How our numbers are made

Market data

Prices, market caps, volumes, supply and all-time highs for 1,154 assets come from CoinGecko and are written into our own database. The largest 250 by market cap are refreshed every five minutes — last completed 2m ago; the rest of the top 1,000 every 30 minutes; and everything below that every six hours. We say which because the three are genuinely different, and a single number covering all of them would be true of one. Pages are rendered from that copy, not from a live call, which is why they load instantly and stay up when an upstream API does not.

Every figure on the site is therefore as of the last sync, never a real-time tick. Where a page can show you that timestamp, it does.

What we check before we publish a number

Reproducing a data feed is not the same as standing behind it. Everything that arrives from upstream is checked against the rest of the market before it reaches a page, and anything that cannot be true is held back rather than published confidently.

This started with a mistake of ours. A tokenized-stock listing arrived with a circulating supply roughly 8.7 million times its real one. That put its market cap at $25.81 trillion and ranked it first on our board, above Bitcoin; it also pushed our headline “total market cap” to $28.07 trillion and reported Bitcoin’s dominance as 4.5%. All of that sat under the words real prices, real market caps until somebody pointed it out. Two things had gone wrong: nothing checked the figure before storing it, and because the source itself ranked that asset #1926 it never appeared in our refresh again — so one bad reading simply froze at the top of our table.

What runs now, on every sync and again on a daily pass over the whole table:

• The source against itself. Each listing carries both a market cap and a rank. When a cap worth a tenth of Bitcoin arrives beside a rank in the thousands, the two cannot both be right, and we believe neither until it is resolved.
• Size against the market. Nothing is several times larger than Bitcoin. A cap that says otherwise is a broken supply figure, not an asset.
• Supply against price. Market cap is price × supply, and price is the hard one to corrupt. A supply that multiplies by twenty-five while the price sits still is an arithmetic error upstream — a real supply event moves the price too.
• Movement against plausibility. Ranks move by minutes of trading, not by orders of magnitude. A cap that jumps twentyfold between two consecutive syncs is held for review.
• Freshness against position. Anything in our top 50 appears in every single refresh, so it cannot quietly fall behind. When one does, the source has stopped standing behind a number we are still printing.
• Each asset against its own kind. The one that would have caught the SPCX row the moment it first arrived, before any of the above. See below.

Not everything here is a cryptocurrency

A tokenized Tesla share, a BlackRock treasury fund, wrapped Bitcoin and Bitcoin itself are four different kinds of claim. We label every asset with which one it is, and where the label changes how the numbers should be read, the coin page says so in a sentence. Ordinary coins get no badge — the label earns its place by being the exception.

That classification is also the sharpest check we have. Crypto market caps follow a power law: Bitcoin is around 11,400× the median coin, and that is correct, so “much bigger than its peers” can never be evidence of anything for a coin. Tokenized wrappers are the opposite. They are bounded by how much of the underlying anyone has actually tokenized, so they cluster: across every tokenized equity we track, the largest is about 3× the median, and for tokenized funds about 19×.

In a group that tight, one member hundreds of times the rest is not a valuation, it is an error. SPCX stood at roughly 370,000× the median tokenized equity. We hold anything past 250×, which leaves more than a tenfold margin above anything legitimate we have ever observed — and the comparison uses the median precisely because a single corrupt member cannot drag it upward. An average would have been pulled up by the bad number until the bad number looked ordinary.

We do not exclude tokenized assets from the rankings. They are real holdings that people are looking for, and hiding them would answer a labelling problem with a disappearance. We label them instead.

A held asset keeps its page, which says plainly that its market data is under review and what tripped. It is removed from the rankings, from the market totals and from our sitemap, and its price stays visible — price comes from trades, not from the supply figure at issue. Right now 1 asset is held.

What we deliberately do not hide. A large market cap with almost no trading volume looks exactly like corrupt data and usually is not: tokenized funds and real-world-asset products genuinely trade that thinly. Those are flagged for us to look at and published unchanged (3 currently). Hiding a real number because it is unusual would be its own kind of dishonesty — and it is worth knowing which market caps are notional rather than realisable, which is the question Sellability exists to answer.

The headline total is no longer our own sum. Adding up the assets we happen to track gave a figure that was both mislabelled and fragile — one bad row landed in it at full weight. The total now comes from a market-wide aggregate computed across every listed asset, and the line beneath it on /prices always states what was actually totalled. We also compare the two on every refresh: our top-500 sum should sit a few percent below the market-wide figure, and a wide divergence means something got past the checks above.

