Each player is valued on what he was expected to produce at the time of the trade, versus what he actually produced for his new team.
Expected WAR blends a player's pedigree (Baseball America rank / draft slot, or a baseline) with his recent track record, projected over the years of team control acquired. The ± band is the uncertainty — wide for unproven prospects, tight for established veterans. Surplus values that production at the FA market price of a win (~$8M/WAR) minus salary — so cost-controlled players carry large surplus and expensive ones little, even at the same WAR. Who won is descriptive, not a skill claim: ~99% of a trade's outcome is unforeseeable at the time.
Historically these expected values are unbiased and land within ±2 WAR of reality 75% of the time — yet the side the model favors actually comes out ahead only 53% of the time. The grade is a calibrated bet, not a prediction. Why trades are an efficient market →
TBR⇄
SEA⇄
STL