Methodology

Coinmico publishes its own market data. Prices, volume, capitalisation, candles and rankings are computed from trades we collect directly from exchange streams and on-chain pools — not requested from another aggregator when you open a page.

Three rules govern everything below.

• Measured, not reported. A venue's own volume figure is never displayed as ours; what we show is the traded value we observed on that venue.
• Labelled where it is not ours. Where a figure comes from an ingested source or an emergency fallback, the page says so instead of presenting it as our measurement.
• Silent where we cannot measure. A window our history does not cover, or a metric a node cannot answer, is left empty rather than approximated into something that reads authoritative.

• Centralised venues: live trade streams from around thirty exchanges, including Binance, Coinbase, Kraken, OKX, Bybit, KuCoin, Gate, Bitget, MEXC, HTX, Bitfinex, Bitstamp, Crypto.com, Upbit, Poloniex and WhiteBIT.
• Perpetual venues: futures and perpetual tapes from the same venues plus Hyperliquid, kept in separate buckets from spot because a perpetual trades at a funding-driven premium or discount.
• On-chain venues: swap events decoded from Uniswap V2/V3, PancakeSwap, SushiSwap, Aerodrome and other Solidly and Slipstream deployments across roughly twenty EVM chains — Ethereum, Base, Arbitrum, Polygon, BNB Chain, Avalanche, Optimism, Tron, Gnosis, Sonic, Mantle, Linea, Scroll, Blast, Berachain, Celo, Cronos, Unichain, PulseChain and HyperEVM — plus Solana pools read over JSON-RPC.
• Order books: top-of-book quotes for every market a venue lists, and rotating depth samples for the deepest markets, used for spread, depth and liquidity figures.
• Contracts: ERC-20 total supply and burn-address balances read directly over JSON-RPC on the chains we run a client for.
• Reference rates: fiat quotes from an independent FX feed; BTC and ETH cross-rates are derived from our own index.
• Ingested metadata: asset descriptions, links, contract addresses, sector tags, reported supply and logos are ingested from public sources and stored on our side. This is also the emergency fallback if our own pipeline is interrupted, in which case the affected view is labelled as fallback data.
• Holder distribution: read from public Blockscout instances and labelled as such, because replaying every token transfer is beyond the public gateways we run on.
• Long-tail quotes: curated assets our venues and pools do not reach yet are quoted from DEX Screener for the token's contract, stored under that source, and never mixed into our index.

The ingester tracks the largest assets by capitalisation — a universe of about two thousand, re-resolved every six hours — together with every pool our on-chain indexers watch: up to 1,500 pools per EVM chain and several hundred on Solana, deepest first, with newly discovered pools folded in as discovery reaches them.

Assets outside that universe are listed from stored metadata until liquidity brings them into range. Venue and pool rankings on the exchanges pages are computed from the same tape as the coin pages, so the numbers reconcile.

Every trade we receive is bucketed per venue over a rolling one-minute window, and the index settles every two seconds:

• Each venue contributes one volume-weighted average price for its window. Spot and perpetual on the same venue are separate buckets, and a perpetual carries a quarter of a spot market's weight.
• The cross-venue median is taken from the spot buckets, so a funding premium cannot drag the reference away from the cash market.
• Venues more than 2% away from that median are dropped as outliers; if that would leave nothing, the full set is kept.
• The surviving venues are weighted by traded notional, so the deepest markets dominate the settle.
• A settle that moves the price by more than 50% is held back unless the venues behind it are the ones that set the standing price, or outnumber them — the guard against a pair or pool mapped to the wrong token. After thirty minutes without a settle the standing price stops vetoing.

Each published price carries the number of venues behind it and a confidence value built from that count and the spread between the venues that agreed. A price with no fresh trades behind it for a minute is marked stale rather than shown as current.

An asset no exchange trades is settled from its pools instead, on the same arithmetic over a wider window: every pool that swapped in the last thirty minutes contributes the volume-weighted price of its own swaps, pools more than 8% from the cross-pool median are dropped, and what survives is weighted by traded value. It is published under its own source, requires at least three swaps and USD 500 of turnover in the window, and always yields to a venue price when one settles — a long-tail token whose pools swap a few times an hour therefore has a price of ours rather than none.

Every market we ingest is re-measured against its asset's own index price every ten minutes, over a seven-day window so that quiet venues stay under review long after they stop trading. Two tests apply:

• Outlier: a market whose last print sits more than 30% away from the level its asset's other markets agree on is excluded from candles, volume and extremes. Every quote leg we track is dollar-pegged, so a gap that wide is a mapping error rather than a currency or a basis.
• Identity: a level more than six times above or below the asset's listing price is read as another token's price, not this asset's. Levels within a factor of two of the listing price are accepted as the asset's own before depth is considered, so a deep collision cannot outvote the level the listing recognises.

Excluded markets stay visible with their exclusion recorded rather than being deleted, and a venue can be excluded from the index entirely while remaining listed.

Volume is traded value we observed, not a figure a venue reports. Trades are folded into one-minute candles per market, and the 24-hour figure on a coin, venue or pool page is the sum of those candles over the trailing day, excluding markets the integrity checks removed.

On-chain volume is the dollar value of the base leg of each swap, taken from the swap's own price. A swap whose implied price sits more than five times away from the index price of its base asset is left out of the roll-up, because a pool whose decimals we read wrong would otherwise dominate a venue's volume.

