Methodology

Observation classes

Every figure carries a class that says what kind of number it is. The class is set by the adapter that read it, travels with the value, and is shown beside it. It is never inferred from how recent the number looks.

LiveA continuously updated market quote.

BTC and ETH from CoinGecko. The asset trades continuously and CoinGecko stamps every update with its own time, which VILIQ shows. A price that arrives without a source timestamp is refused rather than stamped with the time VILIQ received it.

Daily referenceAn official once-a-day reference rate, not a tradeable quote.

Euro reference rates from the European Central Bank, published once each working day at around 16:00 CET, and the U.S. Treasury yield curves. The ECB says its rates are published for information only. They are not prices anyone can deal at, so VILIQ never presents them as live and never draws movement within the day.

Official fundamentalA measured quantity from an official statistical release — not a price.

Measured quantities from official releases: U.S. Energy Information Administration weekly inventories, production, trade and refinery utilisation; the Treasury General Account; the New York Fed’s overnight reverse repo; central bank policy rates. An oil inventory is barrels in tanks, not what a barrel costs.

VILIQ calculatedVILIQ computed this from published inputs, which are named.

Numbers nobody published, which VILIQ computed from published inputs: FX crosses, curve spreads and policy-rate differentials. Each one names its formula and its inputs, with the date of each input.

Every figure shows its source, the time the source observed it and a freshness grade. When no approved source can supply a value, VILIQ says the value is unavailable. It does not substitute a zero, an estimate, or an old value presented as current. Absent, never zero.

What VILIQ serves today

• BTC and ETH spot prices from CoinGecko.
• ECB daily euro reference rates, and crosses VILIQ derives from them.
• U.S. Treasury daily par yield curve and real yield curve, including the real 10-year yield.
• New York Fed effective federal funds rate (EFFR) and overnight reverse repo (ON RRP).
• The Treasury General Account from the Daily Treasury Statement.
• Policy rates from six central banks: the Federal Reserve (effective rate, via the New York Fed), the European Central Bank, the Bank of Canada, the Reserve Bank of Australia, the Bank of England and the Swiss National Bank (via the Bank for International Settlements, because the Swiss National Bank’s own terms allow only non-commercial use).
• EIA weekly energy fundamentals — quantities, not energy prices.
• Headlines and links from eight primary publisher feeds; no article text is stored.

What VILIQ does not serve

Equity indices, single stocks, metals, the dollar index (DXY), VIX, energy prices, futures, options, a daily or live USD/VND rate and Vietnam dealer gold quotes. No approved, licensed source is connected for any of them, so they are shown as unavailable rather than estimated. (The dong has only an official monthly reference from the BIS, about three months behind.)

Freshness

Freshness is judged against each publisher’s own calendar, not a single clock. A figure is stale when a publication that should have happened has not — not because a number of hours passed.

For the business-day publishers, one or two missed publications make a figure stale; three or more make it out of date, and it is then not presented as current. Public holidays are not modelled, so on a holiday a figure can read one publication older than it is — an error in the cautious direction.

VILIQ calculations

Where VILIQ computes a number, it says so, shows the formula, and only combines inputs that describe the same moment.

Licences and attribution

A source that can be reached is not a source that may be shown. Each publisher’s terms were read, and an adapter whose terms are not settled does not run in production.

• Refused before the request is made: fetching data VILIQ may not display would still leave a copy in a cache and a log.
• Approval is per series, not per provider. FRED’s VIX series is copyrighted by Cboe and is not shown; its high-yield credit spread needs ICE’s prior permission and is refused everywhere.
• Stooq (index levels, metals, DXY, VIX) is built but refused: its terms do not settle republication.

Attribution each source requires

• CoinGecko — shown as “Powered by CoinGecko” wherever its prices appear.
• European Central Bank — cited as the source, crosses labelled as VILIQ’s modification, and this notice shown with every FX table:These euro reference rates may be obtained free of charge from the European Central Bank at ecb.europa.eu.
• Federal Reserve Bank of New York — not public domain. Its terms require its copyright notice, which VILIQ shows verbatim:© 2026 Federal Reserve Bank of New York. Content from the New York Fed subject to the Terms of Use at newyorkfed.org.Reproduced in English, as the publisher’s terms require.
• U.S. Treasury and U.S. Energy Information Administration — public domain. VILIQ names them as a courtesy.
• Bank of Canada and Reserve Bank of Australia — attribution, plus a notice that the rate is available from the bank free of charge.
• Bank for International Settlements (the Swiss National Bank’s policy rate) — the BIS cited as the source, no additional charge to users for the figure, no implied BIS endorsement, and, on the Vietnamese pages, a statement that the translation is not an official BIS translation.

Charts

A chart is drawn only from publisher history VILIQ has stored. Nothing is interpolated, smoothed into a trend, or filled in where a publisher said nothing.

• A daily reference is drawn one point per publication. No intraday movement is invented between two fixes.
• Candles appear only where the provider supplies genuine open, high, low and close. For BTC and ETH, CoinGecko supplies 30-minute buckets for one day and 4-hour buckets for one month; the bucket width is stated on the chart.
• Curve spreads are drawn only on dates where both tenors were published.
• No volume is drawn: the provider’s volume is a rolling 24-hour total, not per-candle volume.

When a chart is not drawn, the reason is named

NO_SOURCE

Coming soon. Coming soon: awaiting an approved source for this series.

LICENCE_BLOCKED

Coming soon. Coming soon: a source exists and VILIQ is awaiting the licence to display it.

NO_HISTORY

No history. The history for this series has not been collected yet.

STALE

Stale. The stored history is too old to present as current.

Images and video

VILIQ shows an image or a video only when it is real, verified and from a source whose terms allow it. Otherwise the content is shown without media.

• Video comes from five central banks’ own YouTube channels — Federal Reserve, European Central Bank, Bank of England, Reserve Bank of Australia, Bank of Canada. Each channel was identified from a link on the bank’s own website, not from its handle. Videos play in YouTube’s privacy-enhanced player, loaded only when you press play; nothing is downloaded or copied. Duration and captions are not stated by the channel feed, so VILIQ does not state them.
• Of the eight news publishers VILIQ reads, one publishes an image that belongs to each article: the U.S. Energy Information Administration’s Today in Energy, whose chart is public domain and is shown with EIA’s dated acknowledgement. The other seven declare only a logo or a site-wide share card, which is not article media, so their news appears as a source card naming the publisher. No logo is copied. A news item may instead carry the same central bank’s own video when it is about the same decision and published within three days.
• Video beside a VILIQ Research answer is chosen by the question: the central banks it names, by name, currency or pair. A question about Bitcoin, energy or a trust check gets no video, because no central bank’s channel answers it — no video is better than an irrelevant one.
• Course cards carry abstract illustrations VILIQ draws itself. They are decoration, never evidence of a market or an event.
• Every video is checked on each collection: it must still be public and embeddable (YouTube’s oEmbed answers only then), and its thumbnail must be a real image of a real size. A video that fails disappears on the next run rather than becoming a broken card.

VILIQ never attaches a stock photo to real news, never generates an image and presents it as documentary, and never shows a thumbnail from a host its rights registry does not list.

Trust Check

Model version: trust-1.0.0

Trust Check reports what it could verify, what it could not, and what it did not check. Those are three different statements and are never merged.

• Checked — the registry answered and holds a record.
• Nothing found — the registry answered and holds nothing. Not found is not the same as false: it means there is nothing to verify against, not that anyone did wrong.
• Not checked — VILIQ could not reach the registry. It never counts as a pass and never lowers or raises the score.
• Not applicable or unknown — the input does not allow the check, or no source is connected for it.

Today one registry is connected: RDAP, the domain registries’ own public protocol. It confirms that a domain exists and when it was registered — not who operates it. Australian ABN and ACN numbers are checked against the published ATO and ASIC checksum algorithms; a valid checksum means the number is well-formed, not that it is registered. Company, licence, warning-list and IP registries are not yet connected, and those components read not checked.

Trust Score

The Trust Score has eight components, each scored from its best evidence. Coverage caps the result — cap = 40 + observed weight × 60 — so a subject verified on two components cannot read as strong. Risk indicators apply an itemised penalty that is shown, not folded into the number.

Wording stays evidential. VILIQ reports a risk indicator or a mismatch; it does not state that a person or business has committed an offence on the strength of an automated check.

Trust Score measures available legitimacy and evidence signals. It does not guarantee investment safety or future value.

Scores and models

The engines below describe measured conditions. None of them is a recommendation, and none is a prediction of price.

Market Score

Model version: market-score-1.0.0+outlook-1.0.0

The 0–100 headline is a composite of measured conditions: observed pressure, money-flow pressure for the asset’s class, regime alignment, momentum, trend position and realised volatility. It describes the present. The outlook is kept separate, as a probability across bullish, neutral and bearish for each horizon; high volatility and thin data widen the neutral band rather than push the distribution in a direction it has not earned.

Buy and sell pressure are measured only from realised price and volume: where each bar closed in its range, volume on up and down bars, and net drift scaled by the asset’s own volatility. It can be checked against a chart by hand.

Market Score describes measured market conditions and model probabilities. It is not a prediction of future price and not a recommendation.

Prediction accountability

Every outlook is written to an append-only ledger with its model version, a fingerprint of its inputs, and the rule it will be judged by — fixed before the outcome is known. Database triggers reject edits and deletes, including of the wrong ones.

• 1D: Close-to-close return over 1 calendar day: bullish above +0.5%, bearish below −0.5%, otherwise neutral.
• 7D: Close-to-close return over 7 calendar days: bullish above +1.5%, bearish below −1.5%, otherwise neutral.
• 30D: Close-to-close return over 30 calendar days: bullish above +3.0%, bearish below −3.0%, otherwise neutral.
• 90D: Close-to-close return over 90 calendar days: bullish above +5.0%, bearish below −5.0%, otherwise neutral.

Calibration is measured with the multi-class Brier score against a uniform baseline, per horizon, always with its sample size.

Market regime

Model version: regime-1.0.0

Available macro factors are collapsed into seven dimensions — liquidity, growth, inflation, rates, credit stress, volatility and breadth — and compared with a published reference profile for each of eight regimes. Confidence rewards separation from the runner-up and is capped by data coverage; below the coverage floor no regime is named.

Money flow

Model version: flow-1.1.0

Node pressure is a weighted blend of normalised factors, rescaled by the weight actually observed so missing inputs do not drag a node toward zero. Edge strength combines upstream pressure with measured co-movement, and lead or lag is measured, not assumed. The chain defines which relationships are examined; it is not a claim that capital always moves in that order.

Cycles and analogues

Calendar cycles — halving, election, expiry, seasonality — are labelled calendar correlation, never causal mechanism. Historical analogues standardise macro features and match by nearest neighbour; the result is the spread of what followed across matched episodes, so one dramatic episode cannot pose as the expected outcome.

Historical analogue — not a guarantee of future performance. Similar conditions have produced a wide range of outcomes.

IP Score

Model version: ip-1.0.0

Registrations are weighted by right type and status, and the chain from legal entity to registration, domain and product is traced. Its most useful output is a mismatch — a name marketed by one party and registered to another. It is not a legal opinion. No IP registry is yet connected for automated lookup, so without evidence supplied from a registry the IP components read not checked.

VILIQ Research

VILIQ Research answers only from tools. Each tool is a deterministic engine or a provider observation, with a strict input and output schema and a permission check that runs inside the tool.

Before an answer is sent, every number in it is matched against the values the tools returned. An answer containing a price, score, rate, registration or other figure no tool produced is withheld, not sent. Dates are treated as provenance rather than values, but a fabricated price beside a date is still caught.

Web pages and documents VILIQ Research reads are data, not instructions. Attempts to redirect it are neutralised and shown to you, because a page that tries to manipulate an assistant is itself a trust signal.

The tools VILIQ Research may call, by their internal names. Some have no source connected yet; they report unavailable rather than being skipped quietly.

• getMarketSnapshot
• getMacroObservations
• getNewsEvidence
• getEnergyFundamentals
• getCentralBankRates
• getMoneyFlow
• getRegime
• getAssetScore
• getHistoricalAnalogues
• getEventImpact
• getTrustEvidence
• checkCompany
• checkDomain
• checkIP
• checkToken
• getGoldVietnam
• getVietnamCompanies
• getAcademyLesson
• getUserSkillProfile
• getMyRewards
• getMyCredentials
• getMyFoundingStatus
• getPartnerStatus
• getMyRewardStatus
• getMyCampaignEditability
• getMyProgression

When no language model is configured, VILIQ Research answers directly from engine output. The grounding rule is the same either way.

Academy credentials

Academy progress is graded on the server from recorded answers. Nothing a page says about completion is trusted.

A lesson quiz passes at 70% correct. Answers are graded on the server, and the correct answers reach the page only after yours have been graded.

Credentials are issued per level. Each level names the courses it requires, a minimum quiz score that every lesson in those courses must meet (the lowest score counts, not the average) and, for advanced levels, an accepted challenge. The server recomputes eligibility from recorded progress before issuing, and each credential records the course versions it asserts.

PRO Candidate is where the Academy stops. It says a member is ready to be considered; VILIQ PRO itself is a human decision and is never earned by completing courses alone.

VILIQ Points

VILIQ Points today is points and nothing more.

• Off-chain points recorded by VILIQ, earned through platform activity.
• Not transferable between members and not tradable anywhere.
• No cash value, no redemption for money, no yield, no ownership, equity, profit share or revenue share.
• Not an input to voting power: a large balance cannot buy a vote.

VILIQ Points recognise eligible participation and contribution within VILIQ. They are off-chain, non-transferable, not purchasable or sellable, and have no market value.

VILIQ Points recognise eligible participation and contribution within VILIQ. They are off-chain, non-transferable, not purchasable or sellable, and have no market value.

Voting

Voting is reputation-weighted and limited to community matters. It cannot decide anything that belongs to the law or to the company.

A proposal touching a reserved matter is refused when it is created, not overruled afterwards. Voting confers no ownership of, or control over, the company.

To vote, a member needs verified identity, an account at least 14 days old and Academy level 1 or above, and must not be under an anti-abuse restriction. Identity verification is not yet available on VILIQ, so no member can vote today.

Voting power comes from demonstrated knowledge, verified contribution and community standing; each input is capped and so is the total. A result stands only if the required voting power took part; a tie carries no option.