Research Summary: NFT Wash Trading: Quantifying Suspicious Behaviour in NFT Markets


  • This paper tries to answer the question of how prevalent wash trading is in smart contract-based NFT markets on Ethereum.
  • Wash trading is defined as a set of trades, without taking market risks, that lead to no change of the initial position of the adversarial agents. As these trades are topologically closed directed cycles, the method used to detect them is by using the deep first search algorithm.
  • At least 2.04% of NFT transactions and 3.93% of all addresses show suspicious behavior.

Core Research Question

To what extent does wash trading occur in smart contract-based NFT markets on Ethereum and to which extent does this practice distort prices?


von Wachter, Victor, et al. “NFT Wash Trading: Quantifying suspicious behaviour in NFT markets.” arXiv preprint arXiv:2202.03866 (2022).


  • NFT: A non-fungible token that represents a unique digital or physical asset on a blockchain. In contrast to other tokens (e.g. ERC20), NFTs are not divisible and have a unique identifier.
  • Washtrading: A set of trades, without taking market risks, that lead to no change of the initial position of the adversarial agents.
  • Collections: Items traded on NFT marketplaces are organized in collections. These are sets of NFTs that usually share some common features.
  • Special characteristics of NFT markets: NFT markets differ significantly from traditional financial markets: Users connect to the market through wallets, thus on-chain trading does not impose extensive onboarding and identity requirements. As users can create an arbitrary number of addresses for free and remain pseudonymous, these characteristics make preventing illicit activities a challenging task.
  • Directed Multigraph: Each blockchain address is treated as a node and each transaction as an edge. Transactions are directional, thus the edges are well.


  • The subject of the paper is the occurrence of wash trading in smart contract-based NFT markets.
  • Wash trading is well-researched and usually prohibited in traditional financial markets. However, not much data exists for blockchain-based marketplaces.
  • The data analyzed comprises the 52 leading ERC721 NFT collections on the Ethereum blockchain.
  • In total, 21,310,982 transactions of 3,572,483 NFTs, conducted between 2015 and late October 2021 are analyzed.
  • These transactions are modeled as directed, cyclical graphs and analyzed using a Deep-First-Search-Algorithm.
  • Path trades with an unusually high velocity are treated as suspicious behavior. Additionally, clusters of addresses that trade with significantly fewer trade partners than others are suspicious as well.
  • The results indicate a high concentration of illicit activities around a few NFTs. Despite finding evidence for a significant amount of wash trading activity, less suspicious activity occurs than the authors initially estimated. (See the section below for further details).


  • Dataset: The dataset contains 21,310,982 transactions of 3,572,483 NFTs conducted by 459,954 addresses. Collectively, the dataset represents $6.9b of the $12.3b total trading volume (49.5%) on all NFT markets. The timespan analyzed reaches from the genesis of the Ethereum blockchain in 2015 to 10/2021.
  • Data collection method: Smart contract transaction data for the 4 largest marketplaces (OpenSea API, Foundation, Rarible and Superrare) is collected. The authors focus on the 52 leading ERC721 NFT collections on the Ethereum blockchain. This event data is parsed and enriched with USD prices from the coingecko API and blockchain data from Etherscan. The dataset is then pre-processed to eliminate technical errors or transactions with faulty data.
  • Analysis: The transaction history of each NFT is modeled as a directed multigraph G_{nft}=(N, E), where N is the set of addresses and E is the set of ERC721 transactions between addresses. The authors utilize the Deep-First-Search-Algorithm to identify closed cycles within the data set. Examples of closed cycles:


  • 3.93% of all addresses are involved in wash trading activities. These flagged addresses processed 2.04% of the total sale transactions, inflating the trading volume by $149.5m or 2.17%.
  • Of the 36,385 flagged sale transactions, 30,467 were conducted in clusters of cyclical patterns whereas 5,918 were conducted as a rapid sequence. 48.4% of all detected cycles happen within 1 day and 13.4% of all detected cycles happen within 7 days.
  • The elapsed time to close a cycle with respect to the number of transactions involved are shown in the figure below:
  • The dominating form of closed cycles are simple patterns with 2-leg (e.g. Alice to Bob and Bob back to Alice).
  • The suspicious activity was executed with just 0.45% of the NFTs in the dataset, indicating a high concentration of illicit activities around a few NFTs.
  • Further, through the analysis of the relationship between executed trades per address and unique trade partners per address, a cluster of suspicious addresses was identified that traded 25-37 times with only 12-17 unique trade partners.

