This article describes market-wide practices documented by academic and regulatory research; it accuses no specific venue or token. Our own volume-handling mitigations are described with their limits. Nothing here is investment advice. Decisions are yours.
Volume is the most manipulated number in crypto, and it is not close. Price at least requires capital to move; reported volume can be manufactured by a venue trading with itself at zero cost. Every liquidity figure, every "top exchange" ranking, and every volume-confirmed breakout you have ever seen inherits that pollution, which is why cleaning volume is not a data nicety. It is a survival skill.
Why crypto volume lies more than anywhere else
Three structural reasons. First, incentives: exchange rankings have historically keyed on reported volume, so a venue that printed fake trades bought itself listing fees, users, and credibility at the cost of a database entry. Second, absence of consequences: securities markets prosecute wash trading; much of crypto trades on venues where no regulator audits the tape, so the deterrent that keeps equity volume mostly honest simply does not operate. Third, cheapness: a venue or market maker matching its own buy and sell orders pays no spread, no fee it does not rebate to itself, and no risk. The landmark evidence arrived in 2019, when an analysis submitted to the US SEC concluded that roughly ninety-five percent of reported bitcoin volume on unregulated exchanges was not real, and subsequent academic studies of unlicensed venues have repeatedly found wash trading constituting large fractions, commonly cited above seventy percent, of their reported activity. The exact percentages move; the direction never has.
Who benefits, item by item
Follow the beneficiaries and the persistence explains itself. Venues inflate to climb rankings and attract listings. Token projects inflate, or hire "market makers" who inflate, because exchanges gate listings on volume and because traders screen for it: fake volume is advertising that looks like adoption. Market-making arrangements with volume commitments generate mechanical wash flow. And pump operations inflate deliberately as bait, because a volume spike is the costume of a breakout, the exact signature momentum traders and, note well, momentum SCREENERS are hunting, per the anatomy in the Telegram assessment. The victims are symmetrical: anyone using volume to gauge tradability overestimates their exit, and anyone using volume to confirm strength is confirming a costume.
Detection heuristics anyone can apply
You cannot audit a venue's tape, but you can triangulate, and four checks catch most of the theater. Compare volume to depth: real markets with high turnover carry order books to match; a pair reporting nine figures of daily volume with a thin book and a wide spread is reporting fiction, because real flow would eat that spread instantly. Compare venues: a token trading ninety percent of its "volume" on one obscure exchange while major venues show a trickle is telling you where the printer lives. Watch the shape: organic volume clusters around news, sessions, and volatility; manufactured volume runs metronomically through dead hours. And prefer medians over averages across time: one manufactured spike pollutes a 30-day average for a month, while a median barely moves, which is precisely why our own liquidity column uses the 30-day MEDIAN of dollar volume from vetted venues, per the monitor methodology.
What our pipeline does, and what it cannot
Three mitigations, stated with their limits. Vetted venues: our volume figures come from a published list of exchanges with regulatory exposure and independent scrutiny, which removes the worst offenders and cannot remove wash flow ON vetted venues, where it still occurs at lower intensity. Median liquidity: robust to spikes, blind to sustained inflation, a market maker printing steadily for a quarter defeats a median. Cross-checks in research: for the derivation program's backtests, liquidity gates use conservative floors and the universe rules exclude the listing-age window where manufactured volume is most intense, per the rules published in the universe methodology. The honest summary: we can filter volume's worst lies, not all of them, and any figure on our board should be read as "volume, after the cheap fraud is removed," never as ground truth. A provider who presents volume without that caveat has either not studied the problem or prefers you did not.
Why this matters doubly for momentum research
A momentum system that uses volume confirmation, and most do, is exposed twice. In backtests, fake historical volume qualifies fake liquidity: the strategy "trades" size in names whose real books could never have absorbed it, inflating results in exactly the thin names where wash trading concentrates. Live, volume-costume pumps trigger volume filters by design, adversarial selection where the filter attracts what it was meant to exclude. Both failure modes argue the same direction: volume belongs in a pipeline as a skeptically-cleaned INPUT, never as a trusted confirmation, and backtest liquidity assumptions deserve the same adversarial bounding we applied to survivorship in our own audit.
The reader's rule
Treat every crypto volume number as a claim by an interested party, because it is one. Where the claim matters to you, triangulate with depth, spread, venue distribution, and time shape; where you cannot triangulate, discount by default. And when a service, ours included, shows you a liquidity figure, the first question is the methodology question: whose volume, which venues, mean or median, and what would this number look like if someone were paying to fake it? The dirtiest number in crypto can still be useful. It just cannot be trusted naked. Decisions are yours.