Differing WAR numbers

[monkeyman343434]

I have a general question for anyone out there who is big into WAR.

 

I have checked the WAR numbers on a few guys on both their Baseball-Reference page as well as their Fangraphs page and I'm getting different numbers.

 

Example:

 

Justin Upton: Fangraphs : 6.4 WAR in 2011. Baseball-reference: 4.1 WAR in 2011.

 

Just wondering why the numbers would be so different and which one is more accurate?

[umphrey]

http://www.fangraphs.com/library/index.php/misc/war/

 

fWAR = FanGraphs

rWAR = Baseball-Reference

 

1. Pitcher Value: fWAR relies totally on FIP, making it a defense-independent metric, while rWAR adds in defensive runs saved as well.

2. Calculating Defense: Each system uses a different defensive metric. fWAR uses UZR, while rWAR uses TotalZone

(developed by Sean Smith). UZR is more accurate but is only available

from 2002 onward, while TotalZone values can be calculated for any

player in baseball history.

3. Baserunning: rWAR includes this, while fWAR doesn’t.

[Ennder]

- WAR is context, league, and park neutral. This means you can use WAR to compare players between years, leagues, and team

 

I know this is correct in concept but I certainly wouldn't trust this to be true in any way.

[monkeyman343434]

Not that this really matters for anything, just found it really interesting on how some WAR (especially Upton, Stanton, and McCann) were close and some were much further apart.

 

Player BR FG

WAR WAR

Brian McCann 2.5 3.7

Mitch Moreland -0.2 0.4

Rickie Weeks 3 3.7

Alcides Escobar 2 2.2

Alexi Casilla 1.8 1.4

Logan Morrison 0.2 1

Andrew McCutchen 5.5 5.7

Shane Victorino 5.1 5.9

Mike Stanton 5.7 4.5

Justin Upton 4.1 6.4

[Baldkin]

Both Fangraphs and Baseball-Reference break down their WAR calculations for you to see (offense, defense, baserunning, positional adjustments). Upton's difference was almost all due to the difference in how the two defensive systems viewed him.

 

Ennder, what makes you skeptical that it is context, league and park neutral?

[thebruce44]

Its awful that BF was allowed to call their calc WAR. I get that's its somewhat arbitrary, but why call it the same thing if you go second. You are going to cause nothing but confusion.

[sveumrules]

The AL & NL have different run environments due to the DH so adjusting at least somewhat for league would seem to make some sense. At the same time my favorite thing about these numbers is they really are just guidelines. It is a fact that Ryan Braun batted .332 last year, it is an opinion that he was worth 7.8 wins above a replacement player.

[Baldkin]

The AL & NL have different run environments due to the DH so adjusting at least somewhat for league would seem to make some sense. At the same time my favorite thing about these numbers is they really are just guidelines. It is a fact that Ryan Braun batted .332 last year, it is an opinion that he was worth 7.8 wins above a replacement player.

Its not an opinion. Its a fact. There are mathamatical calculations used to come up with that value. Saying it is an opinion just makes it seem like its some guy randomly deciding that Tim Tebow was a better LF than Ryan Braun.

 

WAR is made up of Batting, Defense, Base Running, Positional Adjustment and Replacement Level. If you don't like the values or the methods used to come up with a particular part, you can use your own numbers to come up with a new statistic.

[Nottso]

I just can't buy it as fact. If it were indeed fact the Brewers 2011 total team WAR would be exactly 6 greater than that of the Cards, 17 greater than that of the Reds, 24 greater than that of the Pirates, etc. I'd be willing to bet that those numbers just won't add up. I'm far too lazy to even attempt it, especially when considering the required extrapolation of the numbers for part time players, but it might make a nice project for one of our numbers guys.

[igor67]

The proper way to look at the issue is that the number is indeed a fact (everybody who got their calculators out would come to the same number). The NAME of the statistic in this case however makes a very strong claim as to how to interpret the number. A claim that is reasonable, but one that still needs to be interpreted with some amount of caution.

[Nottso]

Excellent explanation, Igor. That makes more sense for me.

[RightFieldCoder]

WAR is good to use for comparisons of different players. It is not literal. Playing Braun last year did not net us 7.8 wins more than an average MiLB LFer. I don't know if it's possible to prove it one way or the other. But we can say that he was about 50 runs (5 "wins") better than Alcides Escobar. WAR makes it easier to compare players at different positions in different leagues. It takes offense, defense, and baserunning into account. It is like OPS, not something you should cling to as the be all end all of ability but something quick to look up to make a rough judgement. Also remember that WAR is a counting stat. Someone who plays 150 games should have twice the WAR as the same ability player who only played 75 games.

[sveumrules]

The defensive component alone has so much room for error that I can't consider the end result a mathematical fact in the truest sense of the word. If all of your inputs have varying levels of reliability your end result cannot be considered absolute.

[umphrey]

The defensive component alone has so much room for error that I can't consider the end result a mathematical fact in the truest sense of the word. If all of your inputs have varying levels of reliability your end result cannot be considered absolute.