Market capitalisation is our reference price multiplied by circulating supply. Fully diluted valuation uses maximum supply where an asset defines one, and is omitted where it does not.

Reported total, circulating and maximum supply are ingested and then curated on our side: allocations that are on-chain but not in the hands of the market — team, foundation, treasury, escrow and unvested grants — are subtracted before a figure is called circulating. Where a token contract lives on a chain we run a client for, total supply and the balances at the burn addresses are read from the contract itself, so a burn is reflected without waiting for a third party to notice it.

Raw trades are batched into the tick store, folded into one-minute candles, and rolled up from there into hourly and daily history. Each pass re-aggregates the last ten minutes of candles and the last three hours of history, so trades that arrived late or during a restart are still counted. Daily history further back than our own stream reaches is walked in from venue kline endpoints, deliberately slowly, and labelled where a chart interval is not yet covered by stored history.

The 1-hour, 24-hour and 7-day changes are measured against a sample our index actually holds near the reference point — within ten minutes, two hours and twelve hours respectively. Without that rule the first minutes after a cold start would compare against whatever the oldest stored row happened to be.

Longer changes (14-day, 30-day, one-year) and the 7-day, 30-day and one-year highs and lows are published only once our stored history spans the window. Until then the field stays empty, or the coin page falls back to the ingested profile figure and says so. Each asset carries the date its measurement begins, so a reader can tell how far back our own record reaches.

Ranks are recomputed across the whole catalogue from our own capitalisation figures, so the number shown on a coin page is the same number the listing uses, and the coin table opens in market-cap order by default. Gainers, losers and screens run over the same rows.

Trending measures what the market is crowding into, from our tape rather than from a search box: turnover (24-hour volume against market capitalisation) contributes 55% of the score, the 24-hour move 30% and the 7-day move 15%. Each term is a percentile rank rather than a raw value, so one hyper-volatile micro cap cannot dominate the board, and only assets above about USD 10 million in capitalisation with measured volume are eligible.

New coins lists assets our catalogue had never carried before, detected by an ingester pass rather than picked from the coins we already cover. Assets that were already in coverage when the detection date was introduced are dated from our own price history and are never treated as new.

A detection is only published once every one of these is true of figures we measured ourselves.

• Detected inside the last 30 days, and observed trading for at least 24 hours before it is published, so a pair that appears and dies never reaches the page.
• Priced on our own index at the last settle, across at least two venues and above the engine's confidence floor. An asset carried only by an ingested or fallback quote is not eligible.
• At least USD 250,000 of 24-hour volume on the markets we measure, and a capitalisation of at least USD 1 million computed from supply we track.
• Trading on at least two venues or on-chain pools, which is what keeps dust-liquidity pairs and single-pool tokens off the page.

A token no venue trades yet takes the on-chain track of the same screen. Its price is the one we settle over the pools we index ourselves, and it is judged on measured swap turnover instead: at least USD 100,000 across two or more pools that traded in the last day, or USD 300,000 where only one pool exists, priced above the settle’s confidence floor and held to the same 24-hour observation window. No market capitalisation is required of it, because a token this young rarely has a settled supply — the column stays blank rather than being filled with an estimate. These rows are marked On-chain on the page.

Qualifying assets are released into separate hourly slots, at most ten a day, with each slot fixed from the asset’s own identifier so the page fills through the day and never reorders itself between requests. The overflow waits for the following day. When no detection qualifies the page stays empty; nothing is rotated in to fill it. Screening is a data-quality filter and not a review of a project, its team or its contract.

A listing request approved by our reviewers is published on that decision instead of the thresholds above, since a person has already checked what the submission claims. It appears as soon as it is approved, outside the daily cap and its slots, and its price, volume and capitalisation stay blank until our own index measures a market for it. An approval is not an endorsement of the project.

Our Fear & Greed reading is computed from the history we recorded ourselves — nothing is copied from another sentiment index — over a basket of the hundred deepest assets. Six components each land on a 0–100 scale before weighting:

A component is published only once its baseline window is nearly full, and the yesterday and last-week readings are the same arithmetic re-run on those days rather than a stored score whose method may have changed since.

Venues are ranked by the volume we measured on them, and each venue's share is computed against the same total, so its rank and its share always agree. Pair and pool tables read hourly summaries that include the hour in progress, which is why a busy pool's figure moves within the hour. Liquidity, spread and depth come from the order book and pool reserve samples described above.

Sector pages, dominance and the global totals are aggregates over the assets we track: the newest index row where the ingester has published one, and stored listing metadata where it has not. Sector membership comes from ingested tags, curated on our side; a coin can belong to several sectors, and each sector's capitalisation and volume are the sum over its members.

Automated measurement has failure modes, and we would rather name them than imply they do not exist: a venue stream can drop, a chain indexer can fall behind its head, a token's decimals can be read wrong, a newly listed pair can be thin enough that its own prints are noise, and an ingested supply figure can be out of date. The stale marker, the fallback label and the exclusions above exist for exactly these cases.

Figures are revised when late data arrives or an error is corrected, so a value read now may differ from the same value read later. Nothing on the site is investment advice — see the terms of service.

If a number looks wrong, tell us: send an asset or market correction through the listing form or write to [email protected]. We publish material changes to this methodology here, with the date above.