BDS / Intraday usability
Paper, Replay and Live Day-Trading Evidence: Keep the Labels
Testing the software mechanism, reconstructing a chart and observing an actual account are different experiments.
Name the experiment
A replay applies a declared rule to retained information. A paper environment applies its simulation assumptions. A live account provides actual accepted and executed operations. A test of the interface can be useful without being any of these performance records.
State the environment and task before reporting a result. A mock webhook response or paper order should not be called a funded broker-fill test.
Preserve the information boundary
A replay must use the input and instruction versions available under its declared convention. A later chart can contain revised or completed information. Keep changes as versions instead of silently substituting the more favorable display.
Internal hypothetical research fills belong to their frozen policy. Actual subscriber fills belong to the account record. Combining the best value from each produces an unsupported evidence class.
Worked paper example
Hypothetical case: a paper environment accepts a rule and produces a simulated close. The supported observation concerns that environment under its settings. It does not establish the broker price, rejected-order behavior or funded costs a subscriber would experience.
The experiment can still reveal a configuration error or incomplete rule. Software usefulness does not require pretending the result is a completed independent strategy test.
Compare only the declared objects
Keep original instruction results, simulated operations and actual account operations in separate records. Each needs its own population and assumptions. A single combined win rate can hide which outcomes were hypothetical.
A live account sample is not automatically an independent audit either. Identify whose account it is, what materials were available and what period the claim covers.
Record limitations and costs
A simulation may model costs or leave them absent. Label that choice. Actual account arithmetic needs actual operations and the relevant costs, funding and open exposure. Unknown fields should not default to favorable values.
A clean chronological split does not validate every trading assumption. Tests of code correctness and UI operation are not independent performance samples.
Publish an honest class
Use explicit labels such as hypothetical teaching example, paper workflow, public-document research or reader-supplied unverified record. Stronger claims require their own artifacts and scope.
No paid bot, private script or funded provider account was run for this guide. The worksheet makes evidence classes inspectable rather than certifying them.
A passing software test is not a trading sample
A form export, correct ratio or successful mock request shows software behavior under its test. It does not independently validate the strategy or provide a subscriber result. Keep release QA receipts separate from performance evidence, even when both use timestamps and pass/fail labels.
Respect frozen historical identity
Older research receipts retain their declared fill and information policy. Do not reinterpret them as actual broker execution or silently change the observation window. New exercises can be created with new identities, but comparisons should describe the differences rather than erase them.
Preserve untested operating failures
An isolated paper environment may not reproduce rejected orders, delayed updates or account-specific constraints. Name those untested boundaries. A simulated mechanism can be useful without being marketed as proof that the funded execution chain will behave identically.
Evidence Packet
| Evidence object | Record to retain |
|---|---|
| Environment | Replay, paper or actual account |
| Convention | Input and fill policy frozen |
| Claim scope | What the experiment can and cannot establish |
Related Session Questions
Primary References
The sources support documented mechanisms and source definitions. Our procedure and teaching examples are not completed paid-feed, latency or account tests.