WAR

[savafan]

I'll be the first to admit, I don't get it, so I decided to do some research to better understand. Now I'm finding out that it's not exactly consistent, so I'm curious, when it comes to discussion on this board, which of these calculation methods are we using?

http://www.baseball-reference.com/blog/archives/6063

or

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

[Cooper]

You make a good point. Brandon Phillips has a huge gap between the way both sites figure it. One site has him in the 3's and fangraphs has him at a 6. I'm not sure why but for some reason i went with the fangraphs number because i thought the way they came up with the defensive numbers was better.

It's the defense numbers which have a large variance.

[dougdirt]

You make a good point. Brandon Phillips has a huge gap between the way both sites figure it. One site has him in the 3's and fangraphs has him at a 6. I'm not sure why but for some reason i went with the fangraphs number because i thought the way they came up with the defensive numbers was better.

It's the defense numbers which have a large variance.

Fangraphs and BR use two very different types of WAR.

Fangraphs uses UZR data for their defensive values, which I prefer. Baseball Reference uses Total Zone Rating, which I think have a bunch of flaws in it.

The same things work for pitchers, both sites use very different criteria for pitcher WAR. Fangraphs uses more of a hypothetical value based on FIP rather than what Baseball Reference uses, which is based on ERA. I can get behind both ideas. One side of me says "what happened" is what should be counted for value, but not something that should be looked upon as a prediction for future value. But the other side of me says that a pitcher can't really control the defense behind him, and his value shouldn't be tied to the defense behind him.

When I talk about WAR though, I use Fangraphs because I am 500% more comfortable with their defensive values in WAR than what BR uses. With pitching, I am comfortable enough with how they go about creating that WAR.

[RedsManRick]

They also differ on baserunning, I believe. B-R uses just SB, CS and the like, while Fangraphs uses a more comprehensive measure that looks at all base running opportunities and base advancement.

And perhaps most importantly, I think they use different replacement baselines.

It's best not to think of WAR as a "stat" but rather as a framework for understanding performance and its value. The implementations generally work the same way, that is, they have the same variables. But they may use different inputs.

[blumj]

Specifying fWAR and rWAR helps avoid some of the confusion with just the 1 little extra letter.

[MikeThierry]

Baseball reference is a great site to go to for counting stats. I generally go there to look at batting averages, home runs, etc. However, I feel that Fangraphs is far superior when determining saber numbers. It is a website created by sabermatricians so they are going to have the most up to day information pertaining to the calculating of said numbers.



It's best not to think of WAR as a "stat" but rather as a framework for understanding performance and its value. The implementations generally work the same way, that is, they have the same variables. But they may use different inputs.

I think that is my main problem with some sabermatricians out there. They use tools like WAR as the holy grail and use it as a stat rather than a tool to understand performance.

[RedsManRick]

Baseball reference is a great site to go to for counting stats. I generally go there to look at batting averages, home runs, etc. However, I feel that Fangraphs is far superior when determining saber numbers. It is a website created by sabermatricians so they are going to have the most up to day information pertaining to the calculating of said numbers.

I think that is my main problem with some sabermatricians out there. They use tools like WAR as the holy grail and use it as a stat rather than a tool to understand performance.

I'm with you on Fangraphs as my primary source for sabermetric info. As for your critique, I think this might have been true to an extent when WAR first came on the scene, but I'm not sure I know of any respectable analyst who uses it as such. Can you name one or two who you feel over-do it? I sometimes see non-saber types make this kind of assertion because they themselves misunderstand the way the data is being positioned -- or the context in which the article is being written. And sometimes the saber types fail to explicitly state the limitations of their analysis because in their minds, those limitations are obvious and should be assumed by the reader. And so a person without that background assumes the writer is coming form a place of arrogance because he doesn't know how the analyst thinks.

All that said, at the end of the day, if you're putting together an analysis -- or more to the point, trying to make a decision -- you have to stop somewhere and simply come to an answer. At some point, you have to say, this is the best collection of information/insight I can get and now I need to make a decision. We'll never have a perfect answer and insomuch as we can always learn more, we should be open minded. But at some point you draw the line in the sand.

We should also be careful not to draw false equivalencies. Some people are want to say things like "WAR is just one more stat that should be considered along side things like batting average, ERA or RBI." That's simply not a fundamental misunderstanding of good analysis. You have a specific question, you come up with a model for integrating the relevant data, you do your math, adjust at the end if you need to account for something not in the model and there you go. Is it THE perfect answer? No, but if the question is overall contribution to a team's W-L record, WAR gets you in the right ballpark -- and trying to add in other information in the form of "old school" stats doesn't really add insight. I'm already tired of seeing an "analyst" on ESPN say things like "Well, player A's WAR was 7 and that player B's was 3, but player B had 15 more RBI, so it's hard to say who had the better year." There are objectively better and worse ways to look at the data. You can choose whether or not you want to make your analysis based on numbers or not, but if is numbers based, some implementation of WAR is pretty much the gold standard right now.

[jojo]

Basically if the question you're seeking to answer with WAR spans 2002 forward, go with fangraphs. If the question you're seeking to answer via WAR spans eras prior to 2002, then you have no choice but to use BR's WAR. Do not mix the two and if given the choice of either, go with fangraphs (or consider both but weight fangraphs greater).

[RedsManRick]

There's a great thread on TangoTiger's website about WAR and its misconceptions:

http://www.insidethebook.com/ee/index.php/site/comments/war_doesnt_work/

Here's the crux of his point (comment #57)


The takeaway should always be this:

Is the way the data is being processed by WAR better or worse than what you would do yourself?