Discussion and Key Takeaways

  • The authors argue that the amount of $149.5m and the median of 2.04% of suspicious sale transactions indicate that wash trading may be less common than previous estimations by industry observers.
  • However, the level of suspicious activity varies significantly across NFT collections.
  • The data shows that adversarial market participants prefer fast and simple cyclical patterns.
  • Age and sentiment are often more relevant to price discovery than illicit trading activities.
  • The results can be seen as a lower bound estimation for suspicious trading behavior on decentralized NFT markets.

Implications and Follow-Ups

  • The paper represents a solid theoretical contribution in the form of empirical statistics on fraudulent behavior in NFT markets.
  • An additional aspect worth studying is the incentivization of NFT trading volume through tokens (e.g. $LOOK) and the impact on price, trading volume, and suspicious behavior.
  • The correlation between sentiment data and suspicious behavior is another potentially interesting research topic.
  • Further, the authors suggest looking into the utilization of flash loans for NFT wash trading.
  • A limitation is that only operations conducted within the Ethereum Virtual Machine (EVM) were analyzed, so any “off-chain” transactions are not included in the dataset and thus out of scope.
  • Only Ethereum mainnet transactions were analyzed, other blockchain networks were out of scope - this represents a research opportunity!


  • The paper provides valuable insights for NFT collectors to prevent them from buying NFTs at potentially inflated prices.
  • The authors contribute to a deeper understanding of the prevalence of wash trading and discuss practical countermeasures for NFT marketplaces.

Thank you @f13r for this summary. Since Wash trading are involved in over 3.95% of the transactions carried out, are they practical steps to mitigate it in NFT transactions since it’s deemd an illegal trading?


How does the number compare with other recent papers? I remember SCRF ran another summary of a different study that also flagged huge wash trading on $LOOK but it seemed to find about 3% elsewhere. Does that feel right? In my own experience trading NFTs, it seems as though there’s plenty of undetectable backroom dealing which would suggest the numbers should be much higher. Curious how all of this has changed given the recent chill in the NFT markets too. Any thoughts? Are the bad actors moving elsewhere?


Hi there, According to an article on , there are signs to watch out when an NFT have been wash traded. These signs include the following:
Price: It’s possible that the NFT you’re looking to purchase has been wash-traded if its price is significantly higher than the collection’s floor price (the lowest price an NFT is selling for in a specific collection). This is especially true if the NFT in question has few to no rare characteristics that might justify a higher price point.

Transaction history: The transaction history of an NFT can be checked using programs like Etherscan and BscScan. These details are also displayed on the listing sites of some marketplaces, such OpenSea. Wash trading may be detected by a rapid, inactive increase in the price of an NFT.

Previous owners: A wallet address like CryptoPunk 9998 that appears several times in the transaction history should be avoided. If the same wallet has made several purchases of an NFT, wash trading may be taking place. Another possible indicator that the wallets may be closely related to one another is to check individual wallet addresses to see whether they have interacted with other wallet addresses included in an NFTs transaction history.

I believe that with those signs in mind, potential NFT buyers will be able to mitigate NFT washed- trade transactions.
@f13r , you can still add more if you have. I really enjoyed reading your summary.


Before @f13r comes in to take this question, I think that a good number of the wash trading happens at a peer to peer level. Anybody can create any amounts of wallets at ease and transact with oneself. This could be hard to control but not impossible.

Perhaps, NFT marketplaces could come in here by watching suspicious transactions to suspend them or deanonymizing wallets on a worst case scenario. But this breaches privacy anyway. In the end it gets complicated.


I believe that carrying put surveillance on fraudulent activities with the aim of curbing it will not amount to data protection and Privacy breach. However, caution must be taken why doing so to ensure that the principles of necessity and proportionality is complied with.

Deanonymisation on its own doesn’t amount to data breach, it only brings the data within the definition of personal data and thus requires best practices in handling it.


Thank you @Cashkid18 for this detailed response. However, i have some reservations on the mechanism that suggests that frequent wallet address should be avoided. What if the frequency of the wallet address is as a result of honest trading? In the event this mechanism is adopted, don’t you think it will introduce unnecessary suspicion on wallets that recur frequently and thus impact overall trading negatively?


Thanks @f13r for the nice summary, in this
research paper, I have just two comments;

a. the legality of wash trading with NFT.

b. Do you think that wash trading has really affected the NFT market so far?

In customary finance, wash trading is unlawful. The US was the main country to proclaim it unlawful in the 1936 Commodity Exchange Act, and presently, the training is illegal in pretty much all aspects of the world.

Nonetheless, in the decentralized universe of NFTs, the lawfulness of wash exchanging is yet to be obviously spelt out. While it is dishonest, there’s no regulation anyplace on the planet enumerating what is correct and what isn’t in that frame of mind as its resource class hasn’t been recognized.

