- Cong, Li, Tang and Yang tested 29 cryptocurrency exchanges for fake transactions and found that on the unregulated exchanges, wash trading averaged over 70% of the reported volume.
- The tests look at statistical fingerprints: the distribution of first significant digits, rounding of trade sizes and the shape of the tail of trade sizes.
- Regulated exchanges showed the patterns normally seen in financial markets; unregulated ones did not.
- Fabricated volume improved exchange rankings and temporarily distorted prices, so reported volume is a weak basis for judging liquidity or choosing where to trade.
Trading volume is the number most people use to judge whether a market is liquid and an exchange is busy. It is also easy to inflate: a trader who buys and sells to themselves creates volume without taking any risk. That practice is called wash trading, and one of the most cited studies of it in crypto concludes that, on part of the market, most of the reported activity was not real.
What the study found
Cong, Li, Tang and Yang introduce “systematic tests exploiting robust statistical and behavioral patterns in trading to detect fake transactions on 29 cryptocurrency exchanges”.1 The main findings, in the authors’ words, are:1
- “Regulated exchanges feature patterns consistently observed in financial markets and nature.”
- On unregulated exchanges, “abnormal first-significant-digit distributions, size rounding, and transaction tail distributions” reveal “rampant manipulations unlikely driven by strategy or exchange heterogeneity”.
- Wash trading on each unregulated exchange “averaged over 70% of the reported volume”, amounting to trillions of dollars of fabricated volume per year.
- The fabricated volume improves exchange rankings, temporarily distorts prices, and is related to exchange characteristics such as age and user base, to market conditions and to regulation.
The study is a working paper from 2021, covers the 29 exchanges that were examined at that time, and does not claim to describe every exchange today. It does show that the problem was large enough to matter, and that it can be detected from the data.
How fake volume is detected
The idea behind the tests is that real trading leaves statistical fingerprints that manufactured trades do not copy well. Three are named in the abstract:
1. First significant digits. In many naturally occurring sets of numbers that span several orders of magnitude, the first digit is not uniform. It follows what is known as Benford’s law: digit d appears with probability log10(1 + 1/d).
| First digit | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
|---|---|---|---|---|---|---|---|---|---|
| Expected share | 30.1% | 17.6% | 12.5% | 9.7% | 7.9% | 6.7% | 5.8% | 5.1% | 4.6% |
Sizes produced by a simple generator may not reproduce this pattern, which is one reason such a test can flag fake flow.
2. Size rounding. The abstract names rounding of trade sizes as one of the patterns that looks abnormal on unregulated exchanges, compared with the way real traders’ sizes are rounded on regulated ones.
3. The tail of trade sizes. The authors also compare the distribution of the largest trades with the pattern observed in regulated markets, and report abnormal tails on unregulated exchanges.
These tests do not prove a single trade is fake. They indicate that the volume of an exchange as a whole does not look like trading by independent participants. Our description is a simplified explanation of the idea based on the abstract; the authors’ tests are more detailed, and the full paper should be read for the method.
What it means for traders
- Reported volume is not liquidity. A busy-looking exchange can have a thin real order book. Measure liquidity by the depth you can actually trade, as in our example of slippage and price impact, not by the headline volume.
- Volume-based signals inherit the problem. An indicator built on volume spikes, on volume rank, or on “the most traded coins” is only as good as the volume. A spike on an exchange with inflated volume may be noise.
- Rankings can be gamed. The study found that fabricated volume improves exchange ranking. A rating based on volume tells you which exchange promotes itself, not which one is safest or deepest.
- Arbitrage and price comparison need caution. Price distortions from fake trades were found to be temporary; a price gap that appears on a venue with unreliable volume may not be tradable, as we discuss in why crypto arbitrage gaps persist.
- Prefer venues and data sources that are transparent and regulated, and cross-check volume with depth, spread and the number of distinct price levels traded.
Practical self-check: take a few thousand trade sizes from an exchange’s public feed, compute the share of each first digit and compare with the table above. A large, systematic deviation is a reason to look closer. It is not proof, because Benford’s law applies only to data that spans several orders of magnitude, and some real markets use fixed lot sizes. This article is educational material, not investment advice.
Footnotes
Sources
- Crypto Wash Trading. Lin William Cong, Xi Li, Ke Tang, Yang Yang. arXiv:2108.10984 (working paper, 2021), 2021
This article is research, not investment advice. Results on history do not guarantee future results.