Study: How to predict pump-and-dump schemes

Two scientists from Imperial College London have taken a close look at the process of pump-and-dump schemes. Based on their findings, Jiahua Xu and Benjamin Livshits developed a prediction model based on machine learning that significantly increases the chances of “successful” participation in a pump-and-dump event.

Who Bitcoin trader doesn’t know this:

At the crypto market, some shitcoin that nobody had on the screen before suddenly goes through the Bitcoin trader ceiling for no apparent reason. As soon as you as a Bitcoin trader start to find out more about the coin (maybe it’s a project with substance?), you can see the course slipping back into red in the corner of your eye. This FOMO-causing course behaviour is mostly based on a fiddling around. This fiddling is called pump and dump.

A pump and dump is about consciously and in a coordinated manner driving a price upwards (“to pump” = inflate, inflate) in order to sell the asset immediately after reaching the highest level with a clear profit (“dump” = throw it away, unload it). Although pump-and-dump schemes are significantly older than Bitcoin & Co., they fall on particularly fertile ground in crypto space. And this despite the fact that the initiators of the pump and the dump benefit first and foremost.

In their paper “The Anatomy of Cryptocurrency Pump-and-Dump Scheme”, Jiahua Xu and Benjamin Livshits have explored this form of market manipulation in more detail. They have found that comparably simple machine learning models can make useful predictions about which coin will be pumped. They observed over 220 pump and dumps between July and November 2018.

Crypto trader Choice: Cryptopia

More than two thirds (67 percent) of the crypto trader manipulations took place on the Cryptopia crypto exchange, followed by Yobit (18 percent), Binance (eleven percent) and Bittrex (four percent). In every fourth crypto trader case there has already been a pump of the respective coin at the corresponding exchange. The authors also note that Binance has significantly more users involved in pump-and-dump activities due to its large reach. However, the corresponding price increases of the pumped coins were significantly lower at these trading locations than at Jobit and Cryptopia. The reason lies in the listing of coins with a low market capitalization, which are sought in vain on the large crypto exchanges.

“Pump-and-dump organisers who prefer Cryptopia are attracted by the wide range of low market capitalisation coins listed on the stock market. It is not surprising that small coins are more likely to be associated with fraud, leading to a possible delisting,

so the two authors.