Current Market Regime Β· 3-state Gaussian HMM on SPY (quant-lab)
Loadingβ¦
Fetching the baked regime state from /api/regime-gateβ¦
Regime confidence
β
smoothed posterior P(regime | all data)
Confidence
β
Days in regime
β
Since
β
Data
β
Gate freshness
β
π Last 120 Trading Days β SPY, Regime-Shaded
SPY close over the last 120 trading days, with the background shaded by the HMM's decoded regime β green Strong Bull, red Strong Bear, amber Choppy. Long same-color runs are stable regimes; rapid flips mean the tape is transitioning.
Loadingβ¦
Compact strip view β each slice is one trading day:
Loadingβ¦
ββ
π― Strategy Styles β In Season vs Benched
The DynamicAllocator only deploys strategy styles that historically work in the current regime: mean-reversion in Choppy, trend + momentum in Strong Bull, defensive/cash in Strong Bear. Everything else sits on the bench until the regime turns.
β Favored now
Loadingβ¦
βΈ Benched
Loadingβ¦
π Per-Stock Gate Check
Type a ticker (or arrive with ?symbol=) to see whether it survived the six-filter walk-forward funnel, which styles carry its edge, and how that edge lines up with today's regime.
π Funnel Survivors β Validated Edges
Strategy Γ asset pairs that passed all six walk-forward filters (OOS Sharpe, drawdown, overfit ratio, trade count, hold time, window consistency) AND the bootstrap stress test. This is the entire tradeable edge inventory across the 30-asset universe β everything else failed validation. Bars are colored by style; the dashed amber line is the 0.5 OOS Sharpe funnel floor every survivor had to clear.
Loadingβ¦
Full survivor detail:
Loadingβ¦
π Regime Occupancy β Full Sample
How the HMM splits the whole ~15-year SPY history across the three regimes β the share of trading days spent in each.
Loadingβ¦
Loadingβ¦
π Learn this system
Two companion resources for the Regime Gate: the flipbook walks the 9,000-candidate strategy funnel this page reports on, and the video is the explainer for how this system was built.
π The 9,000 Strategy
Flipbook β how 9,000 candidate strategy Γ asset combinations get distilled through the six-filter walk-forward funnel into the survivor list above.
π¬ Building an Institutional-Grade AI Backtester with Claude
The explainer video for how this system was built β the quant-lab HMM regime engine, the walk-forward funnel, and the Regime Gate you are looking at.