risk-management

Risk Management

Overview

Risk management in TradeMemory is behavioral, not just mathematical. Traditional risk management calculates position sizes and stop losses. TradeMemory adds a behavioral layer: it monitors your execution patterns, detects emotional drift, and flags when you're deviating from your own rules.

The system tracks two kinds of risk:

1. Position risk — How much capital is at stake on each trade
2. Behavioral risk — Are you making decisions rationally or emotionally

Affective State Model

TradeMemory maintains a real-time emotional state model for the trading agent:

Dimension	Range	What It Tracks
Confidence	0.0 - 1.0	Self-assessed confidence, calibrated against outcomes
Drawdown	0% - 100%	Current peak-to-trough equity drawdown
Win Streak	0 - N	Consecutive winning trades
Loss Streak	0 - N	Consecutive losing trades
Risk Appetite	low / normal / high	Derived from confidence + drawdown + streaks

How Affective State Updates

• After a win: Confidence += f(P&L magnitude), win streak ++, loss streak reset
• After a loss: Confidence -= f(P&L magnitude), loss streak ++, win streak reset
• Drawdown crossing thresholds: Risk appetite auto-reduces at 5%, 10%, 15% drawdown
• Daily review: Confidence recalibrated against actual hit rate

Using Affective State

Check get_agent_state before every trading session:

get_agent_state() → {
  confidence: 0.42,
  drawdown: 8.3%,
  win_streak: 0,
  loss_streak: 3,
  risk_appetite: "low"
}

Action rules:

• risk_appetite == "low" → Reduce position size by 50% or skip marginal setups
• loss_streak >= 3 → Stop trading for the session. Review, don't revenge trade.
• confidence < 0.3 → Paper trade only until confidence recovers
• drawdown > 15% → Hard stop. No new positions until daily review.

Behavioral Risk Indicators

1. Disposition Effect

What: Cutting winners short and holding losers too long. Detection: get_behavioral_analysis → disposition_ratio

• Ratio < 1.0 = Good (holding winners longer than losers)
• Ratio > 1.5 = Problem (losers held 50% longer than winners)
• Ratio > 2.0 = Critical (classic retail trader failure mode)