Validation & Research

GoldMind validation tooling is designed to challenge strategy evidence, not manufacture an attractive equity curve. Research results stay separate from live and DEMO journals and do not authorize execution.

Validation workflow

1. Historical replay — run the same market-intelligence, decision, and risk concepts against available historical data.
2. In-sample / out-of-sample — hold later completed trades out and report them separately.
3. Walk-forward — test chronological windows, including purged / embargoed boundaries where supported.
4. Stress testing — perturb spread, slippage, costs, and skipped-trade assumptions.
5. Monte Carlo — resample trade sequences as a statistical robustness check, not a forecast.
6. Readiness decision — record deterministic evidence labels and overfit warnings.

What readiness does not mean

• ▸A readiness label does not unlock Auto or REAL trading.
• ▸One profitable run is not robustness; OOS, walk-forward, sample depth, and stress behavior matter.
• ▸Monte Carlo output is a sequence simulation, not a prediction of future returns.
• ▸Canonical symbol profile presence is not equal validation or a public performance claim.
• ▸Scalping requires dedicated evidence and must not inherit Normal-style readiness assumptions.

AI and learning boundary

AI may explain journal or research context, but it does not alter backtest results, promote parameters, rewrite live strategy logic, or bypass deterministic readiness and safety gates.