Fake Volume Is a Data Quality Bug: Spotting Crypto Wash Trading With a Few API Calls

If you have ever built a dashboard based on crypto market data, you might have sorted it by 24-hour volume. This figure is often seen as a signal of market activity. However, there is one significant issue with volume: it is one of the easiest numbers to manipulate in the cryptocurrency world. This process of creating artificial volume is known as wash trading.

In wash trading, one individual or group trades with themselves repeatedly, giving the appearance of a busy market without any real ownership change. This can create a false impression of high trading activity and inflate volume numbers. Detecting wash trading is crucial because it can mislead data-driven applications, such as trading bots and scanners.

The post discusses how developers can identify wash trading as a data quality problem. Wash trading gets its name from the act of washing and re-selling the same assets repeatedly. A simple example would be a team controlling two wallets, A and B. Wallet A sells 100,000 tokens to Wallet B, and then B sells them back to A a few seconds later.

This loop is repeated numerous times daily, making the token appear to trade for tens of millions of dollars. The team remains in control of the same coins throughout the process. The motivations behind wash trading are varied. Higher volume can boost a token's position on trending lists and top-volume rankings. It can aid in exchange listings, set up pump-and-dump schemes, inflate exchange rankings, and accumulate volume-based rewards or airdrop points.

Centralized exchanges (CEX) and decentralized exchanges (DEX) both present different challenges for wash trading. CEX manipulators typically use multiple accounts or market maker bots to place matching buy and sell orders, while DEX manipulators have to deal with higher gas costs but can still pull off wash trades. One common technique in DEXes involves creating many fresh wallets and trading in the same pool to make activity appear more widespread.

Developers should be cautious when writing detection code, as fast back-and-forth trades do not necessarily indicate wash trading. To avoid false positives, developers can use external heuristics to flag suspicious data without definitively proving wash trading. The post provides a small Python script using CoinGecko's public API to identify potential red flags in token volume data.

This script looks for two common warning signs: volume that is disproportionately large compared to the token's market cap, and concentration of volume on a single venue. These heuristics can flag potentially manipulated data, but they are not foolproof. The third check mentioned in the script is to compare volume against market cap, as a small token trading several times its entire value daily with few holders deserves suspicion. Developers should also consider venue concentration and trust scores when analyzing volume data.