Our own price history

From 2026-08-09 we record one row per asset per UTC day — open, high, low, close, market cap, volume — built from our own repeated observations rather than a vendor's daily bar. 37,259 rows so far. A day we fail to record is a gap we cannot fill later, so the job that writes it is deliberately conservative: if the market data is more than thirty minutes stale, we record nothing for that interval rather than stamping an old price with a new timestamp.

Sellability

Sellability estimates what a position could realistically be sold for, as opposed to what it is quoted at. For a token and a holding size we request live routed sell quotes at a ladder of order sizes and compare each against the reference price.

The rungs are spaced by dollar size, not by proportion of your holding, because price impact responds to how much hits the pool rather than to what fraction of somebody’s bag it represents. From those measurements we interpolate how much could leave under a 1%, 3%, 5% or 10% cost — never beyond the sizes we actually quoted. Where the quotes stop, the estimate stops with them and says so.

We also watch for the point where the market stops responding. Past a certain order size an aggregator returns the same proceeds no matter how much more you offer it: that is the limit of what the pools will absorb, not the price of selling more. Reported as a ceiling — “about $46,000 is all this market will take” — because turning it into a percentage produces a figure that looks like an expensive exit and is really an absent one.

• Reference price — from GeckoTerminal. Never from a small quote: at dust sizes a quote is rounding noise, and during development that produced an exit apparently worth more than paper value.
• Quotes — the KyberSwap aggregator, falling back to ParaSwap, both routing across Uniswap v2/v3/v4, Aerodrome and others. Two providers rather than one because the single keyless source we used until 22 August 2026 went behind a bot wall without notice and took every measurement with it; a rail with one provider is a rail with an outage.
• Contract checks — GoPlus, for honeypot behaviour, sell taxes and transfer restrictions. A missing flag is treated as unknown, never as safe.
• Liquidity and pools — GeckoTerminal.

If a contract is flagged unsellable we withhold the exit figure entirely rather than print a number beside a warning, because a number beside a warning invites you to believe the number.

Exchange order books

Where an asset trades on an exchange we can read, we ask the same question of its order book: how many bids are actually resting, and what would walking them return? We read Coinbase and Kraken. Binance answers HTTP 451 from our infrastructure, so it is not consulted and we make no claim about it — that absence is a technical limit, and it is stated here so it can never be mistaken for an editorial one.

We never merge exchange and on-chain figures into one number. They are different kinds of claim. An on-chain sale is permissionless: anyone holding the tokens and some gas can make it. An exchange sale needs an account, cleared identity checks, a supported jurisdiction, the asset already deposited there, and deposits and withdrawals open at the time. Those are five conditions we cannot verify from here, so the exchange figure is labelled as conditional on access and the two are shown side by side, free to disagree.

Venue order is a measurement and is not for sale. Exchanges are listed by what they would actually pay out for your position, and nothing else. A sponsor can be listed in our directory and can pay for placement there; no amount of money reorders this comparison, adds a venue to it, or removes one from it. If that ever changes, it will be written on this page before it appears anywhere else.

There is exactly one place on a Sellability result where a link to an exchange appears: when the on-chain exit has failed and an exchange one would work. If your position cannot leave through the pools but the resting bids on a venue would take it, we say so and link there. It never appears on a good result, because somebody who has just been told their position sells easily has no use for an account they probably already have.

The venue named in that box is chosen by measurement before anything commercial is looked at, and each such link states plainly whether we are paid for it. Today we are paid for neither Coinbase nor Kraken, and the page says so. When that changes the sentence changes with it, because it is written from the same record that pays us rather than typed into the page.

Exit cost on exchanges

Many tokens — Bitcoin, PEPE, SHIB, Chainlink, Uniswap and most of what people actually hold — trade mainly on centralised exchanges, where a sell on a DEX would not show what a holder can really get. For those we read the exchanges’ public order books instead.

• Venues: Kraken and Coinbase, in US-dollar markets (or USDT/USDC where that is the only one). Binance is not included: it refuses requests from the region our servers are in, and we do not route around that.
• When: the largest 25 assets every half hour on Kraken; the top 250, plus every token the on-chain sweep skips as trading mostly off-chain, once a day on both venues. A figure is shown only if it is at most two days old; with several readings in a day we show the median.
• What: we walk the resting bids selling $100,000 and compare what comes back with the same amount valued at the mid price. That includes both the spread and the depth, the same basis as the on-chain figure. It is before trading fees, which depend on the account.
• Checked: a venue’s price must agree with ours within 10%, or the reading is discarded: two exchanges can list different assets under the same ticker.
• Never blended. Each exchange is its own answer, shown beside the on-chain figure and never averaged with it. Using a book needs an account, cleared identity checks, a permitted country and the tokens already deposited — none of which we can check for you.
• A snapshot: a book is the orders resting at one moment. Selling over minutes can do better as the book refills; a sudden sell-off can do worse.

