⏰ Hourly Performance
Brain accuracy by time of day · the brain grows positions during strong hours, shrinks during weak ones
📊 Per-bucket performance
Accuracy / edge / size multiplier for each session bucket. Buckets with <20 samples show no multiplier yet (returns 1.0).
📚 Why time-of-day matters
Markets behave differently across the trading day:
  • 9:30–10:30 — Opening volatility, emotional flow, model often confused by noise.
  • 10:30–12:00 — Trends establish, informed money positions itself. Usually best edge.
  • 12:00–14:00 — Lunch lull, thin liquidity, choppy.
  • 14:00–15:30 — Algo positioning, institutions re-balance.
  • 15:30–16:00 — MOC auction approach, heavy flow, often unpredictable.
  • Outside RTH — Premarket / after-hours, sparse data.
This module stratifies prediction accuracy by bucket and computes a size multiplier in [0.5, 1.2] keyed to the historical edge. The Unified Predictor applies this multiplier to every position. So if the brain has historically been +6pp on mid-AM trades but -3pp on close-hour trades, mid-AM positions grow by 12% and close-hour positions shrink by 6%.

Compound effect: the hourly multiplier multiplies on top of the Bayesian dropout multiplier, ensemble agreement multiplier, OOD multiplier, and bootstrap divergence multiplier. The brain shrinks size compoundingly when multiple uncertainty signals fire.