Skip to system research
Build A Trading SystemRules / Evidence / Replay

Trading system rule sheet

Write the hypothesis, rules, data and rejection conditions before any test.

Original edition: 8 October 2026. Source limits appear below. Expanded edition: 13 October 2026.

Educational research design and invented arithmetic examples. No strategy returns, broker execution or production certification are supplied.

Turn an idea into a testable specification. Define exactly what triggers a trade, what invalidates it, how much it can lose, and what evidence would make you stop using it.

Free, no sign-in. Your entries stay in this browser; export them before closing the page. This worksheet is not a backtest or investment advice.

1. State the hypothesis

A useful hypothesis says what should happen, where, and why. Leave marketing words out.

2. Write executable trade rules

A second person should reach the same decision from the same timestamped data.

3. Put the loss limit in writing

A conviction label is not a position-sizing formula.

4. Predeclare the test

Choose the data and split before seeing the result. Keep revisions outside the holdout.

5. Define the kill switch

A live system needs a rule for stopping, not just a rule for starting.

Readiness check

Fill in the sheet, then check it for missing decisions. A complete form is a specification, not proof of profitability.

A complete draft can state its limitations

The filled teaching rule sheet writes every required decision but explicitly states that the instrument is synthetic, the hypothesis concerns reproducible mechanics only, and there is no untouched economic validation interval. The original form's completion check means those choices are written, not that they are financially adequate. The example can be loaded, edited, exported or reset locally. Text import is not executable code and does not request an account credential. Its fixed next-open convention belongs to the tutorial; canonical observation-window fills and proprietary entry/exit policies remain unchanged.

Source identity travels with a later edit

A changed field produces a different reader draft rather than silently changing the fixed public replay. Matching-format JSON import must use only known fields and respect their length bounds; invalid records leave existing input intact. The original export format and print function remain. The new checksum tool addresses file identity, while the release workbench addresses operational evidence. Neither is a substitute for an actual market-data contract, complete candidate history or independently verified master. A local saved specification does not create a cloud production dependency or an authority to execute orders.

A filled record and an unfinished financial question

The sample names causal 3/5 means, one long unit, three completed holding bars, fixed per-side cost and input hashes. It declares that the bars were invented and no economic premise was tested. That is a useful finished software-specification exercise even though real instrument eligibility, capital policy, financing, calibration and deployment are absent. A real builder should retain the appropriate missing sources instead of replacing the synthetic label with a ticker and calling the same packet validated. The static publication provides a teaching object, not a shortcut around the authorised research and production processes.

Completed synthetic references