AR-IDX-001·PL-01 — Every measure, in one index. Stylised brand illustration of the laboratory, not a photograph of a physical facility. The charts and figures inside the plate are simulated.

Record AR-DOC-001 · Reference

Every term, measure,
identifier and route.

The Foundry uses a number of ordinary trading words in a narrower sense than usual, and a number of measures that have no standard definition anywhere. This page fixes all of them in one place: a thirty-two term glossary, a definition and unit for every quantitative measure on the site, the notation key, the identifier scheme and the complete route map.

Glossary terms 32 Measures defined 28 Identifier prefixes 11 Routes 28 Updated 2026-07

01GlossaryAR-DOC-001·01

A to Z.

Thirty-two terms, defined as this site uses them. Where a definition here is narrower than common usage that is deliberate — drawdown, risk and validated in particular all carry looser meanings in general trading conversation, and the looser meanings are the source of a great deal of confusion downstream.

Search the glossary

Jump to a letter

Showing 32 of 32 terms

02Measure ReferenceAR-DOC-001·02

Every quantitative measure
used on this site.

Each measure with its definition, the unit it is expressed in, and where it appears. Several are specific to the Foundry and exist nowhere else — where that is the case the definition here is the definition, and it should not be assumed to match a similarly named figure in other software.

Quantitative measures with definition, unit and the pages on which each appears
MeasureDefinitionUnitWhere it appears

The eight components of the AlphaRail Preview Score

The composite is a weighted blend of eight sub-scores, each on a nought-to-one-hundred scale and each computed from a different part of the evaluation. The weights are fixed and published here so that a score can be reasoned about rather than merely reported. Several ceilings also apply: a configuration with negative net expectancy, excessive per-trade risk, a decline beyond a survivable depth, friction consuming most of the edge, or high correlation across concurrent positions is capped regardless of what the blend would otherwise produce.

The eight AlphaRail Preview Score components with weight and what each measures
ComponentWeightWhat it reads

Interactive demonstration — not a historical backtest

03NotationAR-DOC-001·03

Symbols,
and what they stand for.

Formula plates across the site use a small, consistent set of symbols. The most important is R, because expressing everything in units of initial risk is what makes a five-minute scalp and a six-week position trade comparable at all.

Percentile notation appears wherever a distribution is being described rather than a single value. A P10–P90 band is not a forecast interval and is not a confidence interval; it is a statement that eighty percent of the simulated outcomes fell between those two values under the stated assumptions.

Notation key

ROne unit of initial risk — the distance from entry to the original stop, expressed as money at risk. A trade closed at +2.4 R returned 2.4 times what it would have lost at its stop. Every execution-layer measure on this site is expressed in R, and currency amounts appear nowhere, because currency encodes position-sizing decisions that contaminate every statistic computed from them.
ρCorrelation between concurrent positions, on a scale from 0 to 1. Used to convert a nominal position count into effective independent bets: at ρ = 0.7, four concurrent positions behave closer to one and a half independent ones. The correlation ceiling control sets the maximum ρ the configuration will accept.
σStandard deviation — the dispersion of a set of outcomes about their mean. σ_R specifically means dispersion measured in R space, which is the form used in every confidence-interval and sample-adequacy calculation on the site. It is the number most often missing from a published expectancy figure.
P10 – P90Percentiles of a simulated distribution. P10 is the value below which ten percent of simulated outcomes fell; P90 the value below which ninety percent fell. A P10–P90 band therefore contains eighty percent of the modelled outcomes and says nothing about the twenty percent outside it, which is the region that matters most.
P25 / P50 / P75The quartile markers used on fan charts and Monte Carlo path fields. P50 is the median path, not the expected path and not the likely path — half the simulated outcomes were better and half worse, and no individual path is being predicted.
p₁ · p_cProbability of reaching the first target, and the conditional probability of continuing beyond it. Together they determine how much of a configuration's expectancy is available before the exit block decides how much of it is kept.
AR-XXX-000The identifier scheme. A three-or-four letter prefix denoting the record series, followed by a number that is permanent and never reused. Section plates carry a suffix after a middle dot — AR-DOC-001·03 is the third section of this record.

04Identifier IndexAR-DOC-001·04

Eleven prefixes,
and what currently exists under each.

