Machine Learning-Driven and Statistical Algorithmic Trading for Cryptocurrency Markets

  • SupervisorProf. Yiu Siu Ming
    MembersChu Tsun Hang3036063747
    Ma Tsz Hin3036104242
Introduction

The cryptocurrency market’s constant operation and high volatility present significant risk management challenges that overwhelm traditional trading approaches. Static strategies become ineffective, making real-time, data-driven algorithmic programs essential for identifying and exploiting short-term opportunities. These machine learning models can process market nuances faster than humans, creating a competitive advantage by adapting to rapid changes.

Results

1. Data Pipeline and Feature Extraction

2. Backtest Performance

Sharpe RatioSortino RatioCalmar RatioAnn. ReturnMDD
Portfolio4.164.766.73133.3%-19.8%

3. Real-trading Performance

Portfolio/FundReturnsDrawdownSharpe Ratio
Project Portfolio(data from Exchange with trading fee included)23.70%-5.12%2.03
BarclayHedge Cryptocurrency Traders Index-7.08%NANA
Quant-Driven Multi-Factor Rota(data from Bybit without trading fee included)-8.29%-35.67%-0.81
COCI Quantum(data from Bybit without trading fee included)-0.08%-4.32%-0.09
Algo Intelligence Pro AI(data from Bybit without trading fee included)-73.58%-77.42%-1.19