What wash trading is, and why volume can be invented

Trading volume is the number that decides which exchanges appear at the top of a list, which tokens get written about, and which venues look like the centre of a market. It is also, almost everywhere in crypto, reported by the venue whose ranking depends on it.

That arrangement — self-reported figures, ranked, with money attached to the ranking — produces exactly what you would expect it to produce.

What wash trading is

A wash trade is a trade in which the same party sits on both sides. The buyer and the seller are the same person, or two parties acting in concert, and when it settles nothing has changed: the beneficial owner of the asset before and after is identical.

The trade is entirely real as a record. It prints on the tape, it increments the volume counter, it appears in the data feed that aggregators collect. It is fictional only in the sense that matters — no risk was transferred and no position changed hands. It is a transaction that exists to be counted.

Why anyone bothers

Because volume is the sort order.

For an exchange, ranking near the top of a volume table is a marketing channel: it is how new users find a venue, how tokens choose where to list, and how the venue argues it has the liquidity a large trader needs. For a token, reported volume is a listing criterion elsewhere, an eligibility threshold for index inclusion, and a proxy readers use for whether anyone cares.

Some venues went further and paid for it directly. Trade-mining schemes rebate traders in the exchange’s own token in proportion to the volume they generate, which converts wash trading from a deception into a product feature: the volume is fake, the reward is real, and both parties know it.

The best-documented measurement of it

On 19 March 2019, Bitwise Asset Management presented to the SEC in support of a proposed bitcoin ETF. The presentation is a public document, filed to the comment record for the NYSE Arca rule change.

Its central slide compares two numbers for the same week. Reported spot bitcoin volume across the venues listed by the leading aggregator: approximately $6 billion a day. Volume the analysis judged to be real: $273 million a day. That is 4.5 per cent, which is where the widely-quoted “95 per cent of crypto volume is fake” comes from — it is a rounding of the other side of that figure.

The deck identified ten exchanges whose reported volume it considered genuine, and showed them trading as a single tightly-arbitraged market with deviations inside the expected band once fees and hedging costs are accounted for.

Two things need saying alongside that, because a number this striking gets quoted without either.

Bitwise was an applicant making a case. The firm wanted its ETF approved, and “the real market is small, orderly and hard to manipulate” was the argument it needed to win. That does not make the analysis wrong, and it was corroborated independently — the New York Attorney General’s Virtual Markets Integrity Initiative approached thirteen exchanges, dominated by the same ten, and the Blockchain Transparency Institute separately flagged fifty-six venues for suspected fake volume, none of them among the ten. But the source had an interest, and a reader should know that before repeating the figure.

It is a snapshot from 2019. It describes one week in March of that year. It is not a current measurement, and anyone citing it as one — including anyone citing this article — is misusing it.

How it is spotted from outside

The methods in that presentation are still the ones that work, and none of them requires access to the venue’s internals.

• Trade size distribution. Humans and real trading systems cluster on round numbers — whole coins, round fiat amounts. Generated volume tends to produce a smooth, bell-like distribution with none of that clumping, because the thing producing it has no reason to prefer round sizes.
• Spike alignment. Genuine markets react to the same events at the same moment. A venue whose volume spikes do not line up with anybody else’s is reacting to something only it can see.
• Spreads that do not make sense. Impossibly tight spreads on a venue with no visible market makers, or trades repeatedly printing inside the prevailing bid and ask, which real order matching cannot do.
• Gaps. Hours or whole days of exactly zero volume that correlate with nothing — not business hours, not volatility, not uptime.

What changed, and what did not

The aggregators responded. Headline tables are now commonly built on adjusted, confidence-weighted or liquidity-derived metrics rather than raw self-reported volume, and web-traffic and order-book depth checks sit behind the rankings.

The structure did not change. The venue still reports its own numbers, the ranking those numbers feed is still worth money, and the adjustment methodologies are themselves things to be optimised against. Improving a gameable metric produces a better metric, not an ungameable one.

Where the rules now bite

MiCA’s market abuse provisions apply to crypto-assets admitted to trading, or for which admission has been requested, on a platform operated by an authorised provider in the EU.

Article 91 defines market manipulation to include entering into a transaction or placing an order that gives false or misleading signals as to the supply, demand or price of a crypto-asset, or that secures its price at an abnormal or artificial level. It does not use the words “wash trading”. It describes it. Article 92 requires persons professionally arranging or executing transactions to operate systems to prevent and detect market abuse, which makes detection an obligation rather than a courtesy.

That is a genuine change from 2019 — for assets and venues inside the perimeter. It says nothing about the ones outside it, which is most of the market by venue count, and what a venue is regulated for varies enormously.

Why this publication cares about it

Because it is the same failure the rest of this site keeps running into. A number that the measured party reports about itself, feeding a ranking that is worth money, will be optimised rather than reported. It is the reason a reserve attestation needs reading carefully rather than trusting, and it is the reason the AI Citation Index publishes its question set, fixes it in advance, and states its method — including the parts that weaken it.

A measurement nobody can check is not a measurement. It is a claim with a number attached.