Index Methodology

Protocol 7 AI publishes three proprietary indices in its daily, weekly, and monthly briefings. These indices are quantitative tools designed to synthesize publicly available market data into structured readings. They measure signal intensity and systemic conditions — taken on their own, they are not buy or sell signals and they generate no trade recommendation.

Two levels must be distinguished, and they must not be confused. At the level of the instruments: no index, score or classification published by Protocol 7 AI is a recommendation to buy, to sell or to hold. At the level of the publication: the briefings that carry and comment on those readings are general market analysis, addressed identically to all subscribers, and they are not personalised. Sections IV to VIII below set out the identity of the author, the methodology, the meaning and time horizon of the classifications used, the update frequency and the conflicts of interest.

The summaries below describe the general approach behind each index. Detailed algorithmic specifications and source code are proprietary and are not disclosed.

AlphaScout aggregates signals from multiple market categories — including risk sentiment, macroeconomic data, crypto systemic indicators, geopolitical developments, and sector narratives — into a single composite score. The score measures the magnitude of market activity, not its direction. A high score indicates elevated signal dispersion across asset classes; a low score indicates calm conditions.

The DEFCON Engine scores systemic stress on a 0–15 scale built from six independent pipelines, each contributing a weighted sub-score up to its own maximum: Liquidity (0–3), Plumbing (0–2), Energy (0–3), options Gamma (0–1), Geopolitical risk (0–3), and Bonds (0–3). The composite is the sum of the six, mapped to four alert bands — 0–3 Risk-On, 4–7 Caution, 8–11 Warning, 12–15 Crash. One override applies: a confirmed kinetic geopolitical event floors the level at Caution regardless of the other pipelines, so a live shock cannot be published as Risk-On. The specific indicators and thresholds within each pipeline are proprietary; the six categories, their weights, and the banding are disclosed so that any published score can be traced to its risk band. DEFCON is a risk-monitoring tool, not a directional indicator.

DAEMON tracks the economic health of the Bitcoin mining industry by combining miner profitability metrics with network hashrate trends. It serves as a structural health and miner supply-pressure gauge. The score reflects whether mining economics are expansionary, neutral, or under stress. It does not predict Bitcoin price direction.

The briefings published by Protocol 7 AI are general market analysis. They are addressed identically to all subscribers; they are not personalised and take no account of any recipient’s objectives, knowledge, experience or financial situation. They are neither investment advice nor a personal recommendation within the meaning of Directive 2014/65/EU (MiFID II), and they are no offer or solicitation to buy or sell any financial instrument. All investment decisions are made at the reader’s own sole risk and responsibility.

Producing and distributing general investment recommendations is an ancillary service under Article L. 321-2 4° of the French Monetary and Financial Code, the exercise of which does not require authorisation (AMF guide DOC-2013-25). Separately, Protocol 7 AI is not registered as a Conseiller en Investissements Financiers (CIF) with the AMF: CIF is not an authorisation but a registration — with ORIAS, together with membership of an approved professional association — required of those who provide personalised investment advice, which Protocol 7 AI does not provide. Protocol 7 AI provides no service on crypto-assets as defined under MiCA (EU 2023/1114). Past algorithmic performance is not indicative of future results.

Author: Cyril CORNET — Protocol 7 AI (EI) — SIRET 102 471 166 00011

For questions regarding methodology, contact [email protected].

Data sources. Every briefing is built exclusively from publicly available data: exchange and index price data, published macroeconomic releases and central bank publications, publicly available positioning, flow and derivatives data, on-chain and mining data for the crypto readings, and public news reporting retrieved through a third-party search pass. Protocol 7 AI uses no non-public information, receives no material from issuers, and holds no data feed conferring an informational advantage.

Production chain. Each edition is assembled by an automated pipeline: collection and normalisation of the public data; computation of the three proprietary indices described in Sections I to III by deterministic code; generation of the written analysis and of the classifications described in Section VI by a large language model, constrained by a fixed editorial and compliance instruction set and supplied only with the data collected at the first step; then automated consistency and compliance checks. The narrative is machine-generated and is sent as generated: there is no human review before publication. The publisher assumes full responsibility for it.

