What is out-of-sample testing on gold?

Testing the rule set on a gold period that was never used to develop or tune it, ideally in a different volatility regime, so you learn whether the pattern generalises rather than whether it was fitted.

How do I test whether a gold model is overfit?

Perturb the parameters. Shift the stop by two dollars an ounce and the session window by thirty minutes. Robust edges degrade gradually; overfit models collapse.

When is a gold model ready to go live?

When it has cleared out-of-sample and cost-realism checks, completed 100-plus demo trades across CPI, NFP and FOMC, and met go-live criteria that were written down before testing began.

GOLD MODEL VALIDATION

Validating a gold model runs in sequence: in-sample description, out-of-sample testing in a different volatility regime, cost realism with real XAUUSD spreads and slippage, perturbation checks, forward demo and small live.

Go-live criteria — minimum sample, minimum expectancy after costs, maximum acceptable drawdown in dollars per ounce and the size-reduction trigger — must be written before testing starts, not after results appear.