Machine Learning, Electricity Prices, and the Cross-Section of Hourly Cryptocurrency Returns
“Using an hourly panel of 43 cryptocurrencies, we find that while point-forecast accuracy is weak across all models, each generates economically significant High–Low portfolio spreads, indicating that machine learning is more valuable for cross-sectional ranking than for precise return prediction.”
AssetCryptosingle stocks, spot crypto
StrategyMachine learningportfolio optimization
GeographyUSUK
FrequencyIntraday
Also stated: universe of 43 cryptocurrencies
Missing from abstract: benchmark, out-of-sample test, costs, multiple testing
Institutional-grade audit of this strategy.
£5 — Request the audit- Audit of the paper claims accounting for costs, luck, crises, originality, geography, etc.
- Reproduction code integratable into Bloomberg, IBKR, etc.
- Bias-free datasets for backtesting included
- Delivered in 2 working days
The paper's neighbourhoodMachine learning return prediction by year
The family is counted across the whole universe. Asset, method and authors are counted inside the family.
| Similar papers | citation | same method | same asset | same authors | |
|---|---|---|---|---|---|
| Does Machine Learning always Beat Traditional Models? Beta Estimation in Cryptocurrency Markets by Aghaei and Ekponon | 2026 | ||||
| Beyond the Hype: A Multi-Layer Machine Learning Framework for Cryptocurrency Return Forecasting by Khoso | 2026 | ||||
| Meta-Learning Reinforcement Learning for Crypto-Return Prediction by Wang et al. | 2026 | ||||
| The Training Set Delusion: Machine Learning Overfitting in Bitcoin Return and Volatility Prediction by Wang et al. | 2025 | ||||
| Evaluating COVID 19 Feature Contributions to Bitcoin Return Forecasting: Methodology Based on LightGBM and Genetic Optimization by Mahmoud and Velichko | 2025 |