrong on CC.

Virtually zero evidence? So sorry, even Fan Graphs is split on Verlander winning, guess they must be beaten with a wet noodle to put them in line...:eyebrow:

Kirel
11-20-2011, 08:42 PM
I know Verlander had some good D behind him, but let's do the rounds...

Prospectus says:

JV FRA 2.95 pVORP 58.3 WARP 6.7
CC FRA 3.53 PVORP 44.1 WARP 4.4

Internet voting says (Verlander with ease):

http://www.baseballprospectus.com/article.php?articleid=15554&mode=print&nocache=1321670338

FanGraphs says:

FanGraphs tERA and SIERA were lower then CC's

FG also didn't raise a fuss when Verlander won, which they usually do when they don't like a vote:

http://www.fangraphs.com/blogs/index.php/its-unanimous-justin-verlander-wins-2011-al-cy-young/

Oh yeah, my quick and dirty look at the performance vs divisions, says that JV did better vs the "tough" AL East then CC did vs the AL Central. I've heard that critique as well as the defense and the "lucky" BAbip. CC and Weaver are worthy 2nd and 3rd to Verlander, but I am afraid the voters got it right and FG (which part of it?) and ESPN Insider got it wrong on CC.

Virtually zero evidence? So sorry, even Fan Graphs is split on Verlander winning, guess they must be beaten with a wet noodle to put them in line...:eyebrow:
let me rephrase, "There is virtually zero statstical basis for Verlander over Sabathia if you use metrics that abide by BABIP being luck."

FRA just doesn't strike me as legitimatly avoiding BABIP, but I'm not intimatly familiar with it.

And voting and fangraph not raising a fuss isn't all that interesting. My point was simply that, in terms of FIP and DIPS theory in general, there wasn't a lot of seperation. It's hardly a tragedy that Verlander won(had I a vote, he'd have gotten mine)...he did have far better *RESULTS*, but that's not what FIP aims to measure.

bagwell368
11-20-2011, 10:03 PM
let me rephrase, "There is virtually zero statstical basis for Verlander over Sabathia if you use metrics that abide by BABIP being luck."

FRA just doesn't strike me as legitimatly avoiding BABIP, but I'm not intimatly familiar with it.

And voting and fangraph not raising a fuss isn't all that interesting. My point was simply that, in terms of FIP and DIPS theory in general, there wasn't a lot of seperation. It's hardly a tragedy that Verlander won(had I a vote, he'd have gotten mine)...he did have far better *RESULTS*, but that's not what FIP aims to measure.

Overall a fair response, I just don't care for FIP undercutting results, since that's what the award is to reward.

Kirel
11-21-2011, 09:36 PM
Overall a fair response, I just don't care for FIP undercutting results, since that's what the award is to reward.
I don't think Fangraphs FIP is really a great tool for awards purposes. It's more useful, to me, to say "Well, if you take Sabathia or Verlander in 2011 and put them in, say, Pittsburgh, I'm not sure who'd have the better year", which isn't something you can say about, say, Roy Halladay and Jonathon Sanchez. Far more useful for making decisions for the future than speaking definitively about today.

bagwell368
11-21-2011, 11:37 PM
I don't think Fangraphs FIP is really a great tool for awards purposes. It's more useful, to me, to say "Well, if you take Sabathia or Verlander in 2011 and put them in, say, Pittsburgh, I'm not sure who'd have the better year", which isn't something you can say about, say, Roy Halladay and Jonathon Sanchez. Far more useful for making decisions for the future than speaking definitively about today.

Yes, no doubt.

tiger337
12-26-2011, 12:45 AM
The main reason why BR WAR is usually lower than FG WAR for batters is because it has a slightly higher replacement level.

Another difference between the two WARs is that BR WAR is calculated so that the total offensive runs credited to the players on a team equals runs scored for the team. That is not the case for FG WAR. I don't have a a preference between the two, but rather like to look at both of them together.

Others have explained that the pitching WARs are totally different. Again, I like to look at both but favor the BR WAR for awards and Hall of Fame consideration. For prediction, FG WAR becomes more useful.

SmithWesten
12-26-2011, 10:40 PM
You have to use a combination of both

I calculate my own personally.

bagwell368
12-26-2011, 11:07 PM
You have to use a combination of both

I calculate my own personally.

You mean you add them up and divide by two - or do you have some more detailed method you could share?

Jeffy25
12-28-2011, 07:18 PM
You mean you add them up and divide by two - or do you have some more detailed method you could share?

SW is my brother-in-law, he doesn't really do any calculations (kick me, I don't care) I mainly do them.

I have spreadsheets where I calculate them all on my own. No, not a combination, I actually calculate them on my own using my own weights for fielding and I even adjust some statistics because I view flaws in their calculations. (wOBA for example should include IBB and not an improved value for reaching on an error that is near the value of a single, as well they have their metrics incorrect for the value of a single to produce a run/double to produce a run etc, I have updated that in a variety of ways that still uses the same constant for league average and consistency).

Defensively, it comes down to how much value and weight you think defensive metrics should be given, and 'r' and 'f' WAR view this very differently, and I don't really agree with either.

Personally, for a snap shot view, rWAR is good for pitching, fWAR is good for position players (as long as there aren't extreme defensive numbers) and a combination of the two with WPA will give you a good idea how good the player probably really is.

I do calculate them myself in a spreadsheet at the end of each year. The biggest difference I have for any one player is actually Brandon Morrow. But I am not ready to release what I have done yet because I'm not convinced in my findings yet that it's actually 100% strong enough to defend yet.

At this point, all I have is anecdotal evidence that I have done it well enough, hopefully in a few months I have something I am proud to release.

The intention of WAR is to find the number of wins that player creates for your team offensively and defensively (and you can include base-running technically). I think there are too many flaws in WAR as it sits, and I am determined to find a better version of WAR that is not only easier to explain, but also makes a lot more sense.

I'm also calculating it for managers based on their decisions and how they affect WPA, but that is super complicated currently, mainly because I don't have enough data.

OneTuzSea
12-29-2011, 03:05 PM
Looking forward to jWAR :D

thefeckcampaign
04-10-2012, 06:58 AM
You guys lost me. I got on this to find out something basic.

If you add all the players WAR on a particular team, did that total equal the amount of wins that team received that year?

Jeffy25
04-10-2012, 07:49 PM
You guys lost me. I got on this to find out something basic.

If you add all the players WAR on a particular team, did that total equal the amount of wins that team received that year?

That isn't quite how it works.

I'll give you an example.

Let's say your team scores 4 runs tonight.

But they can have a runs created total of 4.96.

That's basically saying the team earned another run, or should have scored another run, but they didn't execute that run (maybe too many guys were left on base, maybe some guys thrown out on the base paths) etc.

Over the course of the season, those Runs Created will come close to the teams actual runs scored, but it won't be accurate.


WAR works as a reference total, it isn't going to be exact. Sometimes, the best team WAR wise isn't the team with the most win. If you want to have a good chance of projecting wins, or looking at a stat and being able to get a good guess what their wins total will be look, take a look at their runs scored vs runs allowed. It won't be accurate (look at the 2011 Giants) but it's usually fairly accurate. Sometimes, a team wins a few close games, or loses a few close games that could have easily gone the other way. These numbers are just to give an approximation or to get you close for an evaluation.