Home/Systematic Trading/Regime Detection

Regime Detection

regime-detectionLevel 2 — Intermediate

Unlocks

Liquidity Cascade AwarenessPortfolio ConstructionRegime IdentificationReturn Expectations and Capital Market AssumptionsReturn StackingSystematic MacroTactical Asset AllocationTrend Following

What It Is

Identifying the current market environment (growth/inflation, risk-on/risk-off, high/low volatility, trending/mean-reverting) so that strategy parameters, position sizing, and asset allocation can be adapted accordingly.

Correct Execution

Practitioner maintains a live, multi-signal picture of regime state: at minimum volatility level (VIX / realized vol), trend direction (moving average slope), and macro backdrop (growth vs. inflation quadrant). Regime state is updated systematically — not discretionary — and drives explicit rule changes in the strategy, not gut-feel overlays.

Progression Levels

Diagnostic Tree

Coaching Cues

  • "Diversify your rebalance dates before you diversify your signals." — when timing luck is clearly large, Corey Hoffstein / Nick Raj framework
  • "What does the market structure look like, not just the price?" — before interpreting any regime signal, check positioning of systematic participants
  • "Ask what game you're playing before you pick a filter." — trend-based vs. mean-reversion vs. vol-targeting, Rodrigo Gordillo

Common Errors

  1. Binary regime thinking: Treating markets as either "risk-on" or "risk-off" → oversimplification ignores the four macro quadrants (high growth/low inflation, high growth/high inflation, low growth/low inflation, stagflation) → use a 2×2 growth/inflation grid as baseline.
  2. Single-indicator dependence: Using only a moving average → high sensitivity to exact parameter choice → combine with at least one independent signal (vol, breadth, or macro).
  3. Ignoring structural market change: Calibrating a regime model on 1990s data and expecting it to work identically now → passive investing, options market growth, and target-date funds have changed market dynamics → periodically re-examine whether your regime's historical behavior still holds.
  4. Treating regime detection as a trading signal: Using regime state to call entries/exits rather than to scale risk → leads to whipsawing → use regime for risk sizing, not trade timing.

Edges

⚡ Conventional Wisdom Is Wrong

Split Your Rebalance Schedule Before You Search For Better Signals

regime-detectionregime-detection →

When a regime filter causes large drawdowns due to late detection, the instinct is to find a better signal. The actual fix is to split capital across two rebalance schedules (e.g., monthly and weekly). The issue is timing luck, not signal quality — the filter rebalances at a fixed date and a crash can occur mid-period before the signal updates.

What most people do
Hunt for a better or faster regime signal when the current one triggers late. Invest in more sophisticated macro or price models.
What the best do
Diversify rebalance timing by splitting the same signal across two offset schedules, halving timing luck without any signal improvement.
Why it's an edge: Eliminates a large source of performance variance that looks like signal failure but is actually calendar luck. Almost no practitioners think to fix the rebalance schedule rather than the signal.
How to exploit: Run any regime filter on two offset rebalance schedules (e.g., 1st and 15th of month). Observe the improvement in max drawdown variance across both schedules. No new signal research required.
Cross-domain parallel
In sports betting, spreading bets across multiple line-move windows rather than one fixed entry time reduces timing luck without needing a better handicapping model.
Corey Hoffstein, "What is Signal Timing Luck? (Regime Filters)," YouTube, 2025-11-07
🔑 Hidden Causal Lever

Systematic Participant Positioning Is The Real Regime Signal

regime-detectionregime-detection →

Macro regime models (growth/inflation quadrant, price trend) miss the dominant driver of the worst drawdowns — simultaneous mechanical de-leveraging by target-vol funds, CTAs, and risk-parity managers. These players don't respond to fundamentals; they respond to volatility, and when they all fire simultaneously, the cascade is endogenous.

What most people do
Build regime filters from price signals and macro data. When the filter fails during a cascade, assume the signal needs improvement.
What the best do
Overlay a positioning model on top of price-based signals. Track estimated leverage of systematic participants. When all are maximally levered, defensively tilt regardless of the macro signal.
Why it's an edge: Correctly identifies when the risk is structural (crowded positioning about to unwind) vs. fundamental. Macro models are blind to this.
How to exploit: Track published CTA net-exposure proxies and vol-targeting fund leverage estimates. When the composite is at extreme long, reduce net exposure as a risk management overlay — not as a trade signal.
Corey Hoffstein, "Liquidity Cascades," Investment Magazine, 2021-07-10
⚡ Conventional Wisdom Is Wrong

Trend Filters Generate Negative Expected Value in Mean-Reverting Markets

regime-detectionregime-detection →

Practitioners apply trend-based regime filters universally across market conditions. But a trend filter has negative expected value in mean-reverting regimes — it does not simply stop working, it actively destroys alpha by whipsawing entries and exits.

What most people do
Use a trend-based regime filter as a universal on/off switch, applying it regardless of the prevailing autocorrelation structure in the market.
What the best do
Use different filter types for different regime characteristics. In mean-reverting markets, switch to vol-targeting or carry-based exposure management rather than forcing a trend filter.
Why it's an edge: Identifies that the filter type must match the market structure, not just the practitioner's comfort with a familiar tool.
How to exploit: Separately measure return contribution of the regime filter during high-autocorrelation vs. low-autocorrelation market periods. If contribution is significantly negative in low-autocorrelation periods, replace the filter with a different mechanism for those conditions.
Rodrigo Gordillo, Return Stacking podcast, 2021-11-15; "Trend vs. Carry," YouTube, 2024-09-12

Sources

  • Corey Hoffstein, "Liquidity Cascades" (Investment Magazine, 2021-07-10) — market structure effects on regime behavior, pro-cyclical systematic participants
  • "What is Signal Timing Luck? (Regime Filters)" YouTube, 2025-11-07 — signal timing luck, split-schedule solution
  • Rodrigo Gordillo & Corey Hoffstein, "Return Stacking" podcast, 2021-11-15 — regime-aware portfolio construction
  • "Trend vs. Carry: Understanding Market Agnostic Approaches" YouTube, 2024-09-12 — carry as regime-agnostic alternative to trend filters