Blog
Blog
Visual temporarily unavailable
This section's visual is under review and a replacement is being prepared.
In 2019, Bitwise told the SEC that 95% of reported Bitcoin trading volume was fake. The study made the news, then the industry moved on. The fake volume is still there. The exchanges that report it are still in the top-20 lists that retail uses to pick where to trade. Here is how the wash trading still works in 2026 and what it does to anyone who builds a strategy on those exchanges' data.
This was not a fringe claim. Bitwise was a regulated asset manager. The SEC received their methodology. Independent analysts reviewed it. The estimate held up. The SEC rejected the ETF application on other grounds, not because the volume analysis was wrong.
Nineteen out of every twenty trades on unregulated exchanges did not actually happen. The tickers moved. The trade history updated. But no real counterparties exchanged assets.
The crypto market has changed since 2019. Major exchanges have gone through regulatory processes in various jurisdictions. Some have faced enforcement actions. The absolute percentage is probably not 95% across the whole market today. But the mechanisms that produced it have not disappeared. They have become more sophisticated, more distributed, and in some cases more profitable than the original crude versions.
This is about mechanisms. Not outrage or regulatory solutions. How volume is manufactured. How it survives in the data. What real volume actually looks like at the microstructure level, so you can tell the difference.
Wash trading has a simple definition: buying and selling the same asset between accounts you control, creating artificial transaction volume without changing your economic position. You start with X coins and Y dollars. After a wash trade, you still have X coins and Y dollars, minus any fees. The only thing that changed is the number that appeared in the exchange's volume counter.
In traditional financial markets, wash trading is explicitly prohibited in most jurisdictions. The Commodity Exchange Act in the United States, for example, has prohibited it since 1936. The prohibition exists because volume is supposed to convey information. It signals that a market is liquid, that prices are being formed through genuine supply and demand. Fake volume corrupts that signal.
In crypto, particularly on unregulated spot exchanges, the legal situation has historically been murky. Wash trading on a platform registered in Seychelles or Malta or the British Virgin Islands, with no connection to regulated markets, typically faces no enforcement. The legal cost of the practice has been near zero.
The economic incentive, on the other hand, is substantial.
Start with listing fees. When a token project wants to be listed on an exchange, it pays. Listing fees on mid-tier exchanges have historically ranged from tens of thousands to several million dollars per token. The primary metric that token projects use to choose which exchange to list on is reported trading volume. Higher reported volume means larger apparent audience, better price discovery, more real traders who might see the new listing. The incentive structure is clean: fabricating volume increases listing fee revenue. This has been documented in trade press reporting during 2018-2019.
Then there are the rankings. CoinMarketCap and CoinGecko both historically ranked exchanges primarily by reported trading volume. High rankings attract traders. Traders generate real fee revenue. So fabricated volume, if believed, converts into real revenue through the attention it commands. CoinMarketCap has since introduced confidence scores and adjusted volume metrics that attempt to filter out suspicious trading, but the underlying data still comes from the exchanges themselves.
The maker-taker fee model adds another incentive layer that is worth understanding precisely, because it creates a scenario where wash trading can be directly profitable, not just indirectly revenue-generating. On maker-taker exchanges, market makers who post resting limit orders receive a rebate per fill, and market takers who remove liquidity pay a fee. If an exchange controls accounts on both sides of a trade, it can structure those accounts to both receive maker rebates while simultaneously collecting the taker fee from one side. Depending on the specific fee structure, the net result per fake trade can be positive. The exchange earns money on volume it manufactured.
Finally, there is regulatory arbitrage. Exchanges operating outside enforcement jurisdictions face no legal cost. The question is purely commercial: does fabricating volume increase revenue more than it costs? The answer, for most of the 2019 period, was yes.
The abstract incentive becomes concrete in a few specific operational patterns. These are mechanisms, not accusations against particular named entities. The patterns are documented across academic literature and regulatory filings.
The most direct method: the exchange runs its own accounts on its platform and trades against itself. Account A posts a buy order. Account B, also controlled by the exchange, posts a matching sell order. The trade executes, appears in the public trade feed, and increments the volume counter. No money changes hands in any meaningful sense. The only external effect is the log entry.
