Raw market coverage can remain visible elsewhere, but selected-pick, no-vig, edge, calibration, and performance claims are withheld until the evidence gate clears.
Public evidence gate closed: Model Brier 0.658 is 0.015 worse than the no-vig closing-market baseline 0.643 on the shared 275-fixture cohort; threshold is 0.000. Public edge claims are blocked until model improves.