Assumptions. The analysis assumes that the public data collected is accurate and complete as published, that observed positioning and flow data are a meaningful description of the current state of the market, and that historical relationships between macroeconomic variables and asset classes remain broadly informative — an assumption that fails precisely at regime breaks. No valuation model is applied: Protocol 7 AI produces no fair value, no price target, no earnings estimate and no discounted cash flow, and nothing it publishes is a valuation of a security.

Limitations. Readings are point-in-time and may be invalidated by any event occurring after the cut-off time stated in each edition. A language model can misread a source, conflate two events, or state a relationship with more confidence than the underlying data supports; the automated checks reduce that risk without eliminating it, and no human reads the edition before it is sent. Third-party data may be revised after publication. The internal specifications of the three indices are proprietary: their categories, ranges and bands are disclosed, the thresholds inside each pipeline are not.

BULLISH / BEARISH / NEUTRAL, displayed on the Crypto and Equities cards, classify positioning observed on the day of publication. They describe the direction of the flow and positioning data read in the sources, never an expectation of future price. BULLISH: the observed positioning leans long. BEARISH: the observed positioning leans short. NEUTRAL: no dominant lean is observable, or the available readings contradict one another. DEFENSIVE is a fourth label, used on the Equities card only, when the observed positioning leans defensive rather than clearly long or short; the sectors concerned are listed on the card itself.

RISK_ON / RISK_OFF / MIXED, displayed on the market regime ribbon, classify the state of cross-asset conditions observed at the time of publication. RISK_ON: observed conditions are consistent with appetite for risk assets. RISK_OFF: observed conditions are consistent with aversion to risk assets. MIXED: the observed signals point in opposite directions, or no dominant state emerges. The regime classification describes conditions across markets; it is not a view on any individual instrument.

SIGNAL CLARITY X/10 measures how unambiguous the observed data is — 1 = a single weak reading, 10 = several independent corroborating readings. It is a property of the data, not a level of conviction of the publisher, and it carries no information about the amplitude, the timing or the probability of any market move. A high score on a BEARISH classification means the observed positioning is clearly short, nothing more.

Time horizon. Each classification is issued for the horizon of the edition that carries it: 24 hours to 1 week for the daily briefing, 1 week to 1 month for the weekly briefing, 1 month to 1 quarter for the monthly briefing. Beyond that horizon a classification is lapsed: it is not carried over, it is not renewed by silence, and it must not be read as a standing position.

None of these classifications is a recommendation to buy, to sell or to hold. None is a price target, a fair value, a valuation judgement or a forecast. They classify what is observable at the time of publication; the reader alone decides whether and how to act on that description, and does so at their own risk.

The classifications described in Section VI are updated with each edition that carries them, and only then: the daily briefing is published on market days, the weekly briefing once a week, and the monthly briefing once a month. No other update schedule is planned or promised. Readers should assume that no classification is refreshed between two editions.

What triggers a revision. A classification is reviewed at each edition and revised whenever the data no longer supports it. A revision is triggered in particular by: a change in the observed positioning or flow data that reverses or neutralises the previous lean; the kill condition stated alongside the original reading being met; a change of DEFCON alert band; or a macroeconomic or market event that leaves the previous reading unsupported. Where a classification differs from the previous edition, the change is stated in the edition that carries it.

Protocol 7 AI does not undertake to publish an update between two editions, including where a classification has lapsed under Section VI. The absence of a new edition is never a confirmation of a previous classification.

Protocol 7 AI operates an algorithmic trading system in production, on the Darwinex platform. That system takes no position in crypto-assets. Positions held by that system may overlap, at the time of publication, with instruments, indices or spreads named in a briefing.

Other relationships and remuneration. Protocol 7 AI is remunerated solely by its subscribers. It receives no payment, in cash or in kind, from any issuer, broker, exchange, index provider, asset manager or crypto-asset service provider, whether for publishing or for the content of what is published; it carries no advertising and no sponsored content; it holds no shareholding link with, and provides no corporate finance service to, any issuer named in its publications. No remuneration of any person involved in producing a briefing is tied to the direction of a classification. Should any of this change, this section will be updated before the next edition.

No link to the trading system, no account name and no performance figure for that system is published here or in the briefings. That silence is deliberate: this section is a conflict of interest disclosure, and it must not become the promotion of a financial contract within the meaning of Article 4-V of French Law No 2023-451 of 9 June 2023.