Confidence Grades Are Not Evaluation Permissions
A signal label cannot replace loss-rule and contract checks.
What a Grade Might Mean
A letter may describe a model bucket or an author's judgement. Ask for the versioned definition and available calibration evidence. Do not assume it is a probability of profit.
What It Cannot Authorise
A grade does not establish permitted external signals, appropriate position size or available headroom. Even a valid signal receipt cannot demonstrate that an evaluation account remained compliant.
Primary Sources
A label belongs to its original version
Keep the supplied grade attached to its original instruction and preserve later changes with their effective version. A changed label should not silently become the grade at an earlier account decision. The original discussion remains in place; the additions concern the relationship between instruction identity and evaluation accounting.
A grade does not provide headroom
Loss headroom requires the current monitored measure, active reference and comparison convention. It cannot be read from a letter alone. The replay takes those numerical fields explicitly and issues no allocation recommendation. A genuine account still needs its own applicable policy and observation path.
Compatibility comes from the programme terms
Holding, automation, copying and external-signal permissions belong to the intended account and phase. A readable signal or an impressive historical result is a different source object. Preserve both without using one as the other's permission. No original personal or provider statement is rewritten in this pass.
Keep entry and native exit identity frozen
An evaluation target does not silently change how an entry model is trained or ranked, and a reporting reset is not a daily model stop manager. Use the authorised entry identity and predeclared native exit when assessing implementation. The new worksheets are accounting exercises, not an exit tournament or production strategy promotion.
An invented grade-versus-permission case
Consider a fictional original instruction with a recorded grade and a native holding interval. A programme permission window can conflict with that interval even if the grade is clearly defined. The review should preserve the original version and the programme source rather than silently shorten the historical hold after seeing the conflict.
This case does not report behaviour by a named provider or question any original personal claim. It illustrates separate objects: instruction classification, economic quantity, account headroom and permission. A letter alone cannot provide the latter fields. The new tools therefore accept explicit account and rule inputs, retain null approval and issue no position changes. A real compatibility conclusion needs the actual source relationship and account terms.
Keep process evidence distinct from probability
A documented grade definition and a repeatable process can support interpretation without becoming a probability of an evaluation pass. Software fixtures test the selected calculation; they are not independent participant or strategy-validation samples. A favourable example also does not retroactively authorise an ignored account restriction.
The evaluation objective should not rank entry candidates by exit-adjusted outcomes or reopen a choice of exits after the result is known. Preserve the authorised entry identity and the selected native exit in any genuine implementation review. The added material is a consumer accounting framework, not a new model experiment or production promotion. It complements the original grade discussion while keeping source classes and administrative questions separate.