is where you assume lineups filled with average hitters. (Someone correct me if I’m wrong here, but I believe this is what fangraphs uses.) So at the start of every game, you go in figuring the odds of winning are 50-50.

Now—your average hitter will hit something like (this is very roughly accurate but only very roughly, don’t taze me) 17% singles, 4% doubles, 0.5% triples, 2.5% home runs, 9% walks/HBP, and 66% outs. Each of those scenarios creates a new situation—whether it be a runner on first with no outs, no runners on and 1 out, or no runners on, no outs, and a 1-0 score.

Then the next batter comes up. Things can get a little more complex here because there are more possible outcomes (example: a single with a runner on 2nd can create a 1st and 3rd situation, or score a run and leave a man on first) but there’s still a certain percentage of the time that each of those things will happen, which you can then multiply by the probability of them happening the first time around to get the probability of two particular events happening in conjunction. For instance, .66*.66 of the time, the first two hitters will make outs.

You can string these probabilities (called Markov chains) together until eventually you’ve got a whole baseball game played out and someone has won and lost. And because every hitter is an independent event, you don’t have to start from scratch. You can start also from the bottom of the ninth with the road team up by three, and keep sending up fictional average hitters until one side wins or loses. Do that a million times and you’ve got a pretty good idea of how often the home team will win.

Now suppose you do this and find out that the home team wins 4% of the time in that scenario. Then you try it again with one out already recorded in the inning, and find that the home team only wins 1% of the time. You can say then that making that first out lowered the home team’s win probability by 3%—or a -0.03 Win Probability Added. On the other hand, say the first batter singles. Now you play out another million fictional games and find the home team wins 10% of the time. So that’s a +0.06 Win Probability Added.

Well, someone actually did this—went through each of the 24 base/out situations (8 base situations times 3 out situations), in each of the 18 half-innings, with each of the possible scores (ok, to a point; I doubt someone calculated WP for a 15 run lead) and ran those average hitters through from that situation to game’s end a zillion times to figure out how often each team won. Figure out how often the batting team wins before the batter hits; then figure out how often it wins after whatever happened when he hit; then subtract the first from the second and you’ve got WPA. Then take all of a hitter’s at-bats and what happened in them and add those up, and you’ve got his WPA for the season.

The stat is not a very good one for predicting future performance, because it’s heavily affected by “clutch” hitting performances. For instance, Marco Scutaro’s home run on April 15 of last year, with 2 outs in the bottom of the ninth, 2 men on first and second, and his team down 2, was worth over 90% of a win in WPA terms. Whereas a homer to lead off a game might only be worth 7% of a win in WPA. Was that Scutaro homer really worth 13 “normal” home runs? In win terms, yes. In terms of who you want to sign as your slugger for next season, not so much.

I don’t think its that unwieldy, “(OBP*1.8+SLG)/4, adjusted for ballpark factor” according to THT.

You’re right, though, EQA is better. Jack Cust is currently 5th in the AL (among qualified) with a .310 EQA.

Brown is simply not a starting-caliber player against right-handed pitching. The problem is that right now the team does not have anyone who IS starting caliber to fill that spot. If Fiorentino was healthy I’d suggest giving him the shot at it. But he, Linden, Denorfia, Buck and Gonzalez are all either injured or coming off injury and struggling.