If you choose to discard WAR, then you have to create your own “smushing” system. Is that personal smushing system more reliable, less uncertain?

***

Now, perhaps SOMETHING in WAR you can do better. So, instead of discarding WAR, you keep most of it, and replace, say, UZR, with, say Fans Scouting Report (i.e., the eye test). And, you say, that is more reliable and less uncertain.

And that’s fine, and acceptable. And that’s why Fangraphs and Rally present the WAR data by its components: to make it easy for you, as a person, to make your smushing system more reliable and less uncertain than what Fangraphs and Rally show.

The framework of WAR is rock solid. The components used in the various implementations can be discussed, but you have to bring an alternative.

And elsewhere, he expounds further, making the point that you can use whatever "smushing" system you want. Want to factor in clubhouse leadership? Go for it. Want to give extra credit for speed putting pressure on defenses? Sure thing. But do it systematically. Don't just use those things to poke holes (WAR doesn't account for ______, there it's no good) or to give extra credit when you feel like it (Jeter is underrated by WAR because ______). The system is useful precisely because it considers so much information in a way that allows to look at all players in a comparable context. You can add variables to the model if you want. You can use different data sources. But at the end of the day, if you want to compare players, you need to do it in a rigorous, systematic way. And doing so such that you others can test and verify your work. And really, that's what Sabermetrics is about.

[savafan]

So what is the definition of "replacement" in WAR, is it a league average player?

[jojo]

So what is the definition of "replacement" in WAR, is it a league average player?

The baseline in WAR is a hypothetical league average player independent of position.

Replacement level for defense is roughly average (which means zero runs over 1200 defensive innings). Replacement level or offense is roughly -20 runs below league average per 600 PAs.

So with all of things considered, a replacement lever player is one that posts a WAR of 0 over a playing time of roughly 600 PAs.

An average player posts a WAR of about 2 give or take.

[RedsManRick]

An average player posts a WAR of about 2 give or take.

http://www.bluebirdbanter.com/2011/10/4/2468426/hate-and-war-the-only-things-we-got-today-does-replacement-level


These graphs might look similar at first glance, but note the differences in scale on the x-axis (!). As expected, the slopes are similar (one more WAR should mean one extra team win), but where that relationship begins is very different depending on whether we're looking at fWAR or rWAR. So a main reason we see a big difference between fWAR and rWAR is that rWAR assumes that replacement-level is higher than fWAR assumes it is. How much higher?

Well, the y-intercept of these graphs should tell us how much a replacement-level team would win (since team WAR would be equal to zero).

The linear model describing the relationship between rWAR and wins is:

Wins = 53.8 + 0.88(rWAR)

So a replacement-level team would win about 54 games (only very slightly more than the 52 games that baseball-reference purports replacement-level teams should win, which could be due simply to variance, since we only used 90 separate team-seasons). We see the model doesn't quite get us to a 1:1 ratio between WAR and wins, but it is pretty close.

The linear model describing the relationship between fWAR and wins is:

Wins = 45.2 + 0.93(fWAR).


http://assets.sbnation.com/assets/735824/fwarvswins.jpg
http://assets.sbnation.com/assets/735830/rwarvswins.jpg



In summation, I'd say that neither version of WAR is necessarily more useful (or "better") than the other, however, it is important to keep in mind that the methods do scale differently. Consider this for a moment: we talk about a 7 WAR player being a likely MVP candidate. Well, add a 7-WAR player to a team full of fangraphs replacement players and they're likely to win no more games than a team full of baseball-reference replacement players without that 7 WAR player. So what do you all think about that?

[MikeThierry]

Can you name one or two who you feel over-do it? I sometimes see non-saber types make this kind of assertion because they themselves misunderstand the way the data is being positioned -- or the context in which the article is being written.

I think the posters within RedsZone do an excellent job at analysis. While I'm still not sold on FIP (for example), I have been convinced of its uses here in the forum by some of the posters, at least use on a limited scale. I have just had a lot of dealings with other people outside of this forum where they make themselves look like total you know whats with their use of WAR or other saber stats. On a national level, I absolutely cannot stand Keith Law. He is one of the most arrogant individuals I have ever read or heard on the radio. There was an interview here locally where he came on and the interviewer challenged Keith Law on his evaluation. Now, the interviewer was not very confrontational and was just asking questions about the logic used in determining his analysis. Keith Law got really defensive about it and said something to the effect of "well you don't have the kind of degree I have to where you won't understand the kind of analysis I do". He later called back into the station and said he would never do an interview on that station again. He is one example of a national guy where I feel falls into the trap of being "the smartest person in the room" and if you challenge him, he becomes defensive.

[savafan]

I would like to go into the 2012 season with a bit better understanding of these newer stats. Can anyone recommend some good books that will break things down to where someone with a lot of game knowledge but little stat knowledge can understand them?

[jojo]

I would like to go into the 2012 season with a bit better understanding of these newer stats. Can anyone recommend some good books that will break things down to where someone with a lot of game knowledge but little stat knowledge can understand them?

There are actually multiple primers available on the internet. I'd recommend gobbling them up before spending money (IMHO, there is not a perfect book offering an introduction to sabermetrics though there are now several out aiming to be just that).

[MikeThierry]

There are actually multiple primers available on the internet. I'd recommend gobbling them up before spending money (IMHO, there is not a perfect book offering an introduction to sabermetrics though there are now several out aiming to be just that).

Yeah, I agree with jojo on this. There are tons of articles online explaining some of the stats. Fangraphs actually does a good job in breaking down some of the advanced numbers. You just have to search on their page for them.

[MikeThierry]

I found this on Fangraphs. Excellent breakdown of WAR and other numbers:

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