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

Research brief and reading order

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

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.

Hypothesis before optimisation

Describe the cause, expected effect and observable alternative explanation. Decide which experiments can reject the idea before searching for parameters. Record every material research choice, including unsuccessful runs; the best final curve is not the whole research population.

Evidence layers stay separate

Unit tests, historical simulation, out-of-sample research, paper execution and actual account records answer different questions. Software correctness does not establish an independent trading sample. A simulation can be useful while still excluding queues, impact, gaps or unavailable historical features.

Reproducibility before promotion

Keep a run manifest with code, data, environment and policies. Require a second reconstruction of the chosen candidate and a documented operational review before considering real deployment. The examples on this site do not certify a model or authorise live trading.

Publishing boundary

Provider descriptions are adjacent-tool assessments based on public documentation, not independently audited subscriptions. The owner states legal independent ownership; commercial arrangements have not been independently reviewed. The publication does not supply an automatic winner or a fabricated research return.

Sources and scope

A reading path with a completed object

Start with the replay workbook if the question concerns available observations, an order intention and a simulated fill. It includes the fixed source, settings, complete synthetic bars, decisions, transactions and checks. Start with the original rule sheet if those decisions are still unwritten; its filled teaching example gives a concrete draft without making a financial claim. The release workbench concerns other operational sources. These are distinct paths, not a ladder on which a good-looking software output automatically becomes a live or economically validated system.

Follow the missing source

If historical availability is unknown, use the data contract and synthetic-versus-real-history investigation. If an exit is ambiguous inside a bar, use the engine contract and intrabar guide. If a period has already been inspected, retain its exposed status rather than resetting it with a new manifest. If an actual account restart is being considered, use the existing recovery material and authoritative broker/cloud records, not the synthetic fill table. The example can finish its narrow software task while leaving every real-market question unresolved.

Read cash and identity together

The six simulated positions net negative 1.20 after 0.20 of total fixed fee per position. Original input and source hashes stay attached to the result. An independent Decimal check verifies the amount within the declared numerical tolerance while preserving the engine's actual floating output. The checksum tool verifies file equality only. A documented source identity cannot prove a causal market mechanism, untouched financial evaluation or subscriber route. This reading brief therefore points to complete records and explicit boundaries, rather than proposing a trading action, forecast or new production owner.

Completed synthetic references