What Sellability cannot tell you

This is the important section.

• Quotes are estimates, not fills. The price moves while you execute, and gas is not included.
• We cover on-chain routes only. For a token that also trades on centralised exchanges, our figure is a floor — you would likely do better on an exchange. Where we can detect this, we say so above the numbers.
• We verify our own arithmetic against on-chain reserves only for constant-product pools. Concentrated-liquidity pools (Uniswap v3) and Uniswap v4 — which keeps every pool's tokens in one shared contract — cannot be independently checked this way, and we report that rather than counting them as verified.
• We do not publish a letter grade. Grades are easy and we have not yet earned one: the back-test that would justify it covers too few pool types. Until then you get a confidence level instead, and "low" is a real answer.
• A quote we can prove is wrong is discarded, not shown. If a smaller slice reports worse friction than a larger one — impossible in a real market — that rung is marked unquotable.

The crypto calendar

Every dated entry at /calendar is added by a person and checked against two sources on different hosts before it is published. A project’s blog post and that same project’s documentation count as one source, not two, and the guard refuses the pair. There is no free feed for token unlocks or exchange delistings anywhere, so the alternative to curating was not automating — it was republishing somebody else’s unverified table.

• We publish only as much of a date as we know. A project that announced “Q2 2027” appears as April 2027 and never as 1 April 2027. Inventing a day is the most common error on every unlock tracker, and it happens because their schema has one date column and a date column always has a day in it.
• Ethereum upgrade times are derived, not reported. An upgrade activates at an epoch; an epoch is 32 slots; a slot is 12 seconds; the beacon chain genesis is a constant. No node and no vendor is involved, and the arithmetic reproduces every upgrade since Altair to the second — which our own test asserts on every run.
• The before-and-after on a past event is our own daily sweep, reported as a median over fourteen days either side, on one market, with the reading count shown. Below five daily readings on either side we publish nothing. The day of the event is in neither window.
• No forecast, on any entry. An unlock page states how many tokens arrive, what that is worth, how many days of real trading volume it represents and how it compares with the depth we measured. It does not say what the price will do, and neither will we.
• Tax dates are statute, not advice. We are not tax advisers. Deadlines move for extensions, agents, weekends and holidays; the authority is linked on every entry.

Projects and official websites

Search knows a list of crypto projects: the coins CoinGecko lists, the exchanges CoinGecko lists, the blockchains in Chainlist, and the listings in our own directory. For each one we keep what kind of thing it is and the web addresses a source gives for it, and every address carries the name of the source that gave it. This is an index, not an endorsement. A project is in the list because a public registry lists it, which says nothing about whether it is any good; the reviewed shelf is the directory, and the two are never presented alike.

• How sure we are that a site is official is said in words, never as a number. There are four levels, from the strongest down: verified by the project (the project proved it controls the domain), verified by Cryptominium (a directory listing whose domain we checked inside its verification window), listed as official by two sources, and listed by one. Two sources count only when they are independent of each other. Nobody can pay to move up a level.
• The website check never visits the site. When you paste an address into search, we reduce it to its registrable domain (app.uniswap.org becomes uniswap.org, using the Public Suffix List) and compare that name with the project list. We do not fetch the page, render it, or look at what it says, and nothing about the address you pasted is stored.
• An address that is not on the list is not marked down. “Not the official site of any project we know” is all we say, because an unknown site is not evidence of anything. We never label a site good or bad and we give no score.
• One thing we do flag: an address whose name resembles an official one. It is compared only with sites at the “listed by two sources” level or better, so one registry’s typo cannot become an alarm. Five rules, each plain string comparison:
  • Same name, different ending: uniswap.com next to uniswap.org.
  • A letter or two off: unlswap.org, for names of five letters or more (one letter under eight, two from eight).
  • The official name plus a word from a short list such as app, claim, airdrop or wallet: uniswap-app.com.
  • Look-alike characters that spell the official name once swapped for plain letters, such as a Cyrillic “і” for an “i”, or a digit one for a lower-case L.
  • The official name buried in a longer address that belongs to somebody else: uniswap.org.claim-now.com is a claim-now.com address.
• Not every project is here yet. DefiLlama’s protocol list is not included: its terms require written consent for commercial reuse, and we have asked. Until it says yes, a protocol with no token of its own may be missing.

