In basketball, passes go to a player's hands — the destination is the player's current location. In football, the most valuable passes go to where the player will be in 1-3 seconds, not where they currently are. A through-ball played into the channel is aimed at a future position the receiver hasn't reached yet. Modeling this correctly requires predicting player movement trajectories and estimating the EPV of future positions, not just current positions. Anticipatory modeling is what separates a through-ball from a misplaced pass statistically.
Implementation: (1) at the moment of pass release, compute the expected position of all potential receivers 1, 1.5, and 2 seconds ahead using velocity vectors; (2) compute the EPV of each future position; (3) model pass value as the EPV of the destination position at arrival time, not at pass release time. This correctly values through-balls, runs in behind, and diagonal balls into space — all of which look wrong if you model destination as current player position rather than future position.
Most EPV computes destination value using the receiver's position at pass release. For through-balls, the receiver is running toward the destination — the model sees an "empty space" pass and assigns low value. Through-balls, runs in behind, and diagonal balls into space are systematically undervalued.