Examples are hypothetical. This lesson is not a live trade recommendation or a promise of returns.
Start with a question somebody else can repeat
An attractive result is difficult to compare when its build or inputs are missing. This lesson creates an understandable baseline in a testing environment. Profit is not the pass criterion.
A single test before optimization
A single test uses specified inputs; optimization explores input combinations. Preserve the first experiment before changing assumptions. Simultaneously changing the rule, costs and interval makes attribution difficult.
Keep four items
Name the experiment folder with a date, identifier and build. Save its question, exported settings, complete report and limitations. Preserve failures alongside favorable results so the next comparison uses the full record.
Record assumptions and investigate errors
Write the version, symbol, broker, dates, model, costs and execution-delay setting. A modeling option does not reproduce every future condition. If no trades appear, inspect rules, inputs, data and journal messages rather than immediately declaring the program correct or broken.
Repeat and explain differences
Read trade count, drawdown, net result and errors together. Repeat without an intentional change and save the second report separately. Explain differences and any randomized component before interpreting them as improvement.
Assignment and completion criteria
Prepare a laboratory notebook with question, build, symbol, broker, dates, data model, costs, delay, inputs, report, journal and limitations. The pass criterion is a repeatable, understandable procedure. Use authorized software only; never enable live trading or unsafe permissions merely to complete an educational exercise.
Next topic: interpreting a test report. This page contains lesson two; it does not claim every course video is complete.