A study abstract reports that Bitcoin’s historical price can be fitted with a long-run power-law curve. That’s an interesting description of past data. It doesn’t prove the curve was identified early, explain Bitcoin’s price, or reliably forecast what comes next.
- Reported price fit: exponent 5.69 ± 0.05; R² = 0.961.
- Proposed mechanism: address growth and a price-to-address relationship.
- Key limitation: the reported results do not establish reliable out-of-sample forecasts.
What the Bitcoin power-law model reports
An abstract for a study titled “Bitcoin’s Price Follows a Robust Power Law: A Mechanistic Derivation of the Exponent” describes Bitcoin’s price with the relationship P(t) ∼ tβ. In plain terms, the model relates price to time. The exponent β determines how quickly the fitted curve rises.
The abstract reports an estimated price exponent of 5.69 ± 0.05 and an R² of 0.961, based on 5, 696 daily observations from July 2010 to February 2026. It identifies those observations as days 560-6, 255 since Bitcoin’s Genesis Block on January 3, 2009.
R² measures how closely a model fits the data used in the analysis. A value of 0.961 indicates a close fit under the study’s chosen model and sample. It does not mean there is a 96.1% chance that Bitcoin will follow the curve in the future.
The abstract doesn’t specify which price series it uses, explain why the sample begins at day 560, or provide enough detail to show how the regression was performed. Each choice can affect the result. The abstract also doesn’t define what the “±” figure represents, so it shouldn’t be treated as a specific confidence interval or error measure without more methodological detail.
Addresses offer a proposed explanation, not proof of cause
The study also reports a relationship between time and address counts: an estimated exponent of 3.046 ± 0.012, with an R² of 0.977. It estimates that price scales with address count with an exponent of 1.838 ± 0.031, with an R² of 0.951.
Multiplying the two address-related exponents gives 3.046 × 1.838 = 5.60. The abstract says this is within 1.6% of the directly estimated price exponent of 5.69. The authors present the calculation as an empirically consistent decomposition: address growth combined with a generalized Metcalfe-type relationship. The idea is that a network’s value may grow with the number of its participants.
That alignment is suggestive, but it doesn’t show that address growth caused Bitcoin’s price to rise. For the multiplication to support the proposed explanation, the relationships need to hold and be comparable across the same sample and definitions. Two quantities that both trend upward over time can also appear closely related without one driving the other.
The abstract calls the mechanism an empirically consistent decomposition and says the proposed adoption explanation is inferred by analogy with spreading on scale-free networks. That isn’t a first-principles proof that network growth determines Bitcoin’s price.
Address counts are also an imperfect measure of adoption. One person can control many addresses. An exchange or custodian can hold coins for many customers, users can reuse addresses, and some addresses may be inactive. An address count is not a headcount of Bitcoin users.
Historical fit is not the same as forecasting
The abstract reports several further checks: a scale-invariance analysis that recovered an exponent of 5.59 from 5, 298 price ratios, a rolling analysis with a median estimate of 5.73 ± 0.58, and a sequential Bayesian analysis with a final estimate of 5.729 ± 0.0132. It also reports no secular drift or structural breaks in its analyses.
These are the study’s reported findings. The abstract doesn’t explain the tests in enough detail to assess their assumptions or reproduce the results, and it doesn’t say what the “±” values represent. The reported checks may support the authors’ view that the historical pattern is stable, but they aren’t independent confirmation. A separate discussion of Bitcoin’s power law and its forecasts isn’t a substitute for independent validation either.
Most importantly, the abstract doesn’t document forecasts made using only earlier data and then tested against later prices. A close fit across a historical sample doesn’t show that the model could have predicted that sample as it unfolded. To make a stronger case for forecasting, researchers would need to disclose forecast dates and methods, measure performance against real outcomes, and compare the results with plausible alternatives. Research on Bitcoin’s unique predictability likewise makes the quality of out-of-sample tests central to any predictive claim.
The abstract reports residual dispersion of 0.302 dex. If that figure is the standard deviation of base-10 log-price residuals, it corresponds roughly to a factor-of-two difference around the fitted relationship. It isn’t automatically a forecast interval, a price floor or ceiling, or a promise that future BTC prices will stay within that range.
The authors interpret speculative movements as cyclical deviations around a structural trend. That may describe the historical data, but the reported fit alone doesn’t show that those deviations reliably reverse or can be traded. A curve that looks orderly in hindsight isn’t a market-timing service. The paper’s framing of weak structure and strong forecasts is a useful reminder that an apparent pattern and a dependable prediction aren’t the same thing.
“Power law” can describe different claims
Here, the study’s main claim is a time-domain relationship: price as a function of time. That differs from a distributional power law, which describes how often values of different sizes occur. Evidence for one doesn’t establish the other. Long-term visualizations, such as this Bitcoin power-law chart, can show the fitted curve, but they don’t validate its assumptions or forecasts on their own.
For distributional power laws, a straight line on a log-log chart isn’t enough to prove that a power law is the best explanation. Researchers may need goodness-of-fit tests and comparisons with alternatives. Those tests don’t automatically validate or invalidate a time-based price regression. The broader point still stands: a convincing curve is a reason to examine the model, not a mathematical crown.
The supplied methodological discussion describes comparisons using AIC, BIC, and likelihood-ratio tests, but gives no comparative results. Without those results, the reported R² can’t establish that a power law describes Bitcoin’s price history better than other approaches.
What would make the model more convincing?
-
What does the model say about Bitcoin’s price?
It reports a long-run statistical relationship between Bitcoin’s price and time, fitted to historical data with a power-law curve.
-
Does the model prove where Bitcoin’s price is headed?
No. The reported historical fit doesn’t establish forecasting skill, and the abstract doesn’t show forecasts tested against later prices.
-
Does address growth explain the price trend?
The study proposes a mathematical connection between address growth and price scaling. But addresses are an imperfect proxy for people and activity, and the reported relationships don’t establish causation on their own.
-
What evidence would strengthen the model?
Clear definitions of the price and address data, transparent methods, tests across different sample choices, and forecasts evaluated against later prices and plausible competing models would make its predictive case easier to judge.
A long-run curve can help organize Bitcoin’s price history and prompt useful questions about adoption, valuation, and market cycles. The abstract reports a close fit and a proposed address-growth decomposition, but the suggestion that an early trend became a dependable long-term forecasting framework isn’t established by those results. Treat the curve as a hypothesis about the past, not instructions from the future.
