The defensive component of WAR is one of the major place where BaseballReference (BR) and Fangraphs diverge in their WAR calculations. As of their 2012 update BR is using John Dewan’s Defensiver Runs Saved (DRS) and Fangraphs is using Ultimate Zone Rating (UZR) created by Mitchel Lichtman. Prior to 2012 for BR and prior to 2002 for Fangraphs, DWAR is calculated using Total Zone Runs (TZ) created by Sean Smith. TZ is calculated from play by play data and is therefore able to calculate values for players as far back as Retrosheet (back to 1914) has available, thus allowing comparison to generations of players prior to the other two metrics. Both UZR and DRS rely on Baseball Info Solutions (BIS) data, which charts various information such as hit type, location of play, etc. play by play using people watching tape of games. That is why it can’t be used going back farther, or at least the labor necessary to go back very far would be considerable.
UZR & DRS
Both UZR and DRS are attempting to do basically the same thing. The main idea is to assign a probability to each defensively play, so a lazy pop fly to second base is caught 99.9% of the time while a line drive into the left-center gap might only be caught 10% of the time. This is turned into a plus minus, so when a player makes a play they get one minus the probability of positive credit, so for the line drive above 1-10% =.9 toward # of plays made above average. If they don’t make the play they get a negative of the probability so the line drive costs them -.1 for missing it. These values are then multiplied by a run value based on how a hit in that area behaves, so on average the line drive would be a double and have a higher run value than a likely single on a pop-up dropping near the second baseman’s area. Multiply the run value for that hit location by the +/- of the play and you get the run value added or lost for that particular play.
If you want more detail on each click through this link to UZR or this link to DRS. Despite them trying to do the same thing with the same data, the answers are not always very similar. For instance, this just completed 2014 season has players like Lorenco Cain as 16.8 runs above average by Fangraphs and 24 above average by BR which is nearly a 50% difference in value and adds almost 1 full WAR to his BR number. At current player value assumptions near $6 million/WAR ,that is a huge difference. That means this is much harder to do than you think, but there is a chance that this will be defunct in the near future anyway. If the new tracking system data is collected in every stadium and is available it will likely alter defensive statistics significantly.
The detailed links give more information as to how they handle values such as throwing, double plays, park affects, et.
This method looks at each play and has two methods depending on the information available. If the fielder responsible for the play is known, for good or ill, they are charged with the outcome. Sometimes this is split if it goes between two players. If this is not available then every fielder gets a portion of the blame or credit based on frequency of the batter hitting to their specific part of the field. If you want more detail click here.
Both UZR and DRS take into account some expected positioning, for instance UZR assumes deeper outfield positioning when above average power hitters are up and DRS takes into account first base positioning with respect to if they are holding a runner or not. Neither of them handle the new explosion of more extreme shifting very well yet. They both drop the data points for plays with “extreme” shifts, meaning that their play sample size is lower and we might be missing some positives and negatives for specific players. This will have more of an impact on players who play for teams that shift a lot.