Wash trade in cryptomarkets — a case of BW exchange
The value of the 24h aggregated volume, which was calculated by CoinMarketCap (CMC), reached almost $25 billion on February 22, 2019. Quite a number. We have already written that CMC double counts around a third of the trading volume across several symbols. However, there is another issue — no one can exactly estimate which part of the volume is fake and generated by the fraudulent exchanges that aim at ranking higher by volume and attracting new customers by showing non-existent liquidity.
One of the Top-10 exchanges, BW, may serve as a notorious model of an exchange that misleads market participants about its trading volume. Our surveillance algorithm has detected systematic wash trading with the vast majority (up to 100% on some symbols) of trades occurring inside the spread. These trades happen at a midmarket price without having any impact on the order book.
What is wash trading?
The Investopedia says,
wash trading is a process whereby a trader buys and sells a security for the express purpose of feeding misleading information to the market.
CME Group has prepared a great video explaining what is a wash trade and why it is prohibited both by its Rule 534 and 4CA of the Commodity Exchange Act (where it is referred to as ‘wash sale’).
Below is the more formal definition proposed by CME.
A wash trade is a form of fictitious trade in which a transaction or a series of transactions give the appearance that authentic purchases and sales have been made, but where the trades have been entered without the intent to take a bona fide market position or without the intent to execute bona fide transactions subject to market risk or price competition.
Who benefits from wash trading?
Wash trading application to the crypto market usually implies illegal activities to artificially increase the trading volume and create a false impression of a liquid market.
There are two main groups of beneficiaries of wash trading:
- crypto exchanges;
- major stakeholders of crypto assets.
Exchanges. There are 240+ exchanges whose trading data is published by CoinMarketCap, although the actual number is higher. However, only 137 exchanges showed the 24h volume exceeding $1M and only 44 showed the reported volume of >$100M. How can a new exchange get more attention from the community? One of the popular ways is to get a higher rank in the list of top exchanges by trading volume. When the actual trading volume is low, the artificial one steps in.
There are several metrics that may help detect inflated volumes without analysing market data:
- the ratio of website visitors to 30-days trading volume (SimilarWeb provides free statistics of visitors while CMC publishes 30-days volume). CoinGecko is a great resource where you can have Alexa rating as well as Twitter followers along with trading volume. Unfortunately, this metric (like many others) does not count for algorithmic volume, which comes directly via API.
- the ratio of followers on Twitter, Facebook etc. to 30-days trading volume. The drawback of this approach is that the number of followers may not reflect the actual popularity of an exchange. An exchange might have been popular in the past and has a lot of followers just by inertia.
- a ratio of mentions (#hashtags) in Twitter etc. to 30-days trading volume. It would be a good metric if there were no bots in Twitter that ruin any statistics.
- a ratio of new posts on Reddit etc. to 30-days trading volume. Unfortunately, such a statistic is not readily available.
- a ratio of blockchain transactions to/from exchange wallets to 30-days trading volume. This approach is good but far from being perfect. First, nobody is immune to blockchain fraud as blockchain transactions can also be artificially inflated. Second, it is not possible to 100% correctly associate an address with a particular exchange.
Assets. As to the coins, the rationale behind wash trading is straight-forward. Higher volumes indicate greater attractiveness of an asset. Higher volumes are often an indication of higher liquidity (however, it is not always true— learn more about it in our article Why Liquidity Matters). Higher volumes usually positively correlate with higher market capitalization, i.e. higher worth of coin’s major stakeholders.
Please note we are focusing only on exchanges in this article.
Good vs Bad market making
Someone outside the industry may wonder whether all market makers are involved in wash trading. Of course, not! A market maker is a market participant whose aim is to provide liquidity by simultaneously placing two-side quotes. The presence of a market maker is positive for other market participants and is supposed to decrease transaction costs for them.
Some of the exchanges stimulate makers by offering lower fees. Other crypto exchanges offer special MM programs (Coinfloor among them), alike classical stock or futures exchanges, where there are certain requirements that a market maker must fulfil in order to benefit from lower fees: the maximum bid-ask spread, the minimum time two-sided quotes must be present in the order book, the minimum size of two-sided quotes etc. These are both examples of good market making practices.