Detection is possible but requires access to the right data. The timing between paired buy and sell executions is implausibly fast, faster than any human trader could respond to an open order. Trade sizes often follow mathematical ratios that do not appear in organic order flow. Most tellingly, Order Flow Imbalance stays near zero on wash-traded pairs. OFI measures the difference between aggressive buying volume and aggressive selling volume. Wash trades are symmetrical by construction: every fake buy has a matching fake sell. The net directional pressure is zero. Over any meaningful window, OFI on a wash-traded pair will be flat. Real markets are not flat; buyers and sellers take turns dominating.
A more sophisticated version involves two affiliated accounts that both post resting limit orders on opposite sides of the book and fill each other. This is structurally identical to internal wash trading, but both sides earn maker rebates rather than one side paying a taker fee. On some fee schedules, the combined rebate exceeds any spread or fee paid, producing a direct profit per fabricated trade.
Detection relies on the same price impact analysis: high volume with near-zero price impact. Kyle's Lambda, which measures price impact per unit of signed order flow, approaches zero on wash-traded pairs. If real, large-scale buying were occurring, it would move the price. Volume that does not move the price is not being driven by buyers who actually want the asset at that price.
Some exchanges display tight bid-ask spreads and deep orderbook depth that is not actually available. The displayed orders are withdrawn before they can be executed by large incoming orders. A trader looking at the orderbook sees apparent liquidity at a certain price. When they submit a large market order, that liquidity has moved. Execution occurs at a substantially worse price than the displayed orderbook suggested.
This is related to but distinct from pure volume inflation. The goal here is to attract traders with the appearance of a liquid, tight market, while the actual execution quality is poor. It inflates the apparent market quality rather than the volume directly. Detection requires comparing displayed orderbook depth with actual transaction prices on large orders. The effective spread, what you actually pay on a round trip, will be much wider than the quoted spread.
The most operationally complex version involves the exchange outsourcing the wash trading to external parties. The exchange sets up a fee structure where high-volume market makers receive significant rebates. The third parties run wash trading operations, earning rebates that exceed their minimal execution costs. The exchange receives a share of the economics through the volume-dependent revenues the fabricated activity enables.
This is harder to detect from trade data alone because the accounts are technically separate entities. Academic researchers have applied clustering analysis to account identifiers, timing correlation across trade sequences, and trade size distribution analysis to identify coordinated patterns. Several peer-reviewed studies have applied these methods systematically and found substantial wash trading patterns concentrated on unregulated exchanges.
The Bitwise 2019 SEC submission (Bitwise Asset Management, "Presentation to the SEC", March 2019) remains the most cited single analysis. Their methodology compared the claimed volume on 81 exchanges against observable characteristics of clean markets: price impact, orderbook depth consistency, trade size distributions, the relationship between volume and volatility. Clean exchanges showed all the expected properties. The exchanges flagging as suspicious failed on multiple dimensions simultaneously. The 95% figure came from aggregating the suspicious volume as a proportion of total reported volume.
What made the Bitwise analysis credible was the methodology, not the conclusion. They were not arguing from intuition. They tested specific, measurable properties of real markets against the data and found the data did not match.
Academic researchers have independently reached similar conclusions using different methods. Peer-reviewed statistical analyses have found that a substantial fraction of volume on unregulated exchanges is not consistent with organic trading. The methodologies have focused on identifying accounts exhibiting wash trading patterns through clustering and timing analysis.
Regulatory enforcement provides additional evidence: what regulators actually found when they had the authority and data access to investigate. The CFTC brought enforcement actions against several crypto derivatives platforms, including BitMEX (CFTC and DOJ, 2020). When regulators with subpoena power examined internal exchange records, they found practices that matched what external data analysis suggested.
The FTX bankruptcy proceedings revealed internal exchange practices that had not been visible from the outside. Documents from the proceedings showed that Alameda Research, FTX's affiliated trading firm, had operational relationships with the exchange that gave it structural advantages over other market participants, including access to customer funds and preferential treatment in exchange systems.
The consistent finding across these independent sources, which used different methodologies, different data sources, and different investigative purposes, is that a substantial fraction of reported volume on unregulated crypto exchanges has not been organic.
The academic methods are not entirely out of reach for a technically capable retail trader or developer. The core tests map onto data that is available or can be estimated.
Real volume moves prices. This is not an assumption; it is a consequence of markets having two-sided depth. When buyers are purchasing in size, they exhaust the available sell orders at each price level and push price higher. Volume that does not move prices is volume that is not expressing real demand or supply.