Credits

• Chain data: ethereum-lists/chains (Chainlist), MIT licence.
• Coin and exchange data: CoinGecko.
• Reading web addresses: the Public Suffix List, Mozilla Public License 2.0.

What changed

Each coin page carries a short block saying what moved this week, and it is built from nothing but the measurements described above: our daily price and volume records, the cost of selling $100,000 on-chain and on exchanges, and the calendar. No news, no social posts, no sentiment. Every Monday the same blocks for every coin are gathered into one dated page, What changed this week, and kept unchanged after that.

• Each figure is compared with a normal month. “Now” is the latest stored day; the comparison is the median of the 30 stored days before the last 7, so a week’s move is measured against the month before it and not against itself. Missing days are left out, never filled in.
• A change has to be large to be shown. The cost of selling on-chain must move by at least 1 percentage point and a quarter of its usual level, or cross one of the published bands; volume must move 3-fold either way; supply must move 5% or more in a month. Most weeks nothing clears the bar, and the block says so in one line.
• Too little history, no block. A coin needs 10 to 14 days of records before anything is compared. A new coin shows nothing rather than a comparison with almost nothing.
• We never say why. The block can put two changes side by side — “the price rose while selling got more expensive” — and that is the whole of its interpretation. Anyone who tells you why a price moved an hour after it moved is guessing.
• Figures under review are withheld. When our integrity checks hold a coin’s market data, the block says so and leaves out price, volume and supply. Exit costs are measured separately and still stand.

What a coin can do

Some coin pages carry a block headed “What it can do”, and Explore gathers them. It asks the same 9 questions of each coin and answers each one yes, partly, no or it depends. Beside every answer is who you depend on for it, because a coin doing something itself and a company doing it for you look the same in an advert and are not.

• Every written answer has a source. The source is the project’s own page or its own terms, it is linked under the answer, and the date beside it is the day we last read it (2 October 2026 at present).
• A blank is not a “no”. Where we could not find a source for an answer, the row is left out. A coin with no block is a coin we have not written up, not one that failed.
• “Can it be sold” is never written. It is the measurement described above: what selling $100,000 returned on-chain or on an exchange, with its date. A sale that lost 15% or more is shown as “partly”.
• Some answers are written once for a kind of asset. A stablecoin, a wrapped token and a tokenized share each get one sentence that is true of every member, which is why those mostly say “it depends” and on what. They appear only where we are sure of the kind.
• “Through whom” names, and never rates. It says who is involved: nobody, an issuer, a custodian, an exchange. It does not say whether they are good at it.
• It is not a reason to buy. The block describes how a coin works. It gives no score, calls nothing safe, and recommends nothing.

Money, and what it cannot buy

Cryptominium will carry sponsored placement and affiliate links, and every commercial link will be marked. Two rules do not bend:

• Sponsorship buys position, never a rating. No payment changes a score, a review or a Sellability result.
• Commercial relationships are disclosed on the page where they apply, not buried here.

What we will never ask for

We will never ask for your seed phrase, private key, or exchange password. No page on this site will ever be a box you type them into. Our recovery tooling runs offline on your own machine for exactly this reason. Any site — including one claiming to be us — that asks for a seed phrase is stealing from you.

Sellability needs no wallet connection. You paste a public contract address and a number.

Privacy in one paragraph

We run no third-party trackers and no analytics scripts in your browser — no Google Analytics, no tag manager, no pixels, no advertising networks. We do count page views, on our own server, using an identifier that is rebuilt daily and so cannot follow you from one day to the next; that is why our own traffic figures are quoted as visitor-days rather than as visitors. The only cookies are the ones that sign you in and protect the forms. Your theme and currency choices are stored in your own browser. Tool checks are logged with a salted hash of your IP address — enough to stop one visitor exhausting the free services we depend on, not enough to identify you. Full detail, including the one third party that sees an IP address, on the privacy page.

Corrections

If a number here is wrong we want to know, and we would rather fix it than defend it. Email [email protected].

Cryptominium provides market information and educational tools. We are not a broker, exchange, custodian or financial adviser, we never take custody of your assets, and nothing on this site is financial advice or a recommendation to buy or sell. See the full disclaimer.