Predictive sorting of cryptocurrencies based on fundamentals and sentiment

This paper examines the predictive power of blockchain characteristics and sentiment indicators for cryptocurrency returns. We construct three weekly factor-mimicking portfolios based on network activity (active users), computing intensity (hashrate), and a sentiment measure from Google search trends. Using an out-of-sample forecasting framework, we find that all three predictors show strong performance across 40 cryptocurrencies. The certainty equivalent returns are often well above the risk-free rate, which supports the economic relevance of the blockchain-driven predictors. We also implement a portfolio sorting methodology that ranks cryptocurrencies by earlier, realized factor-based predictability scores and forms long-short portfolios accordingly. The resulting return spreads confirm the value of combining blockchain and sentiment-based signals. Overall, our findings emphasize the joint relevance of both fundamental and behavioral factors in predicting cryptocurrency returns.