Every feature in MorningEdge is validated through live paper trading before going to production. This page tracks what's been built, when it was validated, and where the evidence lives.
14 features built — 10 validated in production, 4 in validation or pending integration.
Validated Features
| Feature | Status | Validated | Evidence |
|---|---|---|---|
| 5-Stage Gap Scanner 11,400+ tickers → 5 picks through price, gap, spread, volume, and catalyst filters |
✓ Production | Feb 26 (Day 0) | Scanner Architecture, Pipeline Spec |
| Market Regime Filter 6-signal traffic light gate — GREEN/YELLOW/RED. Only trade GREEN days. |
✓ Production | Feb 27 (Day 1) | Traffic Light Gate, Exit Strategies |
| V3 "Unshackled" Exit System ATR disaster floor + 15:20 time exit. Sharpe 4.53 OOS on 160 unseen days. |
✓ Production | Feb 27 (Day 1) | Sell-Timing Analysis |
| Multi-Source Catalyst Intelligence 6 sources: Alpaca, Finviz, Yahoo, Perplexity, EDGAR, Grok. Eliminates blind spots. |
✓ Production | Feb 27 (Day 1) | Catalyst Intelligence, Day 0 Validation |
| DAY Market Orders Replaced OPG orders after Day 0 failure. 100% fill rate in paper trading. |
✓ Production | Feb 26 (Day 0) | OPG vs DAY Analysis |
| Two-Phase Limit Close Limits at ask first → market fallback. Saves ~0.05% per day on exit slippage. |
✓ Production | Feb 27 (Day 1) | Day 1 Results |
| ML Gap Predictor v1 Random Forest hold/fade classifier. 68.6% accuracy. SPY direction = top feature. |
✓ Production | Feb 27 (Day 1) | ML Predictor Report, Model Spec |
| Intraday Position Manager V3 auto-manager: monitors ATR floor + fires 15:20 time exit. Runs unattended. |
✓ Production | Feb 27 (Day 1) | Sell-Timing Analysis |
| Intraday Bar Collector 5-min OHLCV bars to Parquet. 4.8M bars across 211 symbols, 250 trading days. |
✓ Production | Feb 26 (Day 0) | System Test |
| 90-Day Backtester V1/V2/V3 exit strategies with configurable slippage and out-of-sample validation. |
✓ Production | Feb 26 (Day 0) | Exit Strategies |
In Validation
| Feature | Status | Target | Evidence |
|---|---|---|---|
| Shadow Tracker Tracks hypothetical P/L on YELLOW/RED sit-out days. Validates the regime filter. |
In Validation | 30 sit-out days ~Mar 30 projected |
Day 2: +$341 hypo 1/30 days collected |
| Late Entry Sweep YELLOW→GREEN upgrade at 10:00 AM. Waits for regime improvement before entering. |
In Validation | 5 YELLOW→GREEN upgrades 0 events so far |
Late Entry Sweep |
Pending Integration
| Feature | Status | Plan | Notes |
|---|---|---|---|
| ML Gap Predictor v2 Ensemble (LR+GBM) with triple-barrier labeling, half-Kelly sizing, SHAP explanations. |
Built | Run in shadow alongside v1 for 2 weeks | Training complete. Needs production integration. |
| Opening Range Breakout (ORB) Separate strategy module for opening range breakout entries. |
Built | Backtest against V3 baseline | Module complete, pending comparative analysis. |
| Relative Strength vs SPY Intraday alpha tracking with RS_Score for each position vs market benchmark. |
Built | Integrate as exit signal modifier | Tracking data collecting. RS > 1.0 correlates with winners. |
Validation Methodology
Features move through a four-stage pipeline:
- Hypothesis — Identify a potential improvement based on trading theory or observed gaps
- Implementation — Build the feature with unit tests (274 tests across 9 files)
- Backtest — Validate against 90-day historical data + 160-day out-of-sample window
- Live Validation — Paper trade in production with real market data. Document results.
Each validated feature has associated research documentation in the Trading Diary, creating a traceable audit trail from hypothesis through validation.
Live Results Summary
| Metric | Backtest (90d) | OOS (160d) | Live (3 days) |
|---|---|---|---|
| Total P/L | +$12,177 | +$17,029 | +$1,081 |
| Sharpe Ratio | 6.05 | 4.53 | — |
| Win Rate | 55.6% | 57.6% | 70% (7/10) |
| Max Drawdown | -$1,377 | -$3,975 | $0 |
| Trading Days | 90 | 160 | 2 GREEN + 2 sit-out |