In cryptocurrency markets, there are however market making firms whose offering is about producing fake volumes rather than providing liquidity. The main activity of those fraudulent market makers is inflating the volume of a specific coin but they may as well offer wash trading across different assets for an exchange. Luckily, there is another type of market makers that place market integrity above all else.
Similarly, there are exchanges that take actions to detect and prevent market abuse. Gemini has partnered with Nasdaq Inc’s market surveillance technology, Bitstamp has announced its plans to implement the Irisium Surveillance platform while Seed CX claims it has a dedicated team of surveillance professionals.
The case of BW
Our previous research has revealed some trading anomalies in Binance and HitBTC (you may want to read Is this real? Fake Volumes on Hitbtc and Binance and All Quiet on the Binance Front). Now the target is a cryptocurrency exchange called BW.com, which has been recently among top-10 exchanges by trading volume according to CMC.
On February 22 the reported trading volume was almost half a billion dollars. Noteworthy, top-5 symbols contributed more than 95%. Let’s have a closer look at it.
The research paper Quantifying fake volumes on cryptocurrency exchanges, which was published in May 2018, claims only 40% of trading volume is not manipulated. It has a nice illustration of market depth with the simple explanation of fake volume detection. The author considers two types of wash trading:
- an exchange reports trades that do not correspond to actual orders at all;
- an exchange (or another market participant) places a limit order and consequently hits it to produce volume (it may be done from a single account or several controlled accounts).
Now let’s revert to the trading activity observable on BW. The vast majority of trades occurs inside the spread, namely at a midmarket price. To make the matter worse, the genius behind this activity does not bother to place limit orders, i.e. we have detected the Type I of wash trading (with some exceptions). Below are the figures showing the pattern for two most liquid symbols BTC/USDT and ETH/USDT.
The less liquid symbols show the same pattern — EOS/BTC, BSV/BTC. The price is presumably calculated as per the formula: Trade price = (bid + ask)/2 + (ask - bid)* random{-0.05,0.05}).
The rough estimation of the artificial volume share is made according to a simple algorithm. First, we calculate the number of trades and the volume of trades (in base currency) that happened within the spread. Second, we calculate the ratio of in-spread trades to the overall trades.
As it was mentioned above, there are some exceptions from the general rule of producing fake trades in the middle of the spread. The next figure shows a case when a lot of trades happened at bid prices. No market impact was observable, though. Such behaviour may be explained by the activity of a wash trading bot.
Further analysis of trading data shows anomalies detected by the following two metrics:
- the time interval between the two consequent trades and its distribution;
- the trade size and its distribution.
The interval between trades. As we know from our previous research, the most popular interval between the trades is usually 0. It corresponds to common sense because it’s common when a hitting order leads to the execution of multiple resting orders. The BW reports the time stamp only with second precision. The distribution is shown below and looks more or less all right. However, if we create a histogram using the time when a trade was received to our server, the picture will be much different. The distribution is two-modal with both modes being positive (around 36 and 42 ms). It is noteworthy that the distributions of BUY and SELL trades are almost identical, which is theoretically possible but not probable.
Trade size. On exchanges with a high share of retail flow, trade sizes are usually dominated by integer numbers (like 1, 300 or 5000) or tenth of an integer (0.1 or 0.9) because it is easier for people to type 0.1 than 0.98237346. However, that is not what we see on BW.
Conclusion
BW has clear signs of market manipulation detected by our surveillance system. The vast majority of trading volume on BW is the result of trades that occur at midmarket prices (up to 100% in some symbols). These trades do not change the order book and occur in a parallel reality. The similar pattern is detected across all observed symbols. As a result of simple reverse engineering, the assumed algorithm has the following parameters:
Trade side = random
Trade frequency, ms= random[random{36 ±36},random{42 ±42}]
Trade price = (bid + ask)/2 + (ask-bid)* random{-0.05,0.05})
However, it is regrettable and worrying that fraudulent market participants are becoming more sophisticated. It may be much more difficult to detect trading anomalies on other exchanges that utilise wash trading as a marketing tool. This notwithstanding, we plan to continue our research and cover more crypto exchanges.
P.S. Your feedback is welcome. It’d be much appreciated if you could share with us any mistake you may find in our calculations. To make our work as transparent as possible we have made the underlying codebase public — visit our GitHub repository.
