Sector / cohort relative-strength tracker. Heatmap of cohort means × multi-window grid (30m / 1h / 2h / 4h) + ranked relative-strength vs SPY at primary window. Composes-with /api/movers/intraday-window v4 cohorts_per_window field — single fetch, full matrix.
Diagnostic surface — NOT a Brier-validated signal. Short-window cohort returns have low SNR (Lo-MacKinlay 1990); a 2h sector-leadership reading can be noise. Relative-strength vs SPY is informational per Signal-Confluence-Matrix T0; treat as context, not direction. Tagged [DIAGNOSTIC-pending-Brier-N≥30-validation] per Pre-Launch Validation T0.
Primary window
Cohort × Window heatmap
Relative strength vs SPY (primary window)
How this works: Each column is one canonical window. Each row is one manifest cohort from data/movers-intraday-universe.json (megacap_tech, sector_etfs, fixed_income, etc). Cells show cohort mean Δ% for that (cohort × window) pair — so at a glance you can see which sectors led the last 30m vs which led the last 4h. Color intensity scales with magnitude: 5-band {strong-neg, neg, zero, pos, strong-pos} at ±0.5% / ±1.0% thresholds.
Relative strength vs SPY = cohort_mean − SPY_delta at the primary window. Positive = cohort outperforming the broad market. Frazzini-Pedersen 2014-style cross-sectional ranking, but at intraday horizon. Methodology hash: movers-intraday-window-2026-05-08-v4 (handler) + intraday-window-substrate-2026-05-08-v1 (substrate).