Adaptive Flow
Adaptive Flow is the final stage of the strategy pipeline. After a strategy passes Walk-Forward Analysis, Adaptive Flow monitors its performance against live market data and automatically triggers re-optimization when the strategy's edge degrades.
Pipeline Position
Backtest → Optimization → WFA → Adaptive Flow
You can only enter Adaptive Flow for a strategy that has completed WFA with acceptable results.
How It Works
- WFA results establish an expected performance envelope — statistical bands for win rate, Sharpe ratio, drawdown, and profit factor based on out-of-sample windows
- You enter Pre-Live mode which shows readiness checks and expected performance bands
- Starting live validation runs a fresh optimization using the optimal training window, then begins monitoring
- Real-time price feeds drive strategy signal computation on each candle close
- The system continuously compares actual vs expected metrics
- If performance degrades beyond thresholds, re-optimization is proposed
Key Metrics
| Metric | Description |
|---|---|
| Health Score | 0–10 composite rating of forward robustness |
| Confidence % | Statistical maturity of the live validation data (increases as more candles are processed) |
| Deviation Score | How far live metrics have drifted from expected. WARNING at >= 0.60, DEGRADING at >= 0.80 |
| Expected Envelope | Win rate, Sharpe, drawdown, and profit factor ranges derived from WFA out-of-sample results |
Runtime States
| State | Description |
|---|---|
| Pre-Live | Strategy eligible but not yet monitoring. Shows readiness checks and expected bands. |
| Collecting | Just started, confidence below 25% |
| Early Sample | 25–70% confidence, partial data |
| Active | Normal monitoring (warmup complete) |
| Warning | Deviation score >= 0.60 — performance drifting |
| Degrading | Deviation score >= 0.80 — significant performance loss |
| Re-optimizing | Re-optimization is running |
| Stabilized | Post re-optimization, accelerated warmup |
Expected vs Actual
The core of Adaptive Flow is comparing six metrics against their expected envelopes:
- Win Rate — percentage of winning trades
- Sharpe Ratio — risk-adjusted return
- Trade Count — number of trades generated
- Average Return — mean return per trade
- Max Drawdown — largest peak-to-trough decline
- Profit Factor — gross profit / gross loss
Each metric is shown with the expected range and the actual live value, making it easy to see where performance is holding and where it is diverging.
Re-Optimization
When performance degrades beyond thresholds (or a configurable candle countdown is reached), the system proposes a re-optimization:
- You receive a notification with the proposed re-optimization
- You can approve, reject, or defer the proposal
- If approved, a new optimization runs using the latest market data
- New parameters replace the current ones and monitoring continues
Re-optimization uses the same parameter ranges and optimization target from your original WFA configuration.
Multi-Asset (Pack) Mode
If your strategy was validated across multiple assets in WFA, Adaptive Flow can monitor all assets simultaneously:
- A Pack Summary view aggregates health across all assets
- You can switch to individual Asset View for per-symbol details
- Pack health uses median aggregation penalised by the weakest asset
Tips
- Let confidence reach at least 70% before drawing conclusions about live performance
- A Health Score above 7 indicates the strategy is performing within expectations
- If deviation reaches WARNING, watch closely but do not intervene immediately — short-term variance is normal
- If deviation reaches DEGRADING, consider approving re-optimization promptly
- Compare the equity curve overlay (expected band vs actual line) for a visual sense of how the strategy is tracking