Framework

Qualification Framework

A reusable framework for building qualification gates that determine when ideas earn the right to execute.

v0.1·working·Updated July 2026

Overview

Most systems treat the boundary between idea and action as a single gate — a threshold, a score, an approval. This works when ideas are scarce and expensive. When ideas are abundant and cheap, a single gate cannot distinguish between ideas that happen to look good and ideas that have genuinely earned execution through accumulated evidence.

The Qualification Framework provides a reusable architecture for building multi-dimensional qualification gates across any domain where decisions must earn execution. It is not specific to trading, finance, or any single application. The framework defines how evidence accumulates, how gates are structured, how confidence decays, and how qualification status evolves over time.

The framework is conjunctive by design — all gates must pass simultaneously for qualification to hold. This is a deliberate architectural choice. Disjunctive gates (any gate passes) allow weak dimensions to hide behind strong ones. Conjunctive gates force every dimension to independently justify itself, producing higher-integrity qualification decisions at the cost of higher false-negative rates.

Design Principles

These are governing constraints, not guidelines. Every architectural decision in the framework must be traceable to one of these principles.

01

Conjunctive Gates

All qualification dimensions must pass simultaneously. No single strong dimension compensates for a weak one.

A strategy with excellent returns but no regime proof is not qualified — it is untested in a critical dimension. Conjunctive evaluation prevents the halo effect where one impressive metric masks fundamental weaknesses.

02

Continuous Re-Qualification

Qualification is not a credential. It is a continuously evaluated state that must be re-earned as evidence accumulates and conditions change.

A strategy qualified six months ago under different market conditions may no longer deserve that status. Without continuous re-qualification, the system accumulates stale trust that does not reflect current reality.

03

Evidence Over Opinion

Qualification gates accept structured evidence, never confidence, narrative, or subjective assessment.

Opinions are cheap and abundant — especially from AI systems that can generate compelling narratives for any position. Evidence is costly and scarce. The qualification boundary must be crossed with evidence, which is the only currency that resists inflation.

04

Fail-Closed Default

When evidence is insufficient, ambiguous, or missing, the default state is unqualified. Qualification must be actively earned, never assumed.

In domains where the cost of unqualified execution exceeds the cost of delayed execution, the safe default is to withhold trust. This principle prevents the system from drifting toward permissiveness as volume increases.

System Map

Four layers from raw observation to managed qualification status. Evidence flows upward from the observation layer. Assessment consumes evidence and produces gate evaluations. Gate evaluation determines status. Status management handles lifecycle transitions, re-qualification, and confidence decay.

1
Evidence Accumulation3 components
Validation EvidenceSimulation EvidenceEnvironmental Evidence
2
Multi-Dimensional Assessment3 components
Cohort AnalysisRegime ProofDuration and Volume Gates
3
Gate Evaluation2 components
Conjunctive Gate EngineConditional Qualification
4
Status Management2 components
Status Lifecycle ManagerConfidence Decay
Intelligence flows down|Evidence flows up

Layers & Components

Each layer has distinct responsibilities, inputs, outputs, and ownership. No component spans multiple layers.

1

Evidence Accumulation

Gathers raw observations from validation, simulation, and execution into structured evidence that qualification gates can consume.

Validation Evidence

Captures historical performance evidence from backtesting

In: Strategy spec, Historical price data, Regime catalogOut: IS/OOS performance metrics, Confidence intervals, Regime-conditional scoresOwns: Backtest result storage, scoring formulas, fee-efficiency multiplier

Simulation Evidence

Captures forward-looking behavioral evidence from paper trading

In: Live market data, Strategy execution stateOut: Paper trading track record, Behavioral observations, Divergence metricsOwns: Paper account state, position history, equity curves

Environmental Evidence

Captures regime context and environmental conditions relevant to qualification

In: Market data, Regime classifier outputOut: Regime labels, Regime durations, Transition historyOwns: Regime classification, market context snapshots
2

Multi-Dimensional Assessment

Evaluates evidence across independent dimensions. Each dimension produces a pass/fail judgment with supporting metrics.

Cohort Analysis

Compares backtest and paper performance to detect divergence

In: Backtest results, Paper trading resultsOut: Divergence score, Cohort comparison metricsOwns: Divergence thresholds, cohort definition logic

Regime Proof

Evaluates whether a strategy has demonstrated performance in its claimed regime

In: Paper trading outcomes, Regime history, Trade timestampsOut: Regime proof status, Per-regime trade counts and returnsOwns: Regime proof criteria, minimum trade counts per regime

Duration and Volume Gates

Ensures sufficient observation before qualification decisions

In: Paper account age, Trade count, Position countOut: Duration gate pass/fail, Volume gate pass/failOwns: Minimum duration thresholds, minimum trade count requirements
3

Gate Evaluation

Applies conjunctive logic across all assessment dimensions. Produces qualification status with full evidence trail.

