Project Rho/
Rule Workshop
Tata Motors Pullback Guard
v3 · TATAMOTORS · 4 tuned parameters · NSE illustrative history
LOOK-AHEAD PREVENTED
Every signal sees only prices and filings available before its decision timestamp. Zero future bars visible.
624 / 624 passed
17 future reads rejected
Rolling Validation Map
Tune once. Freeze parameters. Evaluate forward.
Every window shares one axis. The vertical hand-off divider marks where parameters stop learning and the unseen forward testing begins.
In-Sample TuneOut-of-Sample Evaluate
Select active window to inspect in-sample tuning and forward evaluation split panes:
Dual-Pane Walk-Forward Split Chart · resizable.tsx
In-Sample Parameter Tuning vs. Out-of-Sample Forward Verification (W1)
Drag divider to adjust split ratio
In-Sample TuningOct-Dec 2025
6 FiringsTuned Return
+4.6%
Win Rate
71.4%
Profit Factor
2.18
In-Sample DD
-0.9%
Optimization Trajectory (T0 → T5)Base: 100.0
Frozen Params:Lookback: 20d EMARSI ≤ 30ATR 1.8x
Out-of-Sample ForwardJan-Feb 2026
5 Unseen FiringsUnseen Return
+3.4%
Forward Win Rate
66.7%
Stability Metric
91%
Degradation Delta
-4.7%
Unseen Forward Evaluation (E0 → E4)Forward Sharpe 1.92
Temporal isolation verified: Zero look-ahead biasDegradation within standard ±12% envelope