UZR is the worst. I really, really like the wOBA side of it, really good for comparing offensive production. Just think about this: last year Casey McGehee would have been a negative WAR player but his fielding was judged so good by UZR that he was supposedly a positive one.

[1992casey]

Its awful that BF was allowed to call their calc WAR. I get that's its somewhat arbitrary, but why call it the same thing if you go second. You are going to cause nothing but confusion.

 

You mean BR? http://forum.brewerfan.net/images/smilies/smile.gif

 

I agree. Baseball Prospectus did the same thing with isolated power, calculating it differently than originally designed. There's value in both ways of calculating, but geez, give the stat a new name.

[thebruce44]

Its awful that BF was allowed to call their calc WAR. I get that's its somewhat arbitrary, but why call it the same thing if you go second. You are going to cause nothing but confusion.

 

You mean BR? http://forum.brewerfan.net/images/smilies/smile.gif

 

I agree. Baseball Prospectus did the same thing with isolated power, calculating it differently than originally designed. There's value in both ways of calculating, but geez, give the stat a new name.

Haha, yes...thanks. I was in such a rage I couldn't even type properly.

I agree, there is value in both stats. But as you said, its been leading to nothing but confusion even on a site like this one where you have pretty knowledgeable fans.

[Thurston Fluff]

UZR is the worst. I really, really like the wOBA side of it, really good for comparing offensive production. Just think about this: last year Casey McGehee would have been a negative WAR player but his fielding was judged so good by UZR that he was supposedly a positive one.

 

That is because they use UZR in a way that is not accurate. It is commonly accepted that defensive metrics need 3ish years to be as accurate as one year of offensive metrics. Then they use one year's worth of defensive metrics and count it like it is as accurate as one year's worth of offensive metrics.

[TooLiveBrew]

So would something with a weighted average of the past three seasons' UZR or UZR/150 make more sense? Maybe valuing players in that way in general makes sense, not just defensively.

[umphrey]

So would something with a weighted average of the past three seasons' UZR or UZR/150 make more sense? Maybe valuing players in that way in general makes sense, not just defensively.

You definitely can't use UZR/150 because that's an extrapolation stat and WAR is a counting stat. For example Craig Counsell had an off the charts UZR/150 last year of 40 because UZR liked him in a small sample and extrapolated it like he was an every day starter. I don't buy into the 'past 3 seasons' logic of UZR either for a few reasons. Things change, in 3 years a player can be a different quality fielder in that time. Mostly though, if a stat isn't accurate over 1 season and over 1000 innings, I don't trust the 3 year average much more.

[sveumrules]

Defensive stats and evaluation have made enormous leaps and bounds in the last ten or so years but they are still nowhere near as reliable and concrete as offensive numbers. If WAR has two players close and one gets all his value from offense and the other gets all his from position and defense I'll take the guy who creates most of his value with his bat because you know EXACTLY how much that is worth. The defensive inputs are nice and I'm sure people work really hard to make them as accurate as possible but they are nowhere near as empirical as offenisve metrics and their creators will generally admit as much.

[brettac1]

Agreed. I'm really not very trusting of any defensive stat at this point. Unless they all agree, I find it hard to use them to make a judgement on a player.

[Baldkin]

---

Things change, in 3 years a player can be a different quality fielder in that time. Mostly though, if a stat isn't accurate over 1 season and over 1000 innings, I don't trust the 3 year average much more.

OF's make about 2.2 plays per game. Go look at what hitters do over a 350 AB sample, and how different they can end up from their final 650 AB sample. That's why defensive stats aren't reliable, just not enough opportunities.

[Thurston Fluff]

So would something with a weighted average of the past three seasons' UZR or UZR/150 make more sense? Maybe valuing players in that way in general makes sense, not just defensively.

 

While that may make it better I don't think it solves all of WAR's issues. WAR is a valiant attempt to get an all encompassing simple number to evaluate a player's ability. I simply don't believe you can boil so many factors down into one accurate simple number. I'm fine with using WAR as a beginning point if they would eliminate, or least change, how they measure defense in WAR but I still think it is not the definitive end all, tell all stat that it sometimes gets used as. Setting aside the defensive issues perhaps my objection with WAR is less the stat itself than how it gets used. As it stands I don't even think it helps much as a starting point. You just can't use an inaccurate metric and expect an accurate result. As the saying goes garbage in garbage out.

[Ennder]

WAR is about giving results not showing how good a player is so once you start using weighted averages over multiple seasons you are creating a completely new stat. The problem with WAR is that baseball is way too random for a single season of any stat to really mean very much and no stat is close to perfect at defining just how valuable someone is, especially a stat that you try to adjust by position or park since those adjustments are flawed to begin with, so you are always going to get results that look weird.

[Thurston Fluff]

WAR is about giving results not showing how good a player is

 

How many times on this site have we seen it used to show how good player is? It seems like the most used stat to show someone is better than someone else.