What's the difference between WAR on FanGraphs.com, and WAR on Baseballreference.com?

[fenrir]

Both seem to produce different results. What is the difference between the two?

[dominik]

Both seem to produce different results. What is the difference between the two?

MatthewC will give you a better answer but the biggest difference I think is that FG uses xFIP for pitchers and another defensive metric for position players. there might be a few more small difference but I don't really know.

[JDanger]

MatthewC will give you a better answer but the biggest difference I think is that FG uses xFIP for pitchers and another defensive metric for position players. there might be a few more small difference but I don't really know.

Fangraphs uses park-adjusted, league adjusted version of FIP, not xFIP. They also have a lower replacement-level.

http://www.fangraphs.com/library/index.php/war/differences-fwar-rwar/

[Bothrops Atrox]

FG uses a lower replacement level, no AL vs. NL league quality adjustment, UZR for defense, and FIP for pitchers. BRR uses a higher replacement level, a AL vs. NL league quality component, TZ/DRS for defense and RA adjusred for team defense for pitchers. There are other small differences with leveraging for relievers and some other smallish factors.

[fenrir]

FG uses a lower replacement level, no AL vs. NL league quality adjustment, UZR for defense, and FIP for pitchers. BRR uses a higher replacement level, a AL vs. NL league quality component, TZ/DRS for defense and RA adjusred for team defense for pitchers. There are other small differences with leveraging for relievers and some other smallish factors.

Which one do you consider more "accurate"?

[Bothrops Atrox]

Which one do you consider more "accurate"?

First of all, when looking at position players, we need to think about where the bulk of the differences come from. Before that, we need to remember that FG uses UZR data 2002 on and TZ 2001 and before. BR uses TZ before 2003 and DRS from 2003 on. So for seasons before 2002, the only big difference is replacement value and the Al vs. NL LQ adjustment BBR uses but FG does not.

For seasons 2003 on...

There is no real agreement as to which is better UZR or DRS. They are calculated slightly differently, key in on slightly different things, and use a different batted-ball data base. This is why you see an occasional big difference in defensive WAR between the two systems for a given player.

There is also not total consensus as to what replacement level should be. Form what I have seen, "experts" tend to side with FGs replacement level. If you chose to make a replacement level adjustment to make the two merge, add in about .5 WAR per 162 game AL season onto a BBR's total to equal FG and about .9 onto a 162 game NL season (for the lacking league quality comparison in FGs). Then what you are dealing with is just the defensive metric of your choice. I prefer the Al. vs. NL league adjustment and the slightly higher replacement level, so I side with BBR, but it is just a preference.

As far as pitchers, since FG uses FIP (and FIP correlates much better year-to-year with future performance than RA), FG is better for a pitcher who hasn't played long enough to know the real impact of their BABIP or LOB skills. Once we have enough sample size to determine if a player really has BABIP and LOB impact (say 7-8 full seasons), I prefer to look at BBR.