Authors:
Vivek Gautam, Lekshmi Kalinathan, Aman Kumar Maurya, Ayush Yadav, Humara Ahmad, Mutaher Ahmed

Addresses:
School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, Tamil Nadu, India. Business School of UOP, University of Portsmouth, Portsmouth, Hampshire, United Kingdom. School of Computing and Artificial Intelligence, The University of Law, Bunhill Row, London, United Kingdom.

Abstract:

The architectural diversity of algorithmic trading methods within cryptocurrency trading is extensive; however, there is little known about the connection between the density of indicators used and the performance of such algorithms. This study rigorously compares two rule-based trading algorithms for Ethereum (ETH-USD) and Bitcoin (BTC-USD) over a four-year market cycle from 2022 to 2025. The first method uses a weight-based approach that includes more than 50 different technical indicators. The second one uses a simple unweighted voting algorithm. To test whether machine learning can selectively improve trade entry quality, researchers introduce the Technical Indicator State Vector (TISV) framework and use an XGBoost classifier in a strict walk-forward setting. Empirical results demonstrate that the simpler, high-frequency strategy outperformed the complex composite model by a factor of 4.27 over four years, driven by geometric compounding and superior capital utilization. Furthermore, TISV-based ML filtering improved the composite strategy’s out-of-sample Sharpe ratio by 739% and its win rate from 35% to 63.76%. However, it provided no measurable benefit to the high-frequency scalping strategy. These findings establish a diagnostic criterion: machine learning-based trade filtering works only when a strategy’s outcomes measurably depend on market entry conditions (defined as an Area Under the Curve > 0.60).

Keywords: Algorithmic Trading; Cryptocurrency Markets; Exponential Moving Averages; Bollinger Bands; Technical Indicators; Technical Indicator State Vector; Walk-Forward Optimization; High-Frequency Trading.

Received on: 07/11/2025, Revised on: 20/12/2025, Accepted on: 13/02/2026, Published on: 16/10/2026

DOI: 10.69888/FTSFDS.2026.000799

FMDB Transactions on Sustainable Finance and Data Science, 2026 Vol. 1 No. 4, Pages: 187-196

• Views : 71
• Downloads : 10

Download PDF