Kyle's Lambda formalizes this: it is the price impact per unit of signed order flow, estimated from the relationship between trade direction, trade size, and subsequent price change. A wash-traded pair will have Lambda near zero over any meaningful sample. Buyers and sellers are matched in fabricated trades, net signed flow is near zero, and price does not move in response to volume.
If you are evaluating an exchange, look at the relationship between volume and price over rolling windows on pairs you suspect. Compare it to the same pairs on exchanges you consider clean. The difference will be visible without requiring Lambda estimation.
OFI tracks the cumulative difference between buyer-initiated and seller-initiated volume. Real markets are directional. Buyers dominate for periods, then sellers dominate. The OFI time series shows this directionality.
Wash trading produces OFI near zero over any extended window. The fabricated trades are symmetric by construction. If you compute OFI on a high-volume pair and the time series is remarkably flat, consistently centered near zero regardless of market conditions, that is a red flag. It does not happen in organic markets.
Organic trading produces characteristic trade size distributions. Most trades are small. A few are very large. The distribution follows a rough power law. The tail is fat and irregular.
Wash trading often produces different distributions: suspiciously round numbers appearing with unusual frequency, repeating size sequences, uniform timing intervals between trades. Compare the histogram of trade sizes on a suspicious pair against the same histogram on a known-clean exchange. Systematic differences in the distribution shape are meaningful.
In organic markets, volume and volatility are positively correlated. When new information enters the market, both volume and price volatility increase simultaneously. Traders with different interpretations of the information trade against each other; price moves as the market incorporates the information; volume increases because both sides are active.
Wash trading produces volume that is uncorrelated with or negatively correlated with volatility. The fake trades happen at a steady rate regardless of whether any information event is occurring. Screen for pairs where high volume is associated with low volatility over time. That combination is not what organic markets produce.
These tests require trade-level data, not just OHLCV candles. Many exchanges make trade history available through their APIs. Some aggregate data providers compile cross-exchange trade feeds. The data exists. The analysis is not technically demanding.
Real volume has price impact. When a large buyer enters the market on a genuinely liquid pair, the orderbook depth gets consumed level by level and price moves. The bid-ask spread widens temporarily under pressure. The orderbook rebuilds as market makers reprice. All of this is visible in the data.
Real order flow has directional character. Buyers dominate for periods, sellers dominate for periods. OFI shows non-trivial excursions from zero in both directions. The market is having an actual argument about price.
Real depth is available when you try to hit it. On a genuinely liquid market, the displayed orderbook depth fills at approximately the displayed prices. Effective spreads match quoted spreads on all but the largest orders. The orderbook is not a fiction that withdraws when approached.
These are not sophisticated claims. They are basic properties of a functioning two-sided auction. A market where buyers and sellers have real opinions, transact at their own risk, and have no coordination on fabricating the appearance of activity will naturally exhibit these properties.
The detection methods above work because they test for these properties. An exchange with real volume will pass all four tests. An exchange with fabricated volume will fail at least one and usually several.
DepthSignal records the underlying microstructure data across exchanges at the trade level: Order Flow Imbalance, Kyle's Lambda, effective spread, orderbook depth. These are the metrics that make these detection methods practical without building your own data pipeline from scratch.
The 2019 Bitwise estimate of 95% fake volume was a snapshot of a particular moment in a young market with almost no enforcement and strong incentives to fabricate. The number is not 95% across the whole market today. But the question "is this volume real?" remains one of the most important questions you can ask before interpreting any crypto market data, and most retail tools give you no way to answer it.
The tools that do answer it are the tools built on actual market microstructure, not on volume counters.
Usually through wash trading, rebate-driven self-matching, affiliated accounts, or displayed liquidity that disappears when someone tries to hit it. The purpose is to make the market look larger and cleaner than it is.
No. The useful question is not whether all volume is fake. The useful question is whether the venue shows the properties of a real market: price impact, directional flow, persistent depth, and plausible trade-size behavior.
Check whether volume moves price, whether order flow imbalance stays unnaturally flat, whether trade sizes look synthetic, and whether displayed depth actually fills.
Yes. It can distort screening, liquidity assumptions, venue rankings, and the confidence traders place in breakout or momentum setups. False size is still dangerous if people make decisions from it.
Yes. The problem is often mixed, not absolute. A venue can have genuine participants while still inflating parts of the tape or displaying unreliable liquidity behavior.
Because it poisons the inputs behind strategy decisions. If the volume, depth, and spread story is false, then any signal built on that venue becomes less trustworthy.