Developing a testable strategy

Specify the experiment

Define the instruments, data source, timeframe, trading window, entry condition, exit condition, position-sizing rule and maximum portfolio risk before testing. A rule such as price looks strong is not reproducible. Freeze a version of the rules and record every subsequent change.

Control overfitting

Trying many parameter combinations and keeping the winner selects noise as well as signal. Separate development, validation and held-out samples. Do not repeatedly inspect and tune against the held-out set. Report the number of variants explored, transaction assumptions and performance across different conditions.

Evaluate distribution, not one score

Compare win rate with average wins and losses, drawdown, sample size and exposure duration. A high win rate can coexist with large losses. LEGION realized drawdown uses recorded close order, not a reconstructed intratrade equity curve. No metric converts an illustrative DEMO result into evidence of investable returns.

Practice exercise

Write a versioned rule set with an explicit no-trade condition, stop rule and held-out validation plan. Keep all experiments virtual.

Open trading terminal (requires a customer account). Compare illustrative indicators.