Trading 101

/

November 17, 2026

Why NFT trading volume gets inflated: how wash trading works

Published estimates of NFT wash trading range from 0.12% of sale value to about 45% of all-time Ethereum volume, and two providers once read the same Blur data as 14% and 80%. This reference explains the mechanics, why volume-based marketplace rewards paid for circular trades, and what an honest volume metric would have to exclude.

Blog Image
★★★★★
CLAIM UP TO
$8,000 USDT
GET DEAL
★★★★★
GET 20% OFF
TRADING FEES
GET DEAL
★★★★★
GET UP TO
$30,050 USDT
GET DEAL
★★★★☆
No.1 DEX
VVIP LEVEL UP  
GET DEAL

Affiliate disclosure: the offers above and the links in this article are affiliate links. KY Solutions K.K., the operator of this site, may earn a commission when you sign up through them. We are an independent affiliate and not the official site of any company shown. 【開示】上記のオファーおよび本記事内のリンクはアフィリエイトリンクです。当サイト運営者(KY Solutions株式会社)は報酬を受け取ることがあります。当サイトは掲載各社の公式サイトではありません。

Ask how much wash trading there is in NFT markets and you can get answers that differ by a factor of three hundred. A peer-reviewed conference paper put it at 0.12% of sale value. An open-source analytics dashboard put it near 45% of all-time Ethereum NFT volume. For one marketplace in one month, two respected analytics providers looked at the same public blockchain and reported 14% and 80%. None of them was lying. They were measuring different things and calling the result by the same name.

This article explains the mechanics of NFT wash trading, why marketplace reward programmes made it rational rather than exotic, why the published estimates disagree so violently, and what a volume metric would actually have to exclude to be worth quoting. It also says plainly where the evidence is thin, including where the strongest-sounding numbers rest on undisclosed methods. This is a factual reference and does not constitute investment or financial advice.

What does an NFT volume number actually count?

A headline NFT volume number counts the sum of prices attached to transfer events that a marketplace contract recorded as sales, converted to dollars, over a period. That is all it counts. It does not count whether the buyer and the seller were different people, whether the price bore any relation to what anyone would pay, or whether the trade had an economic purpose other than producing the record itself.

This is a structural property, not a bug in any particular dashboard. A blockchain records that address A sent a token to address B and that value moved the other way. It does not record who controls A or B. Creating a new address costs nothing and requires no identity check. So the gap between volume and demand is not a measurement error that better software fixes. It is the gap between what the ledger observes and what the reader assumes it observes.

The same gap appears wherever a public metric is cheap to manufacture. The DeFi version of the problem is total value locked, where the same deposited dollar is counted at several protocol layers. NFT volume is the harder case, because in DeFi the inflation is usually double counting of real capital, whereas in NFT markets the underlying trade can be entirely circular.

How does NFT wash trading work mechanically?

The basic mechanism is a sale between two addresses under one person's control, repeated. The operator funds address B from address A, lists an NFT from A at a chosen price, buys it from B, and can then reverse the trade. Value moves out and back. The ledger now holds a sale at that price, and the collection's volume, floor history and last-sale price all move.

Published detection work groups the observable patterns into a few families. The large-scale study The Dark Side of NFTs (arXiv 2312.12544) defines three. Round-trip trading is a closed cycle in the transfer graph, which the authors detect with depth-first search and a threshold of at least ten back-and-forth trades. Unprofitable trading is a sale whose price is matched by a separate transfer of funds between the same parties inside a short window, 20 minutes for Ether and 80 minutes for tokens, which suggests the money was refunded. Hidden trading is three or more consecutive private trades, flagged through the marketplace's own private-sale field.

Nansen's commercial filter, as described in Unchained's April 2023 write-up, uses six flags: self trades, circular trades, same-funder trades, flash loan trades, bounce trades and multi-trade premiums. The open-source filter maintained by the pseudonymous analyst hildobby uses four: buyer and seller are the same wallet, an NFT is traded back and forth between two wallets, a wallet buys the same NFT three or more times, and the buyer's and seller's wallets were both first funded by the same wallet.

