Iteration 75 — ADMIT #14: the Wasserstein Meter CERTIFICATE #14 · BEYOND-NULL 4.0× · 21 GATE SLOTS

In plain words: the transport-geometry meter — which weights how far probability mass must MOVE between regimes, so tail events cost more — passed its final held-out test with the clean margins its kernel cousin lacked. Its forecast improvement is not just statistically positive: it is four times larger than anything random re-pairing produces (certificate #13's improvement sat INSIDE that noise band), and 3.2 times the size of #13's. The era check ran in the same breath (now standard procedure): the improvement holds in both halves of the held-out period, no sign flip, stronger early — the same profile #13 showed, now seen twice and flagged as a pattern. Its direction reading is the THIRD distributional meter to say the same thing: features whose behavior shifts most across regimes tend to LOSE their edge — six data points now behind the both-extremes-lose mechanism. And a pointed note for the operator's pending question about raising the admission bar: under the stricter bar being considered, #13 would have failed exactly where this one passes.