FYI -- While doing some research on Saves and Win Probability Added, I noticed that the WPA values on Baseball-Reference.com are incorrect for 3 games from the 2011 season.
In the 3-game series in Seattle on June 24-26, Florida acted as the home team. For some reason, all the play-by-play events have the wrong sign -- things that aid the winning team have negative values in the wWPA column, and things that hurt the winning team have positive values.
Those values, in turn, are reflected in the individual player totals, and for some players this has a significant impact on their season WPA. For example:
-- On June 26, Seattle rookie Dustin Ackley was the hitting star of a 2-1, 10-inning win, going 3 for 5 with a triple and a double in the 10th that set up the winning run. His listed WPA for that game is minus-.393, which drags his season WPA down to a .076, 21st among MLB second basemen. Switching the sign on that one game's WPA puts his season WPA at .710, which ranks 17th.
-- In that same June 26 game, Seattle SP Doug Fister allowed 1 run in 8 IP, but his listed WPA is minus-.257. Switching the sign bumps his season WPA from 1.890 (19th among qualifying SPs) up to 2.404 (14th).
-- On June 24, Felix Hernandez allowed 1 run in 8 IP but is assigned a WPA of minus-.304. Correcting the sign raises his season WPA from .907 (34th) to 1.515 (25th).
This came to my attention when I searched for Saves that had a negative WPA. Two of the four "hits" came from that series, and neither pitcher allowed any runs (his own or inherited).
Here are the 3 games: