The engineering approach
Inside the decision engine.
Rules, order handling and recovery need separate tests. Here is how GeoFinX makes each one easier to inspect.
Make the rule unambiguous
A strategy can read completed candles or respond to updates while a candle is forming. The specification should pick one behaviour, explain how to align timeframes and state what to do with missing or stale data.
If an H1 indicator provides context for an M15 decision, a closed-candle design uses the last completed H1 value. It can log the bar timestamp so the decision is traceable. This is an engineering example, not a trading recommendation.
Separate signals from execution
A valid signal does not always mean an order can be placed. Keep the signal, position sizing and order handling separate. That makes it possible to test the effect of symbol settings, minimum lot size, spreads or broker rejections without changing the original strategy rule.
- State
- Track what the software believes is open and reconcile it with platform state when appropriate.
- Interruption
- Test a restart with an existing position and check that a previously processed signal cannot be acted on twice.
- Diagnostic trail
- Record the input, the decision and the reason for any rejection in a form that can be reviewed.
Show the assumptions behind a test
Historical simulations depend on the dataset, spreads, fees, slippage and the fill model. Record those choices alongside the settings. When possible, use a demo or paper environment to check execution and alerts closer to real time.

A passing test shows that specified behaviour was observed under the test conditions. It does not predict returns or remove the risk of loss.