Conjunctive Gate Engine

Evaluates all dimensions simultaneously — all must pass for qualification

In: Dimension assessments (all)Out: Qualification decision with evidence trailOwns: Gate logic, conjunctive evaluation rules

Conditional Qualification

Handles partial qualification when some but not all dimensions pass

In: Dimension assessments, Missing evidence dimensionsOut: Conditional qualification status with conditions for full qualificationOwns: Conditional status definitions, remaining evidence requirements
4

Status Management

Manages the lifecycle of qualification status — transitions, re-qualification, confidence decay, and disqualification.

Status Lifecycle Manager

Governs transitions between qualification states based on evidence changes

In: Gate evaluation results, Current status, Time since last evaluationOut: Updated qualification status, Transition eventsOwns: Status state machine, transition rules, event emission

Confidence Decay

Reduces qualification confidence when fresh evidence is not accumulating

In: Time since last trade, Time since last evaluation, Environmental changesOut: Decayed confidence score, Re-qualification triggersOwns: Decay functions, staleness thresholds

States & Transitions

A decision moves through these states as it accumulates evidence. Each transition is gated — there are no free promotions.

Unqualified
Conditionally Qualified
Qualified
Disqualified

Unqualified

Default state. Insufficient evidence to justify execution. No gates have been evaluated or all have failed.

Entry: Initial state for all candidates, or re-entry after disqualification

Exits: Evidence accumulates, some gates pass -> Conditionally Qualified

Conditionally Qualified

Some qualification dimensions pass but not all. Evidence is accumulating in a positive direction.

Entry: Core performance gates pass, but regime proof or duration gates remain open

Exits: All gates pass -> Qualified · Core gates fail -> Disqualified · Evidence stales -> Unqualified

Qualified

All conjunctive gates pass. Strategy has earned the right to execute with real resources.

Entry: All dimensions pass simultaneously in a single evaluation

Exits: Evidence degrades -> Conditionally Qualified · Severe failure -> Disqualified

Disqualified

Active determination that the strategy should not execute. Distinct from unqualified — disqualification implies evidence against, not just absence of evidence for.

Entry: Severe divergence, catastrophic paper performance, or behavioral anomaly detected

Exits: Significant new evidence after cooling period -> Unqualified (restart)

FromToTriggerGuard
UnqualifiedConditionally QualifiedPartial gate passageAt least core performance and duration gates pass
Conditionally QualifiedQualified6h evaluation batchAll conjunctive gates pass simultaneously
QualifiedConditionally QualifiedRe-qualification evaluationOne or more non-critical gates fail, core gates still pass
QualifiedDisqualifiedSevere evidence failureCatastrophic divergence, negative regime proof, or behavioral anomaly
Conditionally QualifiedDisqualifiedCore gate failurePerformance or divergence gates fail
DisqualifiedUnqualifiedCooling period expires with new evidenceMinimum 14-day cooling period, new validation results

Boundary Rules

Boundaries define what crosses between layers and what does not. Every boundary has an explicit failure mode.

Evidence Accumulation -> Assessment

Crosses

Structured evidence records with timestamps, regime labels, and provenance

Does Not Cross

Raw market data, intermediate calculations, execution engine internals

Failure Mode

Missing evidence -> assessment dimension returns 'insufficient data', which fails the conjunctive gate.

Assessment -> Gate Evaluation

Crosses

Per-dimension pass/fail with supporting metrics and confidence levels

Does Not Cross

Assessment methodology details, threshold tuning rationale, intermediate scores

Failure Mode

Ambiguous assessment -> treated as fail (fail-closed default).

Gate Evaluation -> Status Management

Crosses

Qualification decision (qualified/conditional/disqualified) with full evidence trail

Does Not Cross

Gate engine internals, evaluation history for other strategies

Failure Mode

Gate evaluation failure -> status unchanged (conservative). No silent status changes.

Status Management -> External Systems (Allocation, UI)

Crosses

Current qualification status, conditions for full qualification, evidence summary

Does Not Cross

Internal status transition history, decay computation details, gate evaluation logs

Failure Mode

Status query failure -> external systems treat as unqualified (fail-closed propagation).

Operational Semantics

How this framework operates in practice — cadence, consumers, and staleness rules.

DimensionValueRationale
Qualification assessment cadenceEvery 6 hours (batch)Balances evidence freshness against compute cost. Qualification is not latency-sensitive — strategies trade on minute cadence, but qualification evolves over days.
Confidence decay half-life14 days without new tradesA strategy that stops trading loses qualification confidence. After 14 days of inactivity, confidence is materially degraded.
Minimum observation period7 days of paper tradingShorter periods produce unreliable behavioral characterization. 7 days provides a minimum window for behavioral signal accumulation; confidence increases with longer observation.
Minimum trade countDomain-specific (trading: 3+ trades for regime proof)Statistical significance requires volume. Three trades is the minimum for any per-regime assessment.
Re-qualification triggerRegime transition, 6h batch, or significant outcome eventQualification must respond to environmental changes, not just scheduled evaluation.
Disqualification cooling period14 days minimumPrevents oscillation between qualified and disqualified states. Forces meaningful new evidence before re-entry.

Knowledge Lineage

Related Research

See this framework implemented in Orqis

Research — Warren Labs | Orqis