Regardless of this noticeable shortfall of NFT guideline and class assurance, a few nations have stood firm on the training. For instance, Bithumb, a South Korean crypto trade, was blamed for permitting wash exchanging worth more than $250 million phony volume in 2018.

Despite the fact that digital money wash exchanging may be unlawful in certain nations, it is still difficult to distinguish the culprits on account of the decentralized idea of cryptographic forms of money.

In my opinion, Wash trading with NFTs doesn’t seem to have affected the NFT market so far. Regardless, the market is showing improvement over and over.

OpenSea, the biggest NFT commercial center on the planet, outperformed $3.5 billion month to month exchanging volume January 2022. While NFT exchange volumes keep on developing in spite of the tricks, Chainalysis’ report uncovers that most wash trading didn’t yield however much the work that went into them.


Hi there,
I think what this scenario is trying to explain is that if you notice the same wallet address have purchased a unique NFT multiple times, then you should take it as red flag.

Notice that I was specific with the same wallet address buying a particular NFT many times.

A case of the same wallet address buying different NFTs many times shouldn’t be taken as a sign or a red flag for NFT wash trading.

I’m still open to suggestions and corrections though.


great summary work @fl3r really learned a lot about wash trading, it even led me to other illicit practices with NFT’S .
i have a question are there ways though, will like to know if there are ways to detect wash trading ?


Exactly, the act in itself doesn’t cause the breach but the means.


Hi @GloriaOkoba, pleasant perception, I believe that where there is a wrong, there is a remedy. I think scour could be a solution to wash trading. Scour is a unique product created by BitsCrunch, an AI-powered analytics company that can actively remove wash traders from exchanges. I think bitsCrunch is the Guardian of the NFT ecosystem.

The Scour index compiles transactions, wallet addresses, and distributions of reward tokens. Using knowledge graphs and AI technology, Scour can identify bogus orders and complex patterns of wash trades. AI-Enhanced Safety Feature (SCOUR),

I don’t know whether my remark has addressed your inquiry. if not, i believe @f13r would do justice to your question.


Thank @Henry for this wonderful suggestion. AI are known to be biased sometimes especially where it is so programmed and no human intervention. Is there a way to detect if such AI is programmed to be bias to certain wallets addresses?

1 Like

Hi @Samuel94 Nice observation, yes, I agree with you that occasionally AI could be biased I think the explanation is on the grounds that External Audits are Challenging because of Privacy Regulations. Don’t worry let me give a context.

in scenario where AI applications are used in high-stakes environments, many believe that external audits should be used to systematically vet algorithms to detect potential biases. This may be an excellent idea but often privacy is an issue. So, to thoroughly evaluate an algorithm one needs not only access to the model but also to the training data. But companies cannot share the customer data they use to develop models, as they need to comply with GDPR, CCPA, and other privacy regulations.

I believe that the EU’s AI Act (still in draft form) will help, as it requires organizations to use fair training data and ensure that their AI algorithms don’t discriminate. More also, the Equal Credit Opportunity Act stops any creditor from discriminating against any applicant from any type of credit transaction based on protected characteristics. I equally hold the view that companies and other operators should guard against violating these statutory guardrails in the design of algorithms.

In addition, Public pressure can play a role in persuading companies to make AI fairness a priority.

I am of the view that algorithm could be a strategy for detecting and possibly curing intended and unintentional biases in specific wallet addresses. I hope I am able to address your concern? I am open to learn if there is contrary view on this.

1 Like

In summary @Henry you believe AI should at all time have human interventions to check its biases?

As part of today’s coffee house discussion hosted by @jringo it came to my attention that an implication can be drawn from the conclusions of the paper that is simply too big to ignore.

A very interesting point was that the final percentage of wash-trading (~2% of total Txs) seems suspiciously low given people’s experience around NFT project communities and the overall permissionless and trustless nature of web3 with the motto “Don’t Trust, Verify”. There seems to be a suspicion about the final numbers which goes against the common narratives, journalistic work, and first-hand experiences of web3 participants. It is very interesting to investigate this sentiment because it hints at the fact that the conclusions of this paper might be critical for howe we think about the social dimension of web3.

Interpreting the results

The conclusions lead to the following possibilities:

  1. False negative: The number of wash-trading is misleading and is an artifact of a methodological fluke in the heuristics used for representing wash-trading (or other methodological parameters).
  2. Folk-psychological bias: The number of wash-trading represents the underlying reality and the discrepancy is due to erroneous prior intuitions and underlying assumptions about human behavior on web3.

