Market Regime Detection

Market regimes are the most important contextual factor in determining which trading strategies are appropriate at any given time. A trend-following strategy applied in a ranging market will consistently lose. A mean-reversion strategy applied during a strong trend will get run over. Regime awareness is what separates adaptive trading systems from static ones.

Why Regimes Matter

Static trading systems apply the same logic regardless of market conditions. This works when the training environment matches the live environment — but markets evolve. Regime-aware systems adjust their behavior to match current conditions.

TRADEOS.tech classifies the market into one of several regime states and uses that classification to:

• Enable or disable specific signal types
• Adjust position sizing (more aggressive in favorable regimes, more conservative in uncertain ones)
• Change execution behavior (wider stops in trending regimes, tighter in ranging)
• Gate or pass signals from the confluence filter

Regime Classification Approach

TRADEOS.tech uses two complementary methods to classify market regimes:

Hidden Markov Model (HMM)

The HMM is a statistical model that assumes the market moves through a series of "hidden states" — underlying market conditions that you cannot observe directly, but which manifest in observable data like returns, volatility, and volume.

The model is trained to identify states such as:

• Trending (bull): persistent upward price movement with characteristic volatility signature
• Trending (bear): persistent downward price movement
• Ranging: mean-reverting behavior, price oscillating within a band
• High volatility / uncertain: elevated volatility without clear directional bias
• Crisis / dislocation: extreme volatility, correlation breakdown, structural instability

The HMM outputs a probability distribution across these states. TRADEOS.tech acts on the most probable state while accounting for regime uncertainty.

Bayesian Online Change-Point Detection (BOCD)

BOCD detects when regime transitions occur — the moments when market character shifts. Rather than waiting for the HMM to gradually update, BOCD provides early warning that conditions are changing.

When a change-point is detected:

• Current signal weights are re-evaluated
• Position sizes may be temporarily reduced (uncertainty regime)
• New entries may be paused until the new regime is confirmed

Combined Output

The two methods are combined: HMM provides regime classification, BOCD provides transition timing. Together they produce regime state estimates with explicit confidence levels. Low-confidence regime classifications trigger conservative position sizing regardless of signal strength.

Regime-Gated Signals

Not all signals are active in all regimes:

Signal Type	Trending Bull	Trending Bear	Ranging	High Vol
Trend following	✅ Full size	✅ Full size (short)	❌ Suppressed	⚠️ Reduced
Mean reversion	❌ Suppressed	❌ Suppressed	✅ Full size	⚠️ Reduced
Order flow	✅ Active	✅ Active	✅ Active	⚠️ Reduced
On-chain	✅ Active	✅ Active	✅ Active	✅ Active
Funding carry	✅ Active	✅ Active	✅ Active	⚠️ Reduced

This regime-gating is one of the primary mechanisms by which TRADEOS.tech avoids trading in unsuitable conditions.