Validate robust-scale prediction of velocity recovery across trackers

Determine whether a robust-scale statistic of velocity error can predict pitch-control recovery across broadcast trackers with different tail mass, bias, and player-correlation structures.

Background

The paper finds that the RMS of velocity error is insufficient to explain the observed recovery of pitch-control performance for its broadcast pipeline: structured, frame-common, and heavy-tailed errors perform substantially better than Gaussian controls matched on RMS. The authors therefore suggest that a robust-scale measure incorporating the residual error structure might be more informative, but they explicitly leave its cross-tracker predictive validity unresolved.

The unresolved issue is whether such a statistic would generalize beyond the single pipeline and single-match broadcast sample evaluated in the benchmark, particularly when trackers differ in tail behavior, systematic bias, and cross-player error correlation.