Queue Bayesian priors as Edge Hunt 2 factorial dimensions
Append new section to docs/specs/edge-hunt.md describing how to promote the 11 Bayesian priors hypotheses into the pod-shop factorial search.
Append new section to docs/specs/edge-hunt.md describing how to promote the 11 Bayesian priors hypotheses into the pod-shop factorial search.
Rationale: standalone failure != useless. Per the pod-shop framework, signals are evaluated by portfolio contribution (alpha/beta/IR). A signal that fails standalone magnitude threshold can still add unique alpha if uncorrelated with existing signals.
Phase 1 (~8h): 16-signal factorial with existing solver cache + new promotedAdj + eloEarlyPrior. Tests: H7, H3, H9 (via shadow backfill).
Phase 2 (~48h): solver-variant factorials for 4 priors (elo/standings/ squad-value warm-start + squad-value regularization target). Tests: H1, H2, H4, H5 in combination with every other signal.
Factorial signal list now includes promotedAdj and eloEarlyPrior. Ready for burst-runner dispatch to compute-worker on Hetzner.
Spec: docs/specs/edge-hunt.md — "Edge Hunt 2: Bayesian Priors"
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>