Re-evaluate the possession-quality index under alternative velocity windows

Determine whether the possession-quality index’s velocity term is harmless at the previously used 0.24-second finite-difference window and beneficial at a 0.6-second window by re-running the index on its own broadcast data under both settings.

Background

The possession-quality index used in earlier work computes threat-weighted control from broadcast tracking with 0.24-second finite-difference velocities. The present benchmark suggests that this short window may be beyond the velocity-noise break-even point, while a 0.6-second window could make velocity beneficial. However, the benchmark does not re-run the index on its original data, so the interpretation remains unresolved.

The authors distinguish the benchmark’s surface-level findings from the index’s own end-to-end behavior and explicitly identify the need to test both windows on the index’s data.

References

Our reading --- that the index's velocity term was harmless rather than helpful at that window, and that a 0.6\,s window would have made it helpful --- is a hypothesis: we have not re-run the index at 0.24 and 0.6\,s on its own data, and the real-broadcast axis here is at 0.6\,s.

— How Much Velocity Does Off-Ball Space Value Need? A Broadcast-Viewport Benchmark  (2609.10801 - Choi, 9 Sep 2026) in Section 6, “Re-reading our earlier indices”