Adaptive Failure Learning Engine
Self-learning failure ingestion, pattern classification taxonomy, and pre-trade memory gating for automated capital defense.
Continuous Failure Learning Lifecycle
Whenever a paper trade closes with a realized loss, Zenth analyzes the context leading to the loss, extracts the root pattern, and writes a synthesized rule into active memory.
Failure Pattern Taxonomy
Common market traps classified and filtered by the learning engine:
Golden crossover on volume < 80% of 20-period average, causing immediate false breakout whipsaw.
Golden crossover with RSI > 70, resulting in overbought momentum reversal before target.
Oscillating MA crossover within tight 0.3% price corridor across 10 consecutive candles.
Entry near 24h high with negative taker volume divergence.
The 4 Adaptive Filtering Modes
Control how aggressively active learned failure rules gate incoming trade signals:
| Mode | Execution Action on Pattern Match | Recommended Purpose |
|---|---|---|
| STRICT (Default) | Instantly converts [BUY] signal to [SKIP] | Production paper trading with maximum capital protection |
| REPEAT_LOSSES | Blocks signal ONLY after pattern caused >= 2 losses | Allows single anomaly while filtering repeating traps |
| DRY_RUN | Logs match telemetry but allows paper trade to execute | Observing pattern impact without intervening |
| DISABLED | Disables memory filter completely | Raw baseline strategy backtesting & benchmarking |