Identifiers are permanent. A record that is superseded keeps its number and gains an Archived status; a number is never reassigned, so any external reference continues to resolve to the same work. The prefixes below are the complete set in use.

Identifier prefixes with what each denotes and the records that currently exist under it
PrefixDenotesRecords that currently exist
Figure 1 · Records currently published under each prefix Site census

Identifier counts across the eleven series.

Reading the identifier scheme

Numbers are not sequential in timeThe AR-EXP series runs from 041 to 118 with gaps. The gaps are studies that were opened and abandoned before producing anything worth recording, and the numbers stay spent so that a reference to AR-EXP-055 never silently resolves to different work.
Section plates are not recordsA plate such as AR-FRM-000·03 identifies a section of a page rather than a research record. The ·000 base number is reserved on every series for the index page of that series.
Three prefixes carry no numbered records yetAR-FAC, AR-DOC and AR-REG identify facility, documentation and register material rather than studies, so their populations grow with the site rather than with the research programme.
Additional plates in useThree further plates appear on individual pages without forming a numbered series: AR-LAB-001 for the laboratory overview, AR-MRS-001 for the MARS governance architecture and AR-MARS-GATE for the gate-state diagram. They are noted here so the identifier index is complete.

05Route IndexAR-DOC-001·05

Every route on the site.

Grouped by area, one line each. This is the complete map — there are no unlisted pages, and no route on this site exists outside this table.

06MethodologyAR-DOC-001·06

How this reference
is maintained.

Three of the four tables on this page are generated from the same data the rest of the site reads. The glossary, the identifier index and the record counts are not transcribed — they are rendered from AR.GLOSSARY, AR.RECORDS, AR.FRAMEWORKS and AR.WIZARDS at page load, which means this reference cannot fall out of step with the pages it describes.

The measure reference is authored, because unit and provenance are editorial judgements rather than data. Its definitions are checked against the engine implementation rather than against the prose on the pages, on the principle that if a page and the code disagree the page is wrong.

The route index is maintained by hand against the site's fixed route list. Every route appears exactly once, every link on this page resolves to a route on that list, and no page on the site is reachable from anywhere except a route named here.

Where a definition and a page disagree

This page is authoritative for vocabulary and units. Where a narrative page uses a term more loosely for readability, the loose usage is the error. If a measure appears on a page with a definition that contradicts the one here, treat the discrepancy as a defect and the table below as correct.

Foundry-specific measures do not travel

Risk efficiency, tail participation, capture efficiency, the survival estimate and the preview score are defined by this laboratory. Other platforms use several of those names for different quantities. A figure computed elsewhere should never be compared with one computed here without checking both definitions first.

Every measure on this site is model output

The definitions are exact. The values they take on any page of this site come from a deterministic preview engine that reads no market data, and they describe the behaviour of that model rather than of any market, account or strategy.

07LimitationsAR-DOC-001·07

What a reference
cannot fix.

Four honest limits on this page specifically, distinct from the limits stated on the pages the reference describes.

A precise definition is not a validated one

Defining tail participation exactly does not establish that it measures anything useful. Several measures here are constructions of this laboratory whose usefulness is a design claim, not a demonstrated fact, and the research archive holds no study establishing that any of them predicts anything.

Thirty-two terms is a working vocabulary, not a lexicon

The glossary covers what this site uses. Whole areas of trading vocabulary — options terminology, market microstructure, fixed-income conventions, order-type taxonomy — are absent because no page here needs them, not because they are unimportant.

Units hide assumptions

R is a clean unit only if the original stop was a genuine invalidation. A record whose stops were placed at convenient distances produces R values that are internally consistent and externally meaningless, and no amount of unit discipline downstream repairs that.

The MARS calibration is not published here

Gate boundaries, pool percentages, tier tables and workbook completion figures are part of the MARS workbook product and do not appear in this reference or anywhere else on the site. The structure and the principle are described; the numbers are not, and no combination of pages will reconstruct them.

The vocabulary is the cheap part.
Applying it is where it gets interesting.

Every measure defined on this page is exposed as a live output somewhere in the laboratory. The fastest way to understand what any of them actually mean is to move a control and watch which ones respond and which do not.

Open the parameter laboratory Research archive Learning centre Laboratory overview

32 terms · 28 measures · 11 identifier prefixes · 28 routes