Equinox 2 is a virtual trading robot that orbits the Ethereum Universe.
Abstract from Original paper
The Ethereum blockchain is a birthplace and cradle of Decentralized Finance (DeFi) projects that seek to usher in a new world of financial technology. Section 1 of this paper extends empirical studies of market efficiency, detection of price bubbles, NLP based herd detection and Diebold-Yilmaz volatility connectedness, to 22 Ethereum blockchain based (or originally based) assets that remain largely unexplored within the purview of academic investigation. Section 2 leverages Section 1’s insights to build a virtual trading robot named “EQUINOX 2.0” that utilizes 3 Deep Learning algorithms to predict future prices while mitigating risks associated with bubble dynamics and volatility spillovers within the Ethereum Eco-system. EQUINOX 2.0 also incorporates Machine Learning to filter out weak trade signals to curb over-trading. Overall, this paper’s findings are four fold:
1.The Ethereum Eco-system is weak form market inefficient following a battery of 8 random walk tests and LSTM, GRU, and TCN Deep Learning models using historical data are effective in predicting asset prices with low Mean Average Error scores.
2.Bubbles exist and occur at the same time for all assets despite sporadic idiosyncratic/isolated bubbles.
3.Ethereum is the main emitter of volatility to other assets in its Ecosystem and smaller assets (small by market capitalization) are characteristically volatility receivers from the larger ones
4.Investors herd away from and not within the Ethereum Ecosystem but noisy herds are still detectable when using Natural language Processing based analysis on Social Media data.