Technical revision: 1 October 2026 · Educational material
The useful test of a trading-system architecture is whether you can explain a decision and reproduce a failure. Start with a small, traceable pipeline. Add infrastructure when a concrete requirement justifies it.
Make the decision pipeline explicit
Verified observations
→ signal evaluation
→ exposure and risk checks
→ broker-valid request
→ server response and transaction reconciliation
→ updated state and audit logThis diagram is conceptual pseudocode. It does not supply a deployable execution engine. Each arrow has a failure condition: missing history, stale quotes, invalid volume, insufficient margin, a rejected request or an unexpected server state.
Keep observation time separate from chart time
Record when information became available. A higher-timeframe candle, revised indicator or confirmed swing marker can look clear on a completed chart while having been unavailable at the original decision moment.
For an initial MT5 implementation, use a closed-bar logger and deliberately delay processing until all requested values are available. Timestamp the symbol, timeframe and copied shifts. See the CopyBuffer lesson for a frequent ordering fault.
Track intent, request and confirmed state
A buy decision is intent. An accepted request is a separate event. A filled deal and current position are server state. Do not infer the final position from an order ticket alone. Netting and hedging account modes also change how positions are represented.
A production design should reconcile transactions, recover its state after restart and avoid duplicate submissions when a response is uncertain. Before retrying, inspect current orders, deals and positions. Repeatedly issuing the same request can create unintended exposure.
Use explicit limits and failure policies
Document estimated per-trade loss, portfolio exposure limits, margin checks and the conditions under which new entries stop. Decide whether a rejected entry is skipped or retried, and how often. Decide what happens if an existing stop modification fails.
Separate a price move to the stop from a guaranteed monetary cap. Spread, gaps, commissions, swaps and changing conversion rates can alter realized loss. Any stress test should state which of these effects it represents.
Log enough to answer “why?”
| Record | Purpose |
|---|---|
| Code version, inputs, terminal build | Identify the implementation being evaluated. |
| Observation timestamps and values | Reconstruct the information available. |
| Risk budget, rounded volume and prices | Explain the requested exposure. |
| Method result, return code, deal/position state | Separate a submission from execution. |
| Skip/retry reason | Explain missing trades without guessing. |
Keep research and deployment evidence distinct
Compile successfully, test behavior, run historical validation and observe demo execution as separate stages. A profitable historical report does not show that recovery logic works. A clean compilation does not establish trading profitability.
Use the MT5 validation checklist to freeze assumptions and retain reports. The Start Here route supplies small examples for the first stages; it does not claim to be a complete production trading service.
