Detecting Manipulated Cryptocurrency Trades
Cryptocurrencies are new, volatile and unpredictable. Due to lack of regulation, market participants are subject to manipulated trades, called Pump-Dump schemes. Thus, detection of such trades are quite important for market makers and traders. In this study, we use deep and machine learning methodologies to classify and detect manipulated cryptocurrency trades on main crypto exchanges like Binance. We also compare predictive power for several models. Our contribution is three folded. First, we begin with identifying the systematic and idiosyncratic components of manipulated trades. In order to dissect systematic movements, we control for variations in Bitcoin price. Second, we look at the coin specific factors such as regime shifts in volume, price that drive the trades in the short run. Finally, unlike existing studies, we focus on more refined time frequencies and horizons.
Mesut is a Senior Lecturer and Course Leader at Finance at Westminster Business School. He holds PhD in Finance (Cass Business School) and MSc in Finance (University of Texas). He taught at London School of Economics (LSE) for several years. He is also founder of SNS Analytics which develops and implements algorithmic trading mostly in cryptocurrencies. Mesut is an expert and avid speaker for various topics in data science, textual/sentiment analysis, and FinTech. He has a big appetite for automation.