Using on-chain data to predict Bitcoin cycles

Abstract

There is limited literature studying the predictive power of on-chain data for Bitcoin price cycles. This paper contributes to this literature by assessing whether three on-chain, trading-based measures help predict the Bitcoin price time series across three major market cycles. We find that these indicators outperform both a buy-and-hold benchmark and random-entry strategies simulated through Monte Carlo analysis. For example, the Sharpe ratio increases from 0.45 for the buy-and-hold benchmark to 1.28 when using the Market Value to Realized Value Z-score measure. This study contributes to the literature by showing that blockchain-based behavioral data provides predictive value in decentralized markets that lack intrinsic valuation anchors. The findings also have practical implications for investors, traders, and regulators, and they challenge traditional notions of market efficiency by providing evidence of recurring behavioral patterns embedded in publicly observable blockchain activity.