Macro Regime Classification Filter

The thesis

Almost every quantitative strategy has regime-dependent performance. Momentum works best in trending markets and underperforms during regime transitions. Mean reversion works best in stable correlation regimes and breaks down during stress. Insider buying works in most regimes but produces faster pay-offs in risk-on environments. A serious systematic stack does not pretend regimes do not exist — it identifies them and adjusts.

Empirical basis

Regime-aware portfolio construction is a well-developed area in quantitative finance, drawing on work by Ang and Bekaert (2002, 2004), Guidolin and Timmermann (2007), and others on Markov-switching models for asset returns. The Alpha Suite implementation uses a more practical five-factor heuristic rather than a fitted regime model, which sacrifices some statistical sophistication for robustness across different market environments.

How Alpha Suite implements it

The macro regime filter is computed daily from five inputs:

• VIX level and term structure — absolute VIX, plus front-month vs. 3-month relationship; backwardation indicates stress.
• Yield curve — 2s-10s and 3m-10y spreads; flattening or inversion signals late-cycle conditions.
• Credit stress — high-yield ETF (HYG) trend and HYG/LQD ratio; deterioration is a leading indicator for equities.
• Market breadth — percentage of S&P 500 components above their 50-day and 200-day MAs; narrowing breadth flags weak internals even when the index is rising.
• SPY trend — price relative to the 50-day and 200-day MA; basic confirmation that the broad index is in an uptrend.

Each factor scores 0 (defensive), 1 (neutral), or 2 (risk-on). The composite score determines the global position-sizing multiplier across every Alpha Suite signal:

• Risk-on regime (8–10): 100% sizing, all strategies active.
• Neutral regime (5–7): 70–90% sizing, all strategies active.
• Defensive regime (3–4): 40–60% sizing, momentum strategies down-weighted, mean-reversion up-weighted.
• Crisis regime (0–2): 0–40% sizing, only highest-conviction signals taken, paper-position cap reduced.

How it interacts with the rest of the system

The regime filter is multiplicative rather than binary. It does not turn strategies off; it scales them. The Form 4 insider engine still ranks candidates in a defensive regime, but the resulting position sizes are smaller. The pairs trading engine actually increases sizing in late-cycle regimes because the underlying mean-reversion edge often improves when correlations break down. The regime classifier is also exposed via the dashboard so users can see what regime the system thinks the market is in at any moment.