Two things are worth noticing about every one of these lists. Each rule is a proxy for control, not an observation of control. And each rule has a threshold that the analyst chose. Nansen's own conclusion, reported by Unchained, was that wash trading "occurs across a relatively low number of trades at a very high volume". The detectable patterns are a small share of trades and a large share of dollars, which means the headline percentage is extremely sensitive to where the thresholds sit.

Why did marketplace reward programmes make this rational?

Because for a period in 2022, a wash trade paid. LooksRare launched on 10 January 2022 with a token-incentive design that distributed LOOKS pro rata by trading volume to both the buyer and the seller of every eligible trade. According to the marketplace's own documentation, a user's share was their trading volume divided by total platform volume, multiplied by the day's LOOKS allocation. That allocation was 2,866,500 LOOKS per day for the first 30 days, then 1,361,587.50, then 537,468.75, then 286,650 across later phases.

Trading against yourself under that rule costs the platform fee and gas and returns tokens proportional to the notional you print. The Dark Side of NFTs records a concrete instance: Bored Ape Yacht Club number 4937 sold at 91 Ether, and "both the seller and buyer received around $3.5K worth of LOOKS". The paper also notes that during the January 2022 peak for the Meebits collection, nearly 93% of that collection's transactions occurred on LooksRare.

Volume rewards were not the only incentive. The Congressional Research Service, in report R47189 of 20 July 2022, notes that OpenSea required a minimum trading volume to verify a collection, and that this "may present challenges for new artists or collections and thereby incentivize users to engage in wash trading". That is a government analyst identifying a marketplace feature, not a token programme, as a wash-trading incentive.

The honest footnote on LooksRare is that when it ended Trading Rewards on 1 October 2023, its published reason was not wash trading. The stated reason was that "the resulting daily sell pressure has been significant". The incentive link between volume rewards and wash trading is well evidenced, but the decision to end the programme was publicly justified on token price grounds, and reading it as an admission would overstate what the source says.

Blur took the opposite design route, rewarding listing and bidding behaviour rather than raw volume across its airdrop seasons. Decrypt characterised this as "rewarding liquidity" instead of volume. We flag this as the weakest-sourced claim in this article, because the detailed accounts of Blur's points design that we could find are commercial blog posts rather than primary documentation or independent research.

How much NFT volume is wash trading?

Nobody knows, and the published range spans more than two orders of magnitude. What follows is every estimate we could verify to its original source, with the method attached, because the method is the number.

Published estimates of NFT wash trading, with method and sample
Source and date Headline figure Sample and method Source type
von Wachter and others, Financial Cryptography 2022 (arXiv 2202.03866) 3.93% of addresses, 2.04% of sale transactions, up to $149.5m of inflated volume 52 largest collections by volume, January 2018 to mid-November 2021. Authors call it a "lower bound estimate" Peer-reviewed conference paper
The Dark Side of NFTs (arXiv 2312.12544) $8,857,070.41, "around 0.12% of the total trading amount ($7,216,023,387.61)" 2,701,883 NFTs across 285 collections, 3,830,141 sale events, 902,571 addresses, 28 June 2021 to 20 June 2022. Strict cycle, refund and private-trade tests Academic preprint
Chainalysis, 2022 Crypto Crime Report 262 probable wash traders, net $8,458,331 across the group Sales to "self-financed" addresses with a 25-sale threshold. Covers Ether and Wrapped Ether only, by the authors' own caveat Industry blockchain analytics
hildobby on Dune, reported by CoinDesk, 23 December 2022 Over $30bn, "almost 45%" of all-time Ethereum NFT volume, 58% of 2022 volume, 80% peak in January 2022 Four open-source heuristics covering self-trades, back-and-forth trades, repeat purchases and shared funding Open-source analytics, press-reported
CryptoSlam, cited by Bloomberg via CryptoSlate, 5 April 2022 95% of LooksRare volume, about $18bn Proprietary and undisclosed method Industry analytics, method not published

Two of these deserve a correction to how they are usually quoted. The Chainalysis figure is a profit number for a flagged cohort, not a share of market volume, and it is frequently repeated as though it measured the latter. The hildobby figures are a share of volume, not of trades, and the same analyst's Blur work found 11% of volume flagged against only 1.56% of trades. Those are answers to different questions and should never be set side by side without that label.

