AR-FRM-000·PL-01 — Framework intake. Each record enters the bench as a specimen, not as a claim. Stylised brand illustration of the laboratory, not a photograph of a physical facility. The charts and figures inside the plate are simulated.

Record AR-FRM-000 · Framework Library

Twelve frameworks,
studied as families.

Each record specifies a complete trading framework rather than an entry rule: what it claims to exploit, the market states in which that claim is plausible, the geometry of a single trade, the management populations it deploys, the parameters that govern it, and the ways it fails. Twelve records now, structured so the library can grow without any of them being rewritten.

Records 12 Families 5 Status Preview Functionality Level 2 Updated 2026-07

01Why FamiliesAR-FRM-000·01

Strategies that share a claim
share a failure mode.

A pullback entry and a break-and-retest entry look different on a chart and are the same argument underneath: that a move which has already happened will continue, and that waiting for a retracement buys a tighter invalidation at the cost of a lower fill rate. They therefore break in the same conditions and leak expectancy through the same term.

Grouping by family makes that inheritance visible. It also makes the library honest about novelty: most published “systems” are a member of one of these five families with a different indicator on top, and the indicator is rarely the part that determines the outcome.

The practical consequence is diversification. Two directional frameworks running concurrently are not two bets — they are one bet on persistence, expressed twice, and the correlation shows up precisely when it is least welcome.

The five families

Family assignment follows the claim the framework makes about price, not the instrument it trades or the timeframe it trades on.

02Record IndexAR-FRM-001…012

The library.

Filter by family or by lifecycle status. Status describes where a record sits in the experiment lifecycle — Proposed through Validated — and says nothing about how profitable the framework is. A Validated record is one whose description has survived scrutiny, not one that is guaranteed to make money.

Filter by family

Filter by lifecycle status

Showing 12 of 12 records

03Cross-Record ComparisonAR-FRM-000·03

All twelve,
side by side.

The most useful column is the last one. Every framework in this library has a positive case; what separates them in practice is the specific discomfort each imposes on the person running it, and whether that discomfort is one they can hold for a year.

Twelve strategy frameworks compared by family, ideal regime, hostile regime, holding horizon and primary risk challenge
RecordFamilyIdeal regime Hostile regimeHorizonPrimary risk challenge
Figure 1 · Illustrative expectancy against the decline that produced it Simulated

All twelve records at one shared configuration.

Interactive demonstration — not a historical backtest

Reading the plane

Every record is run through the preview engine at the same configuration — one hour trigger, two-times ATR stop, half position banked at first target, four concurrent positions, ordinary friction. Horizontal position is the modelled maximum decline that configuration produces; vertical position is the illustrative net expectancy left after cost.

The point of the figure is the disagreement. A shared configuration is a reasonable default for exactly one of these frameworks and a poor fit for most of the others, which is why several sit below the zero line. A framework designed around a wide invalidation is being asked to work with a stop it would never have chosen.

Read it as a statement about calibration, not about quality. Nothing here ranks the frameworks; it demonstrates that the same settings applied across a library produce results that say more about the settings than about the strategies.

04Regime CompatibilityAR-EXP-071

Twelve frameworks
against eight market states.

A compatibility grid derived directly from each record's declared ideal and hostile regimes. It is a map of stated design intent, not a measurement: the cells say where each framework claims to belong, and the empty cells are the honest majority — states in which a framework is neither advantaged nor disadvantaged, and in which it should therefore expect to earn nothing in particular.

Figure 2 · Declared regime compatibility Design intent, not measured

Rows are framework records, columns are the eight classified market states, abbreviated: Compress, Expansion, Trend, Exhausted, Range, Vol range, Shock, Recovery. Each cell reads Ideal, Hostile or a dash for no declared relationship. Read the columns rather than the rows — a state carrying many hostile cells and few ideal ones is a state in which most of this library should be standing down.

05What a Record ContainsAR-FRM-SPEC

Thirteen fields,
the same thirteen every time.

The value of a library is comparability. A record that describes its edge beautifully and omits its failure modes cannot be set beside one that does both, so the schema is fixed and every field is filled — including the uncomfortable ones.

Two fields do most of the work. Typical failure modes is written before any result is produced, so it cannot be retrofitted to explain a disappointing outcome. Record limitations states what would have to be true for the record itself to be wrong.

