Technical revision: 1 October 2026 · Educational material
A trading system is a set of rules, data assumptions and execution behavior. A checklist helps make those parts observable. No combination of features makes a system “perfect” or guarantees profitability.
Write an executable specification
Describe the symbol universe, timeframe, signal calculation, information available at decision time, entry policy, exit policy and account constraints. “Buy when momentum is strong” is not precise enough. State the exact inputs, comparisons and bar shifts.
For example: evaluate a moving-average crossing between bars 2 and 1 on the chart timeframe, once a new bar begins. Skip any signal while that symbol already has exposure. Exit through the original stop or target. The downloadable demo implements this narrow specification, without claiming an edge.
Separate five responsibilities
- Data: acquire rates and indicator values, verify counts and timing.
- Signal: calculate a deterministic decision from verified observations.
- Risk: estimate volume using the final rounded stop and account-currency loss.
- Execution: apply symbol constraints, submit requests and reconcile server state.
- Observation: log inputs, decisions, rejection reasons and results.
Keeping these responsibilities distinct makes a fault easier to reproduce. A correct signal can coexist with a rejected order; logging should preserve that distinction.
Use MQL5 indicator handles correctly
// Initialization: creates a handle, not an ATR value.
int atr_handle=iATR(_Symbol,PERIOD_M15,20);
// Later, after history/calculation readiness checks:
double atr[1];
if(CopyBuffer(atr_handle,0,1,1,atr)!=1) return;
// atr[0] belongs to the last closed M15 bar.
// Release atr_handle during deinitialization.This is an API illustration, not a standalone EA. A production read also checks valid values and maps the higher timeframe’s availability to the decision timestamp. Mixing MQL4 call signatures into MQL5 is a common source of incorrect examples.
Treat broker specifications as inputs
Read volume minimum, maximum and step; tick size; stops and freeze levels; and supported execution/filling policies. A fixed IOC choice may be unsupported. The included EA selects filling by symbol and checks server return codes.
Margin availability and an estimated loss budget answer different questions. Passing a margin check does not prove that the planned risk is acceptable. Nor does sending a stop guarantee the price at which it will execute.
Design validation before optimization
Freeze a baseline rule and reserve a chronological holdout. Test correctness with known small cases before using the tester to compare results. Record costs, data source, tick model, broker symbol and terminal build. A large optimization search raises the chance of selecting a result fitted to noise.
Timers and repeatable code alone do not guarantee identical trading outcomes. Data, quote sequences, platform settings and broker state also affect a run. The validation checklist provides a reproducible record.
Choose the next concrete task
Begin with the learning route. Use the non-trading logger, then the closed-bar EA and diagnostics. Once behavior is understood, extend one module at a time and compare against the frozen baseline.
References: iATR, filling by symbol.