There is also a gap in the current record that matters. Chainalysis published a market manipulation study covering 2024 that estimated $2.57bn of suspected wash trading, described as "an upper bound estimate", across Ethereum, BNB Smart Chain and Base. That study covers ERC-20 and BEP-20 fungible tokens only and does not cover NFTs at all, so there is no recent authoritative NFT-specific estimate from a major analytics provider. Anyone quoting a current NFT wash-trading percentage is quoting something older than they probably realise.

Why did two providers report 14% and 80% for the same marketplace?

Because one of them counted market-making patterns as wash trading and the other did not. In February 2023, Decrypt reported that CryptoSlam flagged roughly 80% of Blur's volume as inorganic, nearly $824m out of $1.02bn since the 14 February airdrop, after updating its algorithm to catch a new class of trades by wallets showing "no regard for the metadata" of the collections they were trading. Over a comparable window, hildobby's method put Blur at about 14%. For OpenSea in the same period, CryptoSlam flagged roughly 2.5%, about $6.6m of $249m.

Both providers were reading the same public data. The divergence is definitional. A trader bidding mechanically on floor NFTs to farm points is doing something economically real, is bearing genuine price risk, and is also plainly not expressing a view on any particular token. Whether that is wash trading is a judgement, and CryptoSlam's chief executive said as much, describing the new criteria as involving "some judgment calls". Unchained's survey of the filters reached the same conclusion, calling them "subjective" and capable of distorting the market they describe.

What each detection approach catches and what it misses
Heuristic Catches Misses or over-flags
Buyer address equals seller address The crudest self-trade Misses anyone using two addresses, which is most operators
Shared first funder of both wallets Cheaply created address clusters Over-flags wallets funded from the same exchange or faucet. Misses operators who fund through a bridge or mixer
Back-and-forth and closed-cycle trades Repeated round trips on one token Threshold dependent. A ten-cycle rule misses a three-cycle operator entirely
Repeat purchase of the same NFT Reacquisition patterns in unique tokens Applied only to ERC-721 in the open-source version, so semi-fungible ERC-1155 editions fall outside it
Price-matched offsetting fund transfer Trades where the cash was quietly refunded Window dependent, 20 to 80 minutes in the published version. A patient operator waits
Metadata-indifferent trading Point-farming and mechanical floor sweeping Captures legitimate market making. This single choice explains most of the Blur divergence

None of these observes intent, and the serious providers say so. Chainalysis states the limit directly: "our approach tracks patterns of behavior and not intent, which means that it is not by itself sufficient to prove market manipulation". Every wash-trading percentage in circulation is a count of suspicious patterns, not a finding of misconduct, and treating one as the other is the most common error readers make with these figures.

Is NFT wash trading actually illegal?

In the United States, the specific statutes usually invoked do not squarely reach a spot NFT trade. This is a narrower answer than most commentary gives, and the primary texts are short enough to check.

Legal instruments commonly cited against wash trading and their actual scope
Instrument What it covers Reach over a spot NFT trade
26 U.S.C. 1091, the tax wash sale rule Losses on "shares of stock or securities", within 30 days before or after the sale A tax deduction rule, not a trading prohibition, and limited to stock or securities on its face
7 U.S.C. 6c(a), Commodity Exchange Act section 4c(a) Transactions "of the character of, or ... commonly known to the trade as, a 'wash sale' or 'accommodation trade'", and prices that are not "true and bona fide" Applies to commodities for future delivery, options and swaps, not to spot or cash transactions
Securities Exchange Act 1934 and Securities Act 1933 Fraud and manipulation in securities Bites only if the particular NFT is held to be a security, which is asset by asset and unresolved generally
Regulation (EU) 2023/1114 (MiCA), Title VI market abuse Market abuse for crypto-assets admitted to trading, with Article 88 prohibiting market manipulation Article 2(3): the regulation "shall not apply to crypto-assets that are unique and not fungible with other crypto-assets, including digital art and collectibles"
Wire fraud and conspiracy, 18 U.S.C. 1343 and 371 General fraud statutes The route actually used against crypto wash trading in 2024, and it turns on proving deceit and intent

MiCA's NFT carve-out is not unlimited. Recital 11 of the regulation states that "the issuance of crypto-assets as non-fungible tokens in a large series or collection should be considered an indicator of their fungibility", and that a unique identifier alone "is not, in and of itself, sufficient to classify it as unique and non-fungible". A ten-thousand-piece profile-picture collection with an interchangeable floor is exactly the case that recital describes, which is why the carve-out's practical edge is contested rather than clean. The same fungibility question drives the analysis in our piece on fractional NFTs as an investment vehicle.

