The integrity of financial markets is a cornerstone of economic stability. Fraudulent actions by a single trader or a coordinated group can cause significant disturbances, directly influencing asset prices or misinforming other participants. Such behavior is a source of systemic risk and erodes trust, highlighting the urgent need for effective countermeasures.
Building on existing literature, this study designed an agent-based market model to replicate the behavior of the Bitcoin market during an alleged price manipulation event between 2017 and early 2018. The model incorporates a limit order book market and various agents with distinct trading strategies, including a fraudulent agent initialized from empirical data. The simulation results provide a compelling fit to historical data, suggesting that the presence of this fraudulent actor was essential for the unprecedented price peak in late 2017.
The Rise of Cryptocurrencies and Associated Risks
Cryptocurrencies represent a digital alternative to government-issued fiat money. Instead of a central authority, their implementation relies on cryptographic principles to validate transactions and generate new currency. Every transaction is recorded on a public ledger known as the blockchain.
While blockchain technology fosters innovation across sectors like supply chain management and finance, its application as a replacement for standard currency remains contentious. A significant issue surrounding cryptocurrencies is the extreme volatility and heavy tails of their return distribution, leading to a history of notorious "bubbles."
Traditional markets have built trust over decades through government institutions, robust legislation, and effective oversight systems. Each case of market abuse has historically led to a better understanding of vulnerabilities and the development of countermeasures. Cryptocurrencies, however, are still in their infancy. New methods are required to establish a reliable and fair market environment that maintains the decentralized nature of blockchain technology.
Common Frauds in Crypto Markets
Several illicit activities are linked to cryptocurrencies, including black-market trading, money laundering, and terrorist financing. This study focuses on fraud that deliberately targets and disrupts the market itself.
A prevalent form of market fraud is wash trading, where a single entity acts as both buyer and seller to create a false impression of high trading activity and mislead investors. A more serious form is the pump-and-dump scheme, which involves coordinated actions to artificially inflate a market's price in a short period.
Studies have attempted to explain price movements as direct consequences of manipulative behavior. For instance, one analysis of suspicious activity on the Mt. Gox exchange concluded that fraudulent actions influenced Bitcoin's price growth from $150 to $1000 in late 2013. More recently, research has argued that the Bitcoin market price might have been significantly inflated by the issuance of Tether.
As investors increasingly include crypto-assets in their portfolios and major companies accept Bitcoin payments, its volatility has become a potential source of systemic risk for the broader economy. While advances are being made in detecting wash trading and pump-and-dump schemes, new models are needed to simulate, predict, and test the effectiveness of policies against such fraudulent behavior.
The Power of Agent-Based Modeling
Agent-based models aim to explain complex phenomena where macro-level behavior emerges from micro-level rules. This paradigm has been enhanced by modern data-driven approaches, where behavioral data specific to each agent is used to construct their decision mechanisms. This increases a model's validity and predictive power, rivalling traditional quantitative methods in economic research.
These models are particularly useful when individual agent parameters are vital, such as testing interventions during a global pandemic. In crypto markets, agent-based models have been created to study widespread cryptocurrency acceptance, speculative trading, and price stabilization mechanisms.
A key strength of these models is that they provide an experimental sandbox for policymakers. Once a behavioral scheme is identified and its consequences are understood, the simulated environment can test the effectiveness of various regulatory measures, monitoring systems, and enforcement mechanisms.
This Study's Contribution
Most research focuses on the statistical relationship between price and a set of exogenous variables. This study, however, prioritizes the qualitative explanation dimension. It constructs a data-driven model focusing on the causal influence of specific fraudulent behavior that allegedly inflated the Bitcoin price.
This approach allows for a broader analysis of the fraudulent trader's role. The market model generates data such as price, volume, and the fraudulent trader's Bitcoin inflow, enabling comparison with empirical data. The model suggests that certain anomalies in market volume and price dips can be attributed to the fraudulent trader's actions.
Furthermore, the model investigates the reasons behind the manipulation's success, exploring the connection between the efficiency of a specific strategy and transaction costs. This required implementing a realistic model of order book liquidity, proposing a new distribution model based on a mixture of two components.
The Alleged Tether-Based Manipulation Scheme
This section elaborates on the alleged price manipulation using Tether in 2017/2018, presenting the technology, its ecosystem, and considerations from relevant literature.
What is Tether and Why is it Controversial?
Tether (USDT) is a cryptocurrency pegged to the US dollar, making it a stablecoin. Its purpose is to facilitate transactions between cryptocurrency exchanges, which often face challenges in establishing traditional banking relationships. Tether Limited, its issuer, claims every USDT is backed by one US dollar held in reserve, publishing monthly attestations to prove this.
This claim has been controversial. A 2019 study pointed to suspicious auditing methods, suggesting the issuer could leverage the time between audits to issue more Tether than its actual reserves. Following investigations and a lawsuit from the New York Attorney General, Bitfinex and Tether agreed to an $18.5 million settlement in February 2021, with the Attorney General stating Tether had misrepresented its reserves.
The Manipulation Scheme Explained
The scheme's possibility arises from the ability to push unbacked Tether into the market, falling into the pump-and-dump category but with a more powerful profit generation mechanism.
The strategy relies on the assumption that the market will respond with positive feedback (an inflow of buy orders) to the large Bitcoin buy orders executed by the fraudulent trader. Once a positive price trend is established, the trader's cash buffer can be refilled to satisfy the monthly attestations.
The profits are generated in two ways:
- The value of the Bitcoins the fraudulent trader already possesses increases by triggering an inflow of new buyers.
- If the price increases sufficiently, the trader can sell a smaller amount of Bitcoin for dollars than the amount bought with Tether to cover the attestations, resulting in a surplus of "free" Bitcoin.
A plausible selling strategy involves pumping the price as high as possible and then liquidating a sufficient amount via a sequence of small sell orders just before the attestation date. This approach takes advantage of high liquidity near the current price while being harder for the market to detect, preventing a drastic price drop.
👉 Explore advanced market analysis tools
Observed Volume Anomalies
The original study concluded that Tether flows from suspicious addresses were correlated with price increases. This analysis extends to trading volume and influence on other traders.
Evidence suggests the fraudulent trader sold unlawfully obtained Bitcoins to satisfy the attestations, causing temporal spikes in traded volume. These spikes, believed to be liquidation events, often corresponded to the end-of-month statements published by Tether Limited around the 15th of each month, particularly in July, September, November, and January of the studied period.
Additional large volume anomalies were hypothesized to be the market's response to price increases or decreases caused by the fraudulent trader's actions, possibly from other investors entering or exiting the market.
Building the Simulation: Exchange and Agent Models
The model's granularity is a limit order book where orders are placed publicly. To simplify processing, orders can be entered every minute, meaning each trading day consists of 1440 intervals.
The Limit Order Book Market Model
The market environment is based on an established model where traders observe the order book—a table of order types, amounts, limit prices, and expiration times. Buy orders are sorted in descending order by limit price, and sell orders in ascending order.
An order-matching mechanism executes trades when a sell order's price is less than or equal to a buy order's price. The market price is updated based on the last match at each time interval.
Modeling Liquidity: Price and Amount Distributions
A crucial property of an exchange is its order book depth, defined by the distribution of Bitcoin amounts and limit prices. This model hypothesizes four main empirical properties:
- A broad, hump-shaped (bimodal) distribution of limit prices.
- Quickly rising transaction costs.
- A relatively small volume concentrated near the mid-price.
- General symmetry between the buy and sell sides of the order book.
The bimodal shape is modeled as a mixture of two distributions: a Gaussian distribution near the mid-price and a Beta distribution for the tail of the limit price distribution further away. The amount distribution is constructed to reflect traders' bias towards "round" values (e.g., 0.5, 1, 1.5).
The Trading Agents
The success of the fraudulent scheme depends on the market's response. The model includes several agent types to simulate this:
- Random Agents (RAs): Issue buy or sell orders with equal probability, with limit prices sampled from the Gaussian component near the mid-price.
- Random Speculative Agents (RSAs): Similar to RAs, but their limit prices are sampled from the Beta distribution, meaning their orders are placed further from the mid-price, speculating on market volatility.
- Chartist Agents (CAs): Observe the average Bitcoin returns over a recent window. If the average return is positive, they issue a buy order; otherwise, a sell order. They become more cautious after the price reaches certain thresholds.
The Fraudulent Agent (FA)
The fraudulent agent's behavior is defined by buying and selling schedules. The buying schedule is constructed directly from empirical data on Tether outflows from identified suspicious addresses. A "cash matrix" defines the amount of capital the FA uses to issue buy orders on a given day and minute.
The selling strategy involves liquidating a portion of Bitcoins to refill the cash buffer for the monthly attestations. The FA sells small amounts every minute via limit sell orders. The amount planned for each day is calculated from empirical traded volume data, normalizing the spikes observed on specific dates.
Incorporating Large Scale Events (LSEs)
Volume anomalies not explained by the FA's actions are treated as exogenous Large Scale Events (LSEs). These are incorporated into the simulation as prior knowledge, with increased trading activity arranged on specific dates to reproduce the observed volume spikes, with parameters set to match the magnitude of the empirical anomalies.
Simulation Results and Analysis
To demonstrate the essential influence of the fraudulent agent, four simulation experiments were conducted:
- Base Scenario: Only random and random speculative agents are active.
- Susceptible Scenario: Includes chartist agents (CAs).
- Susceptible Scenario with LSEs: Includes both CAs and large scale events.
- Manipulated Scenario: The fraudulent agent is active alongside all other agents and events.
The results were clear. In the non-manipulated scenarios, the market generally remained in equilibrium or exhibited well-behaved fluctuations. Even with speculative agents and external events, the probability of the price reaching the historical heights of late 2017 was exceedingly low.
In the manipulated scenario, however, the introduction of the fraudulent agent successfully created a price bubble that matched the empirical data. The model's generated market price, traded volume, and Bitcoin inflow to the fraudulent agent all showed a satisfactory fit to historical records.
The simulations showed that the FA's actions directly explained several price dips and volume spikes that coincided with the end-of-month attestation dates. The model also demonstrated how the FA could generate a surplus of Bitcoin simply by executing the scheme, effectively having other market participants foot the bill.
The Critical Role of Market Liquidity
The study found that liquidity—specifically, its distribution—is a strong predictor of the manipulation scheme's success. Not only the total amount of liquidity but also how it is distributed in the order book is crucial.
In illiquid markets with low agreement on price, orders are dispersed further from the mid-price. This dispersion creates favorable conditions for manipulation. The study tested this by adjusting a model parameter (α) that controls how close to the mid-price orders are placed. As α increased (simulating more orders concentrated near the mid-price), the efficiency of the manipulation strategy decreased, resulting in lower inflated prices and a higher rate of scheme failure for the FA.
This finding suggests that measures promoting better price agreement and liquidity concentration near the mid-price could make markets more resistant to this type of manipulation without necessarily reducing overall trading activity.
Implications for Regulation and Market Integrity
The economic understanding provided by this model has significant implications for the cryptocurrency market. A key regulatory recommendation is that stablecoin providers should be required to prove their capital reserves more frequently than once a month to protect customers and prevent market manipulation.
Policymakers are gradually catching up. The European Union's "Regulation on Markets in Crypto Assets" proposal and proactive steps from the U.S. Biden administration target stablecoin regulation. Beyond legislation, self-regulatory approaches can be adopted, such as constraints on maximum order sizes or the number of orders per trader, similar to mechanisms in FOREX markets. More invasive interventions like circuit breakers (trading halts) could also be considered.
A dynamic, less restrictive approach could involve real-time market surveillance systems. Exchanges could monitor liquidity distribution and predict the market impact of large orders, refusing those suspected of aiming to create sudden price swings. They could also search for fraudulent behavioral patterns, such as periodic spikes in traded volume followed by liquidation on different exchanges.
The immutable and public nature of blockchain transactions offers a unique advantage for such monitoring systems, providing a full transaction history for analysis. This opens innovation potential for sophisticated AI models trained on historical or simulated data to oversee market behavior.
However, a significant challenge is the lack of incentive for exchanges to combat fraud aggressively. The short-term benefits of high trading volume may outweigh the long-term benefits of a stable, reliable market. There is evidence that exchanges were likely aware of the 2017 manipulation but did not intervene, as the price increase contributed to Bitcoin's widespread popularization.
👉 Get insights into market monitoring strategies
Frequently Asked Questions
What is an agent-based model in finance?
An agent-based model is a computational simulation that creates a virtual market populated by autonomous "agents" programmed with specific rules and strategies. These agents interact with each other and the market environment, and their collective actions generate complex, emergent market behavior that researchers can study. It's a bottom-up approach to understanding how macro-level market phenomena arise from micro-level decisions.
How did Tether supposedly manipulate Bitcoin's price?
The alleged scheme involved issuing Tether that was not fully backed by US dollars. This Tether was then used to place large buy orders for Bitcoin on certain exchanges. This artificial demand created a positive price trend, which attracted genuine investors. The fraudulent trader then sold portions of their Bitcoin holdings for dollars at these inflated prices to create the illusion of reserves for monthly attestations, generating a surplus of Bitcoin and profiting from the inflated value of their holdings.
What are the signs of potential market manipulation in crypto?
Common red flags include:
- Wash Trading: A single entity buying and selling to create fake volume.
- Pump and Dumps: Coordinated efforts to inflate a price before dumping holdings.
- Spoofing: Placing large fake orders to create false supply/demand signals.
- Unexplained Volume Spikes: Large, periodic volume anomalies that don't correlate with major news events, especially around specific calendar dates (like monthly attestations).
Why is the Bitcoin market particularly susceptible to manipulation?
Bitcoin's susceptibility stems from its relative illiquidity compared to traditional markets and fragmented liquidity across numerous exchanges. This leads to low agreement on price among traders, causing orders to be dispersed widely in the order book. This dispersion allows large orders to have a greater price impact and makes it easier for malicious actors to create significant price movements without immediate arbitrage correction.
What regulatory measures could prevent such manipulation?
Effective measures could include:
- Frequent Audits: Requiring stablecoin issuers to prove reserves daily or weekly.
- Real-time Surveillance: Exchanges implementing systems to detect suspicious trading patterns and order book activity.
- Order Limitations: Setting caps on order size or frequency for individual traders.
- Enhanced Transparency: Mandating clearer reporting of trading volume and order book data.
- Cross-Exchange Cooperation: Creating mechanisms for exchanges to share data and identify coordinated manipulation across platforms.
Can this model be applied to other cryptocurrencies?
Yes, the fundamental principles of the agent-based model and the manipulation scheme are transferable. The model could be adapted to study manipulation on other cryptocurrency markets, provided there is sufficient empirical data to initialize the agents (especially any fraudulent actors) and validate the simulation results. The approach is particularly relevant for other illiquid assets or newer stablecoins.
Conclusion and Future Research
This study demonstrated that a data-driven agent-based model, incorporating a fraudulent actor with a specific price manipulation strategy, could recreate the Bitcoin price bubble of 2017-2018, which was otherwise statistically unlikely to occur. The model successfully explained quantitative phenomena like volume spikes and price dips, linking them to the fraudulent agent's need to liquidate holdings for monthly attestations.
The findings underscore the vulnerability of illiquid, nascent markets to manipulation and provide valuable insights for policymakers and regulators. The methodology highlights the promise of deep integration between distributed ledger technologies and artificial intelligence for market oversight.
Future research directions include:
- Incorporating full information from all addresses linked to the manipulator.
- Using detailed, real-order book data from the involved exchanges.
- Designing more sophisticated agents that can adapt their strategies based on market conditions and potential countermeasures.
- Expanding the model to simulate multiple exchanges simultaneously to better capture cross-exchange arbitrage and influence.
- Generalizing the model to serve as a testbed for evaluating the effectiveness of various regulatory interventions before they are implemented in the real world.