Digital Assets
Can On-Chain Data Predict Bitcoin Cycles?
Financial markets are moved by the psychology and actions of institutions and retail investors. This means that analyzing transaction volume, patterns, and other data can be extremely valuable for traders and investors in predicting future price moves.
Bitcoin has been one of the most successful new categories of assets in history, and differs from traditional assets like stocks and gold in a few ways, such as instant transactions and mathematically capped supply. One less discussed difference is how transparent the market and transactions of Bitcoin and cryptocurrencies in general are.
In theory, this can give investors a different way to study market cycles, with every transaction registered and “memorized” open for access in the public ledger.
A new economic research paper published by researchers at the University of Vaasa (Finland) and the University of Turin (Italy) investigates the potential of this method. It was published in Research In International Business and Finance1, under the title “Using on-chain data to predict Bitcoin cycles“.
Predicting Markets
Because predicting financial market prices can be extremely lucrative, considerable attention has been paid to this topic. However, traditional financial models struggle to explain the price movements of crypto assets.
This is because, contrary to stock, cryptos lack intrinsic value tied to a company and potential future dividends. Similarly, they are not the same as national currencies, impacted by the central bank’s decision and a national economy’s strength or weakness.
Instead, the price of cryptos is largely sentiment-driven, even if the underlying utility for transactions or as a store of value is, of course, the deeper reason for cryptocurrencies’ value.
In traditional markets, sentiment-driven price moves are typically inferred from indirect proxies such as surveys or media-based indicators. But blockchains provide a transparent and tamper-resistant ledger of transactions, offering a verifiable record of investor behavior.
In order to answer the question of whether on-chain data are useful to predict Bitcoin prices, the researchers used three on-chain, trading-based measures. They measured them across three major market cycles.
Measuring Bitcoin’s Sentiment
Metrics Overview
The researchers analyze Bitcoin prices from December 7, 2013, to April 12, 2025, which spans three complete market cycles: 2015, 2018, and 2022.
The three indicators used in this study are:
- Net Unrealized Profit/Loss (NUPL) ratio
- Market Value to Realized Value Z-score (MVRV Z-score)
- Cumulative Value Days Destroyed (CVDD).
The first two metrics relate prices to holders’ aggregate cost basis (realized value) and can be interpreted through behavioral finance mechanisms.
CVDD reflects long-term holder behavior as it captures the spending of long-held coins and therefore provides information about capitulation by long-term holders during periods of extreme pessimism.
Overall, the idea is to evaluate investors’ sentiments, where over-optimism may trigger excessive risk-taking and price run-ups that can evolve into bubbles, which then burst when investors panic, and prices fall well below intrinsic value.
In cryptos, search engine activity and social media are among the most prominent sources of sentiment analysis. But on-chain data ultimately contains the proof of such sentiment converting into actions.
Net Unrealized Profit/Loss Ratio
The NUPL ratio approximates the share of coins that are currently held at an unrealized profit or loss.
As such, high values imply (above 0.75) a potential market top, with euphoric sentiment leading to holding substantial unrealized gains. Similarly, low values are typically linked to fear and capitulation at a market bottom.
Market Value to Realized Value Z-score
MVRV Z-score assesses whether a coin is undervalued or overvalued relative to its “fair value,” and is a widely used on-chain metric.
To do so, it combines 3 metrics:
- Market value (MV): Bitcoin price multiplied by the number of coins in circulation.
- Realized value (RV): Valuing each coin at the price at which it was last transferred on-chain and summing across all coins in circulation
- Z-score: Standardizes the deviation between MV and RV by the standard deviation of market value
This indicator suggests that market participants are holding large unrealized gains during bull market phases, when Bitcoin’s market value rises substantially above its realized value.
A score below -0.2 is considered a state of heightened fear and uncertainty. An exit threshold of 5-7 marks that the average participant holds large unrealized gains, generating the behavioral pressure toward profit-taking that historically coincides with cycle tops.
Cumulative Value Days Destroyed
CVDD is built on Coin Days Destroyed (CDD), a metric that weights transactions by both the amount of coins moved and how long they were held.
More precisely, it measures the number of coins transferred multiplied by the number of days since those coins were last moved. CVDD aggregates this activity over time
It can be especially useful to measure the market bottom, as it assesses when long-term holders capitulate.
Can On-Chain Data Predict Bitcoin Price?
Published Results
Several tested NULP strategies all outperformed a buy-and-hold strategy. In addition to higher returns, they all displayed smaller drawdowns as well. The most aggressive NULP strategy turned out to be the most profitable.
MVRV Z-score also exhibited superior and robust risk-adjusted performance relative to the buy-and-hold benchmark. They outperformed the NUPL-based strategies across all metrics, although with some extra volatility in some cases.
The CVDD strategies were proven able to identify cycle bottoms across all trades and window ranges, beating most randomly timed entries.
With a p-value of 99%, it suggests that although CVDD typically enters very close to the bottom, its holding periods are sometimes longer than ideal, which reduces annualized performance.
These results indicate that all three measures contain predictive value, with MVRV Z-score producing the strongest overall risk-adjusted performance and CVDD appearing especially informative for identifying market bottoms.
Overall, the study indicates that yes, on-chain data contain economically meaningful information about Bitcoin market behavior.
Limitations
It should not come as a surprise that market indicators of an overbought or oversold situation of Bitcoin markets help trading better than a buy-and-hold strategy. After all, if such indicators provided no extra advantage, traders would have stopped using them a long time ago.
They are, however, not a crystal ball, and most likely a more sophisticated approach combining multiple indicators will have superior performance, including other types than on-chain indicators.
The research article also admits that more work is needed for analyzing the link between on-chain data and prices for other assets, such as Ethereum, Solana, and XRP.
Similarly, other on-chain metrics are yet to be evaluated scientifically.
AI Dirusption?
Lastly, the emergence of LLMs (Large Language Model) and AI in general could disrupt the pattern back-tested from 2013.
LLMs are increasingly used by retail and institutional investors to interpret market conditions and process information, with the potential to amplify behavioral biases. This could radically alter the dynamics of the on-chain sentiment signals examined here.
So crypto investors should be careful not to get overly confident in the reliability of indicators that worked in the past, as markets are ever evolving, today even more so, as new analytical tools like AI can also change the structure of markets.
So, as always in investing, diversification and remembering that “past performances are not a proof of future results” will matter.
Study Referenced
1. Klaus Grobys, Sebastian Näsman, and Davide Sandretto. Using on-chain data to predict Bitcoin cycles. Research in International Business and Finance. September 2026. Article: 103486. Volume: Volume 89. 10.1016/j.ribaf.2026.103486.