Enforcement so far has landed on fungible tokens. SEC press release 2024-166 of 9 October 2024 charged ZM Quant Investment, Gotbit Consulting and CLS Global, plus nine individuals, over what the Commission called "market-manipulation-as-a-service" in crypto assets offered and sold as securities, stating that "the schemes were intended to induce investor victims to purchase the crypto assets by creating the false appearance of an active trading market for them". The parallel criminal action, which a law firm note from Mintz dated 15 October 2024 describes as covering 18 individuals and entities under wire fraud and conspiracy charges, was reported as the first criminal charges of its kind against financial services firms in crypto. None of it concerned NFTs.

The one piece of United States government text that addresses the NFT case directly is the Treasury Department's February 2022 study of money laundering through the art trade, which states that "NFTs can be used to conduct self-laundering, where criminals may purchase an NFT with illicit funds and proceed to transact with themselves to create records of sales on the blockchain". That is a money laundering typology rather than a market integrity rule, and the study's overall conclusion was comparatively mild, finding "some evidence of ML risk in the institutional high-value art market" and little evidence of terrorist financing risk. The institutional treatment of digital assets under rules such as the Basel III crypto capital requirements and the custody framework discussed in FDIC crypto custody approval assumes valuations that a manipulated trade history degrades, which is one reason this stops being an academic question once regulated balance sheets hold the assets.

What would an honest NFT volume metric have to exclude?

It would have to exclude six categories, publish the excluded amount separately, and state its thresholds. CryptoSlam already does the second of those, filtering flagged trades out of its headline figure and showing the exclusion as its own line, which is better practice than a single unexplained total.

Exclusions a defensible NFT volume metric would need, and the data each requires
Exclude Why What it requires Feasible on public data
Self-trades and shared-funder clusters No change of beneficial owner Address clustering with a published threshold Yes, imperfectly
Closed cycles and refunded trades Value returns to its origin Transfer graph plus fund-flow matching inside a stated window Yes, threshold sensitive
Trades priced far off the contemporaneous floor Notional is chosen, not discovered Time-matched floor series per collection Yes
Volume created while a reward programme paid for it The trade's return came from the incentive, not the asset Dated register of marketplace incentive programmes Partly, no public register exists
Aggregator and bundle double counting One economic trade recorded more than once Router and bundle decoding per marketplace Yes
Lending liquidations and internal protocol transfers Forced or mechanical, not demand Contract-level labels for NFT lending protocols Partly

Three honest limits remain even after all six exclusions. Beneficial ownership is unobservable, so clustering is always a guess. Point-farming and market making sit on a genuine boundary that no rule resolves, which is the Blur problem and it does not go away. And semi-fungible ERC-1155 editions escape the repeat-purchase test as currently published, which means the measured share is understated for collections that use that standard. A metric that states these three limits is more useful than one that reports a single clean percentage, in the same way that the metric-integrity problems we covered in ESG crypto funds and sustainable digital asset trends were less about measurement technique than about who chose the boundary.

How strong is each claim in this article?

Below is our own assessment, including the places where we are relying on weaker material than we would like.