False Negatives and Methodological Hurdles

If 1 is the case, (as the authors are careful to point out as a possibility) it means that additional studies are needed to form clear hypotheses based on the statements and conclusions of the current paper and find where the contention lies. The most obvious culprit is that the heuristics used such as closed loops and high transaction frequency might be either too constraining or not suitable. There is no reasoning in this paper as to why these two heuristics were chosen in the first place although it seems that the quantified representation originates from (Das et al, 2021). But as mentioned in that original literature, they were the first to construct a quantified model of malicious behavior in the context of NFTs and not enough research exists on the specific ways in which wash-trading actually takes place in this context. I would like to ask @f13r whether they know if any of the quantitative work is based on ethnographic analyses on how this behavior actually takes place and if such ethnographic studies have been specifically conducted on some of the communities behind the NFT projects and marketplaces chosen for the current study. If not, then that presents a very important methodological hurdle because the current methods of quantification might be failing to account for real behavior in NFT communities. NFT communities in the context of degen culture have unique coalition dynamics and collusion patterns and probably engage in wash-trading in ways that have not previously been considered. For example, there is no clear explanation as to why round-trip trades modeled as closed-loop graphs might be a suitable heuristic for identifying wash trades in NFT communities. If @f13r or any of the original authors would provide more in-depth reasoning, as to why this is it would be much appreciated.

Folk-psychological Bias, Bad Faith and Views of Human Nature

If 2 is indeed the case, then this hints at a very interesting sociological implication. On the one hand, the default stance in web3 seems to be that fraudulent behavior is expected and doesn’t constitute that big of a problem since culturally, a lot of strain is placed on individual responsibility in terms of security, vigilance and risk. The default stance in terms of the question of human nature and behavior in a social setting seems to be one of acceptance of greed and opportunism. The design of blockchain validation systems in PoW systems that use greed and personal profit as incentives to maintain the integrity and truthfulness of transactions seems to presuppose such an underlying philosophical presupposition.

Tactically building systems that utilize the bad faith and opportunism of users for producing outcomes that serve a greater good (from securing a network to reducing carbon footprint) is the underlying presupposition that also informs a great deal of innovation in Ethereum and other blockchains with smart contract functionality (Cosmos, Celo etc.). But if the conclusion of the paper is legitimate, it means that the pessimistic intuitions about human nature that inform the game-theoretic design of the infrastructure of web3 are based on a wrong assumption. This is big and cannot be overstated.

It seems that because of the way bad faith behavior seems to occupy the forefront of the space’s attention and because of the inherent cognitive bias to overestimate the negative consequences of certain actions, our underlying perception of how humans act in the anarchic and permissionless environment of web3 is clouded.

The possibility that such an interpretation of the conclusions is even possible means that the study cannot stop here. Having been trained in the history of science, simple implications such as this are often enough to append whole established ways of thinking and follow-up research will be needed to address the issues raised. The original researchers such as @f13r will need to address the possibility of methodological errors or replicate given an updated data set including additional sources and heuristics and if no statistically important discrepancy is observed then a follow-up to address the philosophical issues of morality and design based on the empirical evidence collected would be important.


@cashkid18 can you throw more light on one of the signs of wash trading ; which is Price .

Yes @Samuel94 very correct and a very great observation it will mostly give away peoples data.

@f13r Thank you for this appropriate wash trading summary, which provides information for NFT collectors to avoid buying NFTs at potentially inflated prices.

Hi there,
In NFT, there is what is known as Floor Price and this is the NFT’s lowest cost or minimum price within a collection. It stands for the least amount a buyer must invest to acquire an NFT within a particular project.

In most cases, two groups of people decide on an NFT’s floor price.
First, by the owners of an NFT project when it is still in its infancy that is when it’s newly minted.
Furthermore, by NFT owners who list it on a secondary market.

Both buyers and sellers of NFTs in the market place understand the importance of using floor price as a criterion for determining the worth of an NFT.
The NFT collection is considered to be more valuable when the floor price is also higher.

Now in NFT, floor price manipulation occurs and this is to make the floor price to go higher than it’s worth. A sign of floor price manipulation is likely when ‘sweep the floor’ happens and this is When all of the digital items in a collection are bought.
It is considered a sweep the floor when project owners buy every item in their collections for the floor price.

In a similar way, when purchasers “sweep the floor,” they have acquired all of the collections offered by that project.

The demand can be made up to give the appearance that an NFT is more valued.
This occurs when a new NFT project is strongly purchased by an individual or group.

This is especially true for NFT projects that have recently launched.

These NFTs will be sold by the buyer at a higher floor price following the sweep.

In order to prevent floor price manipulation and NFT wash trading, it is very essential to understand how NFT projects’ floor pricing are established before investing in NFTs to avoid losing your money.