Test the proposed explanation for deep-zone imputation bias

Establish whether compression of the B2 team-centroid relative-offset anchor toward visible teammates causes the consistent deep-zone under-crediting of the defending team observed in the broadcast off-screen-player imputation benchmark.

Background

The benchmark observes a structured imputation bias: the off-screen anchor places hidden defenders too high, producing substantial under-crediting of the defending team in its own deep zone. The paper interprets this pattern as resulting from the anchor being pulled toward visible teammates near the ball.

That causal interpretation is not independently tested. Confirming or rejecting it would clarify whether the observed bias arises from the anchor mechanism itself and would guide the design of learned occluded-player imputers.