← all research

shadows scored: 61/64 (3 immature/skipped), names total: 1748

mean IC: +0.0001 (t=+0.00, n=61, bootstrap p=0.5046) sign hit-rate: 53.0% (null 50%) top-half minus bottom-half realized: +0.1739% per 21 sessions

per-date IC (first 10): 2024-01-15:+0.28, 2024-01-31:+0.43, 2024-02-15:+0.06, 2024-02-29:-0.67, 2024-03-15:+0.15, 2024-03-31:+0.05, 2024-04-15:-0.00, 2024-04-30:-0.07, 2024-05-15:+0.01, 2024-05-31:+0.38

VERDICT: WEAK (positive but insignificant — keep shadow, no enforcement) Caveat: backfilled windows may overlap the model's pretraining corpus; treat as directional, confirm on live forward shadows.