Detecting EA decay before it destroys your account

The most dangerous drawdown is not the one that happens fast. It's the one that happens slowly — a gradual drift from profitable to marginal to unprofitable, playing out over six months while you keep telling yourself it's just variance.

What EA decay actually is

EA decay is the gradual erosion of a strategy's edge over time. It's distinct from a temporary losing streak: where a losing streak is a statistical fluctuation within a still-functioning system, decay represents a structural change in the relationship between the EA's signal and market outcomes.

The distinction matters because the correct response is different. A losing streak warrants monitoring. Decay warrants intervention. The problem is that from inside the drawdown, they look identical — until one of them stops.

The three types of decay

Market regime shift

The most common type. Many EAs are optimized for specific market conditions — trending, ranging, high-volatility, low-volatility. When the dominant market regime shifts (e.g., a low-volatility choppy market transitions to a high-volatility trending market), strategies built for the old regime stop working. This isn't the EA breaking — it's the environment changing.

Curve-fitting decay

An EA over-optimized on historical data will show excellent backtest metrics but fail in forward testing because it learned noise rather than real patterns. Curve-fitting decay typically appears immediately in live trading — the performance gap between backtest and live starts large and often gets larger.

Structural market change

Some EA strategies work because of structural market inefficiencies — specific spread patterns, liquidity conditions, or broker arbitrage. When brokers change conditions, algorithm regulations tighten, or market microstructure evolves, these edges can disappear permanently. This is the rarest but most irreversible type of decay.

Early warning signals

• Profit Factor dropping below its 90-day rolling average for 30+ consecutive days
• Win rate declining without a corresponding increase in reward-to-risk ratio
• Equity curve slope flattening or turning negative over a 60-day window
• Average trade duration increasing without a change in strategy settings (EA is spending longer in losing positions)
• Trade count dropping sharply (EA is filtering out trades that would have triggered previously)
• Maximum adverse excursion increasing — trades are going deeper underwater before recovering or stopping out

Using Stability Score for early detection

Stability Score measures how closely an EA's actual equity curve tracks a smooth, ideal growth line. A score near 1.0 means the equity curve is almost perfectly linear — consistent, predictable growth. As decay sets in, the curve becomes increasingly jagged and deviates from the ideal line, dropping the Stability Score.

The useful property of Stability Score is that it's sensitive to changes in consistency rather than just absolute level. An EA can still be profitable (positive P&L) but show a declining Stability Score because its wins and losses are becoming more erratic. That early signal — profitability is fine but consistency is degrading — often precedes actual losses by weeks.

Pause, review, or kill?

Define thresholds in advance, before you're emotionally invested in a specific EA's outcome.

• Pause: Stability Score drops 25%+ from its 90-day average. Investigate market regime change. If regime shift explains it, wait for regime to return.
• Review: Profit Factor drops below 1.0 for 30 consecutive days. Analyze trade-level data for changed behavior patterns.
• Kill: Two or more signals trigger simultaneously, or equity drops 50%+ from all-time high. No EA is worth running at that point.