Open a worked record · Pullback Continuation →

Record schema

IdentityRecord code, lifecycle status, family and holding horizon. The status is a process state, never a quality judgement.
Core edgeWhat the framework claims to exploit, stated precisely enough that it could be shown to be false.
Ideal regimesThe market states in which the claim is plausible. Declared in advance, from the eight-state classification.
Hostile regimesThe states in which the same logic is structurally disadvantaged and should not be deployed at full size.
Market universeWhich instruments the logic is permitted to touch, and implicitly which liquidity conditions it assumes.
Entry architectureTrigger, permitted chase, confirmation requirement and fill assumption.
Stop architectureWhere the idea is proven wrong, expressed as a relationship to volatility rather than as a fixed distance.
Exit architectureTargets, partial policy, trailing activation and giveback tolerance — the block that decides tail participation.
Branch possibilitiesWhich management populations the framework can deploy, from the four in the branch architecture record.
Relevant parametersThe subset of the 159 engine controls that materially move this framework's behaviour.
Planned diagnosticsThe measurements declared in advance as the ones that would reveal deterioration in this specific framework.
MARS compatibilityHow much a governance layer changes this framework's behaviour, and which layer of authority binds first.
Failure modes and limitsThe specific ways practitioners break this framework, and what would have to be true for the record to be wrong.

06MethodologyAR-FRM-000·06

How these records are written.

Four sources feed a record, and they carry very different weight. Knowing which parts are authored judgement and which are model output is the difference between using this library and being misled by it.

SOURCE 01 · STRUCTURAL The public trading literature Every framework here is widely described in public material. The library contributes structure rather than discovery: it takes a commonly discussed approach and forces it into the seventeen-component specification until every component has an answer. Weight: high for the description · none for any performance claim
SOURCE 02 · AUTHORED Regime assignment Ideal and hostile regimes are assigned by reasoning about the mechanism: a framework that depends on continuation is disadvantaged where drift is absent. This is a defensible argument, not a classification produced by measurement, and the compatibility grid inherits every weakness of that judgement. Weight: moderate · would be revised by regime-partitioned evidence
SOURCE 03 · MODELLED The preview engine archetype Each record maps to an engine archetype carrying a baseline edge, a tail multiple, a preferred volatility band, a design ATR coefficient and a shock probability. Every number on the framework pages comes from running that archetype through the same deterministic model used everywhere else on this site. Weight: directional only · magnitudes are illustrative
SOURCE 04 · OBSERVED Practitioner failure patterns The failure-mode field records how people actually break each framework — widening a chase distance after missed fills, reading a high hit rate as low risk, drawing structural levels after the move. These are the most useful lines in the library and the least verifiable. Weight: high for practice · unquantified by design

What is deliberately absent

No record carries a win rate, a profit factor, an expectancy figure presented as a property of the framework, or any statement that one framework outperforms another. Those numbers would require market data and out-of-sample separation, and neither exists in this phase.

07LimitationsAR-FRM-000·07

Where this library
is weakest.

Five specific problems, listed because a reader who does not know about them will overread the comparison table and the compatibility grid.

Family boundaries are not clean

Break and Retest is grouped as Directional because it enters with the move, yet its execution problem — a fill that may never arrive — is closer to the Structural family. Failed Breakout is filed as Reversal but is traded as a structural event by most of its practitioners. The taxonomy is a lens, not a fact about markets.

Compatibility is declared, not measured

Figure 2 shows what each record claims about itself. A framework asserting that it suits Structured Trend has not been shown to suit Structured Trend; it has been argued to. The archive already contains a failed attempt to classify these states from price alone, which is a fair indication of how hard the measured version is.

Twelve records is a small library

The planned library is substantially larger, and the current twelve are chosen for coverage of the families rather than for completeness within them. Whole approaches — statistical arbitrage, options structures, carry, event-driven trading — have no record here at all.

Implementation variance swamps framework differences

Two traders running “pullback continuation” may share nothing except the phrase. The record specifies structure, and most of the outcome dispersion in practice comes from choices inside that structure — which is precisely why the parameter laboratory exists alongside the library rather than beneath it.

Nothing here indicates what you should trade

The library is a comparative description of approaches. It contains no recommendation, no ranking by expected profitability and no route to one. A record that suits your temperament and schedule is a better starting point than a record that scores well in a model that reads no market data.

A framework is a starting point.
The parameters decide the outcome.

Two operators can select the same record from this library and end with systems that behave nothing alike — one banking early with a stable curve and no tail, the other holding through giveback for the outcomes that pay for everything else. That difference is set in the parameter laboratory, not in the choice of framework.

Calibrate a framework Branch architecture Regime taxonomy Laboratory overview

Twelve records · five families · every output a labelled model result