Confirmed bars, no lookahead
Lifecycle changes are evaluated from information available at the decision timestamp. A live, unfinished bar may be useful context, but it cannot be quietly promoted into confirmed evidence after the fact.
Methodology and trust
Evidence before narrative.
Good review starts before the outcome. Lumi preserves the setup state, the first rule holding it back, and what was knowable at the time—then keeps the versions, data quality, and correction history attached.
Private owner pilotPlanned
Evidence controls
Build a record you can explain.
Every conclusion should be traceable to the version and data that produced it.
Versioned inputs and engine identity
An evidence record should identify the engine version, configuration version, setup family, and inputs that produced its state. A later rule change belongs to a new cohort, not an edited past.
Freshness and data-quality gates
Missing, stale, delayed, or internally inconsistent data should block a readiness conclusion. Freshness is evidence about reliability, not a decorative timestamp.
Correction lineage
Corrections create a new lineage linked to the original record. They do not silently rewrite what the trader saw or what the engine concluded at the earlier time.
The first failed gate
Turn “not ready” into a useful next condition.
The first rule preventing a setup from becoming ready, shown with the condition that would need to change. It is a diagnostic state—not a prediction that the setup will eventually trigger or work.
LifecycleWaiting
First failed gate15m structure not confirmed
What changes nextConfirmed close above the review level
Illustrative methodology example · not current market dataHonest review
Keep uncertainty honest—and useful.
Correct WAIT and BLOCKED decisions count.
Ambiguous fills stay ambiguous
When partial fills, option legs, timestamps, or position intent cannot be reconciled confidently, Lumi should preserve the uncertainty instead of inventing a clean story.
Trading frictions belong in review
Spread, slippage, alert-to-decision delay, order-to-fill delay, fees, and other transaction costs can materially change an outcome. Review should not assume an idealized fill.
No-trades remain visible
Waiting, blocked, skipped, and invalidated candidates remain in the denominator. Correct restraint is part of process evidence, not missing data to discard.
No winner gallery
Lumi does not use cherry-picked outcomes, isolated win rates, or post-factum screenshots as proof of profitability. Any future performance study must state its sample, costs, assumptions, and limitations.
Four separate questions
Outcome is not the same thing as decision quality.
A sound review keeps setup, decision, execution, and outcome quality distinct.
- 01
Setup quality
Did the market-derived setup satisfy its defined rules?
- 02
Decision quality
Was the plan reasonable given what was knowable before the result?
- 03
Execution quality
How did the actual fill compare with the recorded plan and available liquidity?
- 04
Outcome
What happened afterward, including costs and path—not whether hindsight liked the decision?
“Every conclusion should be traceable to the version and data that produced it.”
The current Lumi Core research product remains private and is not for sale. Native Evidence Cards, independent data replay, journal reconciliation, and cohort analytics shown on this site are planned concepts or illustrative previews until each capability has completed the approvals required for a public release.