The Crypto Prediction Divergence: When Accuracy Doesn't Equal Profit

This research explores the economic potential of integrating machine learning with technical analysis by analyzing major cryptocurrencies—Bitcoin, Ethereum, Litecoin, and Ripple -using long-term historical data from 2018 to 2024. While earlier studies have predominantly focused on short-term forecasts utilizing standard technical indicators, the broader performance of machine learning-assisted technical strategies across varying timeframes and market conditions remains underexplored. To bridge this research gap, XGBoost classification models are employed with a notably large feature set of 771 technical indicators, a scope rarely seen in prior literature. These models are used to evaluate predictive accuracy over forecast horizons ranging from 5 to 120 trading days. The most consistent and accurate predictions are observed for selected cryptocurrencies, especially over the 120-day period. Unexpectedly, the highest annualized returns emerge from shorter forecast horizons, such as 5 days. Elevated market volatility further enhances both predictive accuracy and trading profitability. This divergence likely stems from the models being tuned for classification accuracy rather than return optimization. Additionally, the study finds that despite strong predictive performance, technical analysis may not be universally applicable or effective across all cryptocurrencies.

Authors

Institutions

Publication Details

Journal
Computational Economics
Published
2026-11-17
DOI
https://doi.org/10.1007/s10614-026-11467-9
Primary Topic
Stock Market Forecasting Methods
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

The Crypto Prediction Divergence: When Accuracy Doesn't Equal Profit

Wojciech Kuryłek
5 links
Computational Economics
Stock Market Forecasting Methods
article

The Crypto Prediction Divergence: When Accuracy Doesn't Equal Profit

Wojciech Kuryłek
article en

Abstract

This research explores the economic potential of integrating machine learning with technical analysis by analyzing major cryptocurrencies—Bitcoin, Ethereum, Litecoin, and Ripple -using long-term historical data from 2018 to 2024. While earlier studies have predominantly focused on short-term forecasts utilizing standard technical indicators, the broader performance of machine learning-assisted technical strategies across varying timeframes and market conditions remains underexplored. To bridge this research gap, XGBoost classification models are employed with a notably large feature set of 771 technical indicators, a scope rarely seen in prior literature. These models are used to evaluate predictive accuracy over forecast horizons ranging from 5 to 120 trading days. The most consistent and accurate predictions are observed for selected cryptocurrencies, especially over the 120-day period. Unexpectedly, the highest annualized returns emerge from shorter forecast horizons, such as 5 days. Elevated market volatility further enhances both predictive accuracy and trading profitability. This divergence likely stems from the models being tuned for classification accuracy rather than return optimization. Additionally, the study finds that despite strong predictive performance, technical analysis may not be universally applicable or effective across all cryptocurrencies.

Computational Economics
University of Warsaw (PL)
Stock Market Forecasting Methods
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon