Wash Trade Type 3 Detection Test
Description
This test uses machine learning algorithms to identify groups of users engaging in coordinated trading activity to artificially inflate market volumes and manipulate asset prices. These users trade within a closed network, executing a high number of trades with significant volume in a short time frame, often exceeding a set percentage of the Average Daily Volume (ADV). This manipulation creates the illusion of increased market demand without genuine economic intent, leading to potential regulatory scrutiny.
Use Cases
-
Market Integrity Monitoring: Identifies market manipulation by flagging suspicious trading clusters that artificially inflate trading volumes or distort asset prices.
-
Risk Management for Crypto Exchanges: Crypto exchanges can use this test to detect coordinated wash trading activities that falsely boost the perceived liquidity of specific trading pairs, helping to mitigate reputational risks and ensure regulatory compliance.
Required Data
Trades Data
| Variable | Description |
|---|---|
timestamp | Timestamp of the trade (YYYY-MM-DD hh:mm:ss). |
order_id | Unique identifier linking the trade to its order. |
user_id | User ID to identify the individual user. |
counterparty_user_id | The user ID of the counterparty in the trade. |
symbol_pair | Asset symbol; e.g., BTCUSDT, SHIBETH. |
side | Whether the trade is a BUY or SELL. |
price_usd | Price of the symbol in USD. |
price | Execution price of the symbol pair. |
amount | Amount of the symbol being traded. |
Order Data
| Variable | Description |
|---|---|
user_id | Unique identifier for the user who placed the order. |
symbol_pair | Trading pair for the order, e.g., BTCUSDT. |
order_id | Unique identifier for the specific order. |
order_start_time | Timestamp when the order was created. |
order_end_time | Timestamp when the order was last updated. |
Parameters
| Parameter | Description | Type | Default Value | Configurable |
|---|---|---|---|---|
adv_window | Time window (days) to calculate Average Daily Volume (ADV). | Integer | 7 | Yes |
analysis_window | Time window (days) to check for wash trades. | Integer | 1 | Yes |
analysis_minimum_aggregate_dollar_threshold | Minimum total dollar value to flag wash trades. | Float | 5000 | Yes |
time_delta_threshold | Threshold for median time delta (seconds) between order placement and execution. | Integer | 60 | Yes |
adv_threshold_percent | Percent of ADV used as a threshold for ADV-related checks. | Float | 0.1 | Yes |
maximum_number_users_within_group_threshold | Maximum number of users within a flagged wash trading group. | Integer | 5 | Yes |
minimum_ratio_dollar_volume_within_group_to_total_threshold | Minimum ratio of intra-group trading volume to total volume. | Float | 0.6 (60%) | Yes |
score_threshold | Minimum score required to flag a suspicious trading cluster. | Integer | 4 | Yes |
Methodology
- Calculate ADV (Average Daily Volume)
- Aggregate historical trading data over a defined time window to establish normal trading activity.
- Apply Time Delta Filtering
- Evaluate the time elapsed between order placement and execution.
- Focus on trades with rapid execution indicative of pre-arranged activity.
- Calculate Dollar Value Traded within Group
- Aggregate total dollar volume of trades per user during the analysis period.
- Highlight unusually high trading volumes indicative of wash trading.
- Map Trades to a Directed Network
- Use machine learning to represent trading relationships as a graph.
- Nodes represent users; edges denote trades with attributes like asset type and trade volume.
- Identify Clusters of Coordinated Activity
- Apply community detection algorithms to isolate suspicious trading clusters.
- Score and Flag Suspicious Clusters
- Compute a composite score based on trade execution speed, ADV thresholds, and intra-group trade ratios.
- Flag high-scoring clusters for further investigation.
- Summarize Trading Patterns
- Compute key statistics for flagged clusters, including total traded value and intra-group transactions.
- Generate Visual and Analytical Outputs
- Provide graphical representations of flagged clusters for compliance teams.