Wash trading: how it’s detected in volume data

Wash trading is placing trades that create the appearance of activity without genuine change of ownership or real economic risk — for example, the same party (or coordinating parties) buying and selling to itself. It inflates reported volume, and because volume is used to rank exchanges and judge liquidity, detecting it matters. This article explains how analysts try to spot it in volume data.

This is an educational, awareness-level explainer about detection and market structure. It is not a guide to performing wash trading, which is illegal in many jurisdictions and prohibited by regulators such as the U.S. Commodity Futures Trading Commission. It is not trading advice.

Why wash trading distorts the data

Reported trading volume is one of the most-cited numbers in crypto: it drives exchange rankings, “top pair” lists, and perceptions of how liquid a market is. If a venue or token can manufacture volume, it can appear more important, more liquid, and more legitimate than it is. That is the incentive. The result is that raw volume figures cannot be taken at face value, and a body of analysis — including a widely discussed 2019 study submitted to the SEC arguing that a large share of reported bitcoin spot volume was not genuine — has focused on separating real from fake activity.

The statistical fingerprints analysts look for

Wash trading tends to leave patterns that differ from organic markets. No single one is proof, but together they raise or lower confidence.

• Volume that does not match order-book depth. Genuine large volume needs a deep book of resting orders to trade against. If a venue reports huge volume but its order book is thin and its spreads are wide, the volume is suspect.
• Trade-size distributions that look unnatural. Real trades cluster around certain human and algorithmic sizes. Data providers examine whether trade sizes follow expected statistical distributions; sharp deviations — too many identical sizes, or too-round numbers — are a flag.
• Price impact that is too small for the volume. Real trading moves prices. Volume that is enormous yet leaves the price almost undisturbed suggests trades that are not consuming real liquidity.
• Timing regularity. Bot-driven wash trades can recur at suspiciously even intervals or in symmetric buy/sell pairs, unlike the burstier rhythm of organic flow.
• Volume/visit and volume/user ratios. Comparing reported volume to independent proxies for a venue’s actual user base can reveal figures that are implausible for the traffic.

Benford’s Law and digit analysis

One well-known technique borrows from forensic accounting. Benford’s Law describes the expected frequency of leading digits in many natural datasets — the digit 1 leads far more often than 9. Genuine trade and volume figures often approximately follow this distribution. Fabricated numbers, especially those generated by simple bots, frequently do not. Analysts test whether an exchange’s reported figures deviate from the expected digit distribution as one input among many. Like every method here, it is suggestive, not conclusive.

On-chain cross-checks

For assets and venues where settlement touches a public blockchain, analysts can compare reported trading volume against on-chain flow data. If an exchange claims vast trading but the corresponding on-chain deposits, withdrawals, and settlement activity are far too small to support it, the gap is informative. This cross-check has limits — much trading is internal to an exchange and never hits the base chain — so it is a sanity check, not a precise audit.

Wash trading versus legitimate high activity

A crucial part of detection is not flagging honest markets by mistake. Several legitimate activities produce high volume and rapid, repetitive trading that can superficially resemble wash patterns:

• Market making. Professional market makers continuously post and update buy and sell orders, generating large volumes of genuine, risk-bearing trades that provide liquidity.
• High-frequency and arbitrage trading. Firms exploiting tiny price differences trade rapidly and often, which is real economic activity even though the individual trades are small and frequent.
• Incentive and rebate programs. Maker rebates or trading competitions can encourage high volume that is real in the sense that ownership genuinely changes, even if the economic motive is the incentive rather than a market view.

Distinguishing these from wash trading is the hard part. The defining feature of wash trading is the absence of genuine change in ownership or real economic risk — the same beneficial owner ends up on both sides. Public trade data usually cannot prove that; it takes account-level identifiers that only the venue or a regulator holds. This is exactly why external detection reports probabilities, not verdicts.

Why this matters beyond exchange rankings

Inflated volume is not a victimless distortion. Newcomers judging where to trade may be drawn to venues that look busy but are thin and hard to exit. Tokens can appear to have organic demand that does not exist. Data products, indices, and even some research can be skewed if they ingest unfiltered volume. And because volume feeds perceptions of legitimacy, fabricated figures can prop up venues that would otherwise struggle to attract genuine flow. Understanding how detection works is therefore less about catching bad actors and more about equipping yourself to discount numbers that do not hold up — a defensive, awareness-level skill rather than an operational one.

Why detection is genuinely hard

• Real markets are messy. Legitimate high-frequency and market-making activity can superficially resemble wash patterns (rapid, repeated, small trades), so honest venues can be flagged by crude tests.
• Sophisticated actors add noise. Wash schemes can randomise sizes and timing specifically to defeat simple statistical tests.
• Data access is uneven. Detectors often see only public trade prints, not the account-level identifiers that would prove the same party is on both sides. Regulators and the exchange itself can see far more.
• Definitions blur at the edges. Incentive programs, maker rebates, and internal transfers can inflate volume without meeting a strict legal definition of wash trading.

Because of this, credible analyses combine several signals and report confidence levels rather than declaring a single number “fake.” Treat any claim that rests on one metric with caution.

What this means for reading volume

• Do not treat headline volume as truth. Prefer sources that weight or filter volume by liquidity and order-book quality.
• Look at depth and spreads, not just volume. A market that is genuinely liquid shows it in a deep book and tight spreads, which are harder to fake than a volume number.
• Prefer corroborated figures. Metrics cross-checked against on-chain settlement or independent traffic estimates are more trustworthy than raw self-reported totals.
• Remember the limits of detection. Absence of a flag is not proof of honesty, and a flag is not proof of guilt; these are probabilistic signals.

What this means

Wash trading detection is a forensic exercise: because a fabricated volume number is easy to print and hard to prove false, analysts look for the fingerprints it leaves — volume that outstrips order-book depth, prices that barely move, unnatural size and timing patterns, digit distributions that break Benford’s Law, and on-chain settlement too small to match the claim. None of these is conclusive alone, which is why serious analysis stacks several and reports confidence rather than certainty. The practical takeaway is humility about volume figures generally, the same discipline that applies to on-chain flow data and to telling apart the spot and derivatives layers that generate those numbers.