Short-term validation needs independent session evidence
Keep dependent observations, inspected variants, prior rules and complete master requirements separate from software test counts.
More rows do not automatically mean more independence
Observations within one session can share information, exposures, events or an already inspected path. A count of chart bars is not automatically a count of independent economic decisions. State the actual study's unit, eligible population and evaluation contract before interpreting a statistic. The desk gathered no genuine validation sample and calculates no confidence interval here. Its software fixtures and nine setup branches are deliberately authored. They cannot be added to market observations as if each were another independently verified strategy outcome or customer experience.
Retain all inspected alternatives
Save candidate variants, inspected sessions, data revisions, exclusions and selection decisions. A final attractive rule cannot erase the trials that produced it. Entry models rank on raw causal boundary recovery, with identity frozen before one selected exit. Exit-adjusted returns, profit factor or win share may assess that object but do not define its entry ranking. Do not reopen an exit tournament after validation or forward access. The workbook specifies no actual entry calibration, economic exit experiment or best-performing strategy, so its illustrations cannot satisfy those research obligations.
Completion is the full factory and master
Actual instrument research means the complete canonical process through its independently verified master unless a narrower diagnostic was explicitly requested. A missing asset integration is part of that task, not permission to erase identity guards or route everything through equities. Compatible verified artifacts can be reused without weakening periods, warmup, fill policy or resource bounds. A verified no-model master can be completed evidence; a component screen, adapter plan or missing data cannot. This publication performs no such instrument task and therefore makes no new market-level success or failure claim.
Separate deployment from research labels
Code existence, grade maps and research LIVE names cannot prove deployed subscriber state. Registry, gate, deployed SHA, subscriber route and certification remain independent evidence objects. Production work belongs in managed cloud services with durable receipts and personal computers off. The static page creates no model promotion, additional paid capacity or account controller. Its Decimal and browser passes verify declared equations and interface behaviour only. The invented register below is a transparency example, not a completed experiment, empirical market rejection or independently audited financial result.
Fictional short-horizon evidence case
A fictional register has a rule definition, a component software pass and an unverified economic master. The correct completion state is incomplete research. No actual session sample or strategy was run. A genuine no-model result would need its own completed independent master, not a zero from one component test.
| Object | Record | Bounded finding |
|---|---|---|
| Rule | Defined fiction | Not calibration |
| Component | Software check | Not financial sample |
| Master | Not verified | Task incomplete |

Primary references and scope
Source review: 14 October 2026. Official mechanism descriptions are not this desk's paid-account test, independently captured fills or vendor-return audit.