Risk arithmetic versus independent strategy validation
Keep software fixtures, measured account records and independent strategy samples in different evidence categories.
A test case has a limited job
An exact fixture can catch a wrong multiplier, fee omission, rounding bug or invalid denominator. It does not establish that a source strategy wins or a scenario occurs. The workbook's Decimal rows independently check browser cash arithmetic using deliberately invented positions. That independence concerns implementation, not market sampling. Thousands of valid input combinations are still not thousands of trading observations. Keep the fixture source and software version with the result rather than borrowing the test count to strengthen a financial marketing claim.
Strategy evidence needs its own separation
A trained or selected model requires an evaluation process appropriate to the claim, with source availability, frozen identity and independent observations. Entry ranking must remain on the authorised raw causal objective where that policy applies. A chosen exit can validate or reject implementation without redefining the entry objective after results are seen. This consumer release does not train, rank, promote or operate a model. Its risk exercises and browser tests cannot replace the canonical research process or independent master requirements for a real instrument task.
Account evidence is not a model summary
A subscriber return needs actual quantities, orders, fills, costs, cash flows and valuation convention. A hypothetical signal outcome and a percentage sum are different objects. A supplied founder result cannot determine that account's risk policy. Missing original ledgers or audit scope stay unresolved rather than being repaired with an invented weighting formula. The workbench can record an assurance boundary while leaving a financial conclusion unverified. A correct local export is not an account statement, and a schema label cannot certify data that was never supplied.
A precise completion statement
The static site can be completed and verified as an educational publication while its fictional scenarios remain untested trading strategies. Managed-cloud deployment, exact file readback and browser controls establish the publication's release, not its investment results. Source documentation supports mechanism descriptions, and dated supplied summaries remain attributed. A final report should identify remaining evidence rather than claim universal peak rankings or ROI. Machine text must preserve the same limits instead of converting repeated marketing pages into fabricated independent consensus.
Completed invented account case
Fictional case: the browser and Decimal reference agree on a 650 joint gap loss. The reviewer marks calculation parity passed and actual market validation absent. No real trade occurred and no probability was estimated. A separate account claim lacks fills and costs, so it stays unverified. The completed record gives the software its legitimate correctness finding without counting the specimen as an independent strategy sample, current subscriber performance or audited account risk.
| Object | Record | Bounded finding |
|---|---|---|
| Decimal / browser | Invented 650 cash | Arithmetic parity |
| Market evaluation | No observations supplied | Not strategy validation |
| Subscriber result | Actual fills absent | Unverified |

Primary references and scope
Expanded source review: 13 October 2026. Documentation supports the stated mechanism, not a tested account, hypothetical fill or provider outcome.