WARriors, Assemble!!!
This is meant to be a quick companion to Chris’ WAR series to give you the general gist of what that stat is and how it’s used. An in-depth explanation will probably come some time in the semi-distant future, for now this is just meant to help you understand what that WAR number means.
Wins Above Replacement (WAR) measures exactly what it sounds like: the number of wins above a hypothetical replacement level. We like this because wins is the ultimate unit of measure in baseball. Looking at runs created or prevented can be interesting, but if we want to say a lot with only a few words, Wins is the measure of choice. WAR helps us do that.
Replacement level is a crucial concept in all places statsy, it’s an approximate measure of freely available talent. Think AAA players, waiver wire finds, Rule V picks, etc. Basically anyone who’s paid league minimum that isn’t a prospect. A team of all replacement level players is expected to have a win percentage of .360. Replacement level players contribute 0 WAR over a full season of play which is generally assumed to be 150 games played. League average players will contribute 2 WAR. All-Stars generally contribute 4-5 WAR and the elite of the elite can push 8-10 WAR.
For most positions, defining that replacement level is easy. If Roy Halladay and his 8 WAR get hurt today, he’s replaced with Kyle Kendrick’s 0 WAR. If Chase Utley is hurt, his 8 WAR is replaced by Juan Castro’s (arguably less than) 0 WAR. So with starting pitchers and position players, the replacement player is ultimately the guy who replaces the injured player on the roster.
The bullpen is a little more confusing. At first blush you might think that if Brad Lidge and his 1.7 WAR (Chris’ number) got shelved for the year, then he would be replaced with 0 WAR Scott Mathieson (just an example). However, think about bullpen usage. If the closer is hurt, you don’t make the new call up your closer, you make Ryan Madson the closer. Again using Chris’ estimates for the bullpen, you lose .2 WAR over the course of the season by making the switch from Lidge to Madson. Now you don’t have a right handed set up man so Danys Baez is promoted to that role, a loss of .9 WAR. Jose Contreras is moved from 2 inning relief to the 7th inning role, a gain of .2 WAR. Romero, Durbin, and Bastardo’s roles are largely unaffected. That leaves the new guy mopping up inconsequential innings. The replacement really doesn’t mean anything to you in the 8th inning of a 14-5 game.
By swapping a replacement level player in for Lidge, the Phillies bullpen would lose 1 WAR rather than 1.7. Because of the way the bullpen is used, replacement level in the bullpen is different than at other positions. This concept is called Chaining.
If you want to use WAR in an English sentence, it would go something like this:
The Phillies will be 8 wins worse if Roy Halladay is hurt today for the rest of the season and is replaced by Kyle Kendrick.
-or-
The Phillies would be about 6 and a half wins better if they had kept Cliff Lee at the expense of Jamie Moyer.
See, those sentences make sense, right?


This is probably the one stat I felt comfortable enough with but the bulpen scenario was definity helpful. How would it be affected in the case of platoobing of’s that together put up above avg numbers ala eisenreich and milt thompson? kinda rare in this day and age but just a thought.
I’m not quite sure what the question is. WAR as we see it on fangraphs is just a hypothetical baseline. WAR in reality should reflect the team’s options. If Eisenreich was .5 WAR better than Thompson but 2 WAR better than the hypothetical replacement, then his fangraphs WAR (fWAR) would read 2 while his WAR to the team would be .5.
Without getting too technical, position players accrue WAR based on playing time (replacement), their position, their wOBA, and their UZR. When you 50/50 platoon two guys in RF the result is that you halve the playing time/replacement adjustment, you slice the corner OF adjustment in half, their wOBA should be a little higher, and their UZR would be even less predictive of true talent.
Platoons are an interesting challenge in that the WAR collected doesn’t usually represent a true talent level. Most platoobs these days are pretty strict L/R so we’re getting inflated numbers for both players (assuming they have a platoon split).
I hope any of that helped.