Evidence strength of the claims made in this article
Claim Evidence type Strength
Volume counts recorded sale prices and cannot observe beneficial ownership Property of the data model, uncontested across all cited studies Strong
LooksRare paid LOOKS pro rata by volume to both sides of a trade Marketplace's own documentation, a commercial source and not independent Strong on the mechanism, since the issuer is authoritative about its own rules
Volume-based rewards drove a large wash-trading spike on LooksRare in early 2022 Academic preprint with a priced example, plus a government report citing an industry aggregator Strong on direction, weak on magnitude
Estimates of the wash-traded share range from 0.12% to about 45% Four separately published studies, each verified to its own source Strong
The 95% and 80% CryptoSlam figures Press reports of a proprietary, undisclosed method Weak. The number cannot be reproduced or audited by anyone outside the firm
No current authoritative NFT-specific estimate exists Absence of evidence, inferred from the stated scope of the 2024 Chainalysis study Moderate. An absence, and a study we have not found may exist
Blur's bid-based rewards were designed to resist volume farming One press characterisation plus commercial blog posts Weakest claim here. We could not verify the design intent against primary documentation
26 U.S.C. 1091 and 7 U.S.C. 6c(a) do not reach a spot NFT trade Statutory text read directly Strong on the text, moderate as applied, since we are not aware of litigation testing it on NFTs
MiCA excludes unique non-fungible crypto-assets from its market abuse regime Article 2(3) and Recital 11 of the regulation Strong on text, contested in application to large series
Our own proposed exclusion list would produce a trustworthy number Reasoning from the published methods, not tested Weak. It is a design proposal and nobody has run it

What should a reader do with a volume figure?

Treat it as a claim with a method attached, and ask for the method before using the number. Four questions settle most cases. Which filter was applied, and is it published or proprietary. Is the figure a share of volume or a share of trades, because the same dataset gave 11% and 1.56%. What period does it cover, given that the extreme readings cluster in a window when volume rewards were being paid. And does the provider publish what it excluded.

A provider that shows its exclusions and names its thresholds is more trustworthy than one reporting a lower number by an undisclosed method. The same scepticism applies to engagement and activity metrics on any platform that benefits from them being large, a pattern we examined in social trading platforms in 2026.

What else do readers ask about NFT volume?

Is NFT wash trading the same as the tax wash sale rule?

No, and conflating them is common. The tax rule at 26 U.S.C. 1091 disallows a loss deduction when substantially identical stock or securities are reacquired within 30 days either side of a sale, which is the mechanic behind the harvesting strategies discussed in AI tax optimisation and tax-loss harvesting. Market-manipulation wash trading is a different thing, about creating a misleading price and volume record. The Congressional Research Service lists them as separate items in its report on NFTs, alongside Commodity Exchange Act and securities law provisions.

Can you tell from a blockchain explorer whether a trade was a wash trade?

You can see suspicious patterns, not intent. Repeated transfers between two addresses, a shared funding source, prices far above the collection floor, and a quick reversal are all visible. Whether one person controlled both sides is an inference. Chainalysis frames its own method the same way, as tracking behaviour rather than intent.

Which marketplaces had the highest reported wash-trading shares?

On hildobby's open-source filter as reported in December 2022, LooksRare was 98% and X2Y2 was 87% of volume. CryptoSlam's proprietary filter put LooksRare at about 95% in April 2022 and about 80% of Blur in February 2023, while hildobby put Blur at roughly 14%. The marketplaces with volume-denominated token rewards consistently sit at the top of every list.

Did the problem go away when the reward programmes ended?

The specific incentive did. LooksRare ended Trading Rewards on 1 October 2023, and that removes the clearest profit motive for printing notional. What has not been established is the current level, because the most recent large wash-trading study from a major analytics firm covers fungible tokens only and excludes NFTs.

Does a high wash-trading share mean a collection is worthless?

It does not follow. Wash trading corrupts the volume and last-sale record. It says nothing directly about whether anyone wants the asset. The practical consequence is narrower and worse: the metrics a buyer would normally use to judge liquidity and price are the exact metrics that were manipulated.

Why do different NFT data sites show different totals for the same month?

Mostly because they filter differently, and some of them filter without saying how. CryptoSlam removes trades it classifies as wash trading from its headline figure, so its totals can sit well below raw on-chain or marketplace numbers for the same period. A difference between two sites is usually a difference in definition rather than a data error.

Which sources support this article?

No items found.
★★★★★
CLAIM UP TO
$8,000 USDT
GET DEAL
★★★★★
GET 20% OFF
TRADING FEES
GET DEAL
★★★★★
GET UP TO
$30,050 USDT
GET DEAL
★★★★☆
No.1 DEX
VVIP LEVEL UP  
GET DEAL

Stay ahead of the markets

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Read More