A mathematical model that explains over 95% of Bitcoin's price variance across 15 years of data—providing expected return estimates where traditional financial frameworks acknowledge they cannot.
Every major institutional framework for Bitcoin allocation—from BlackRock to Fidelity to Morgan Stanley—confronts the same fundamental challenge: Bitcoin has no cash flows, no earnings, and no traditional basis for estimating expected returns.
BlackRock's 2024 portfolio sizing paper states it directly: Bitcoin has "no underlying cash flows for estimating future returns." Fidelity relies on Monte Carlo simulations of historical data. Morgan Stanley uses mean-variance optimization with acknowledged limitations. All of these approaches treat Bitcoin's future price path as essentially unknowable.
The power law model offers a different perspective. It doesn't claim to predict exact prices. But it does provide a mathematically grounded framework for expected return estimation—something no traditional approach offers.
A power law is a mathematical relationship where one quantity varies as a power of another. In nature, power laws govern city sizes, earthquake magnitudes, metabolic rates, and network effects. They emerge wherever systems exhibit scale-invariant growth.
Applied to Bitcoin, the power law models the relationship between price and time since the genesis block (January 3, 2009). On a log-log scale, Bitcoin's price traces a remarkably straight line:
Where β (the power law exponent) captures the growth rate and A is a scaling constant. Using the Santostasi parameterization:
This single equation, with just two parameters, explains more than 95% of the variance in Bitcoin's price over its entire 15+ year history—and that figure has been rising. For context, most financial models consider an R² above 0.3 useful.
The power law trend line represents fair value. By measuring how far historical prices deviate from this trend (the log residuals), we construct a complete valuation envelope:
Fair Value (trend line)—the power law regression itself. This is the model's estimate of where Bitcoin "should" be priced at any given point in time, based on its age.
Floor (−2 standard deviations)—the adoption growth baseline. Historically, Bitcoin has never sustainably traded below this level. It represents the minimum price consistent with the network's organic growth trajectory.
Ceiling (+2 standard deviations)—the speculative overshoot boundary. Cycle tops have historically touched or exceeded this level before mean-reverting.
Key insight: The floor line rises every day, regardless of market sentiment. Even during the deepest bear markets, price has closed below the floor on only about 2.4% of trading days (135 of 5,713 daily closes)—recovering above it every time. This provides a mathematically defined margin of safety that improves with time—a property no other Bitcoin valuation framework offers.
The multiplier (current price ÷ fair value) provides an instant read on Bitcoin's position within its historical valuation range. This is the metric our tools use to drive allocation recommendations, Kelly criterion sizing, and rebalancing signals.
Note the visible compression: early cycles saw the multiplier swing from 0.1× to 16×; recent cycles oscillate between roughly 0.5× and 3×. This is Bitcoin's volatility declining on a log scale—a structural feature the power law framework predicts and quantifies.
Deriving σ = 0.2 (log10)
The σ bands used throughout our tools are based on the standard deviation of log10(price / trend) across recent halving cycles. Combining Cycle 4 (May 2020–Apr 2024) and the ongoing Cycle 5 (Apr 2024–present) yields σ ≈ 0.21, consistent with the σ = 0.2 value used independently by Perrenod and Krueger. We use this “mature Bitcoin” volatility rather than the all-time value (σ ≈ 0.35), which is inflated by Bitcoin’s chaotic early years. As the volatility eras chart shows, each successive halving cycle produces a tighter distribution—using only recent cycles ensures our uncertainty bands reflect what Bitcoin actually does today, not what it did in 2011.
A common objection to any statistical model is overfitting: the model merely memorizes historical data and will fail going forward. The power law model provides a compelling counter to this concern. When we calculate R² using an expanding window—fitting the model on data available up to each year and measuring goodness of fit—the fit has improved monotonically:
An overfit model degrades as new data arrives. The power law model improves. Each passing year adds out-of-sample data that the original regression never saw, yet the fit tightens. This is the statistical signature of a model that captures a genuine underlying process rather than noise.
What this means for allocators: The model's reliability has a measurable track record. Every year that Bitcoin's price continues to conform to the power law corridor strengthens the case for using it as a valuation framework. The trend is toward more certainty, not less.
The most serious statistical objection to the power law model deserves direct engagement. The critique, articulated most thoroughly by Tim Stolte, runs as follows: when two non-stationary variables (variables that trend over time) are regressed against each other, R² can approach 1.0 regardless of whether any genuine relationship exists. Both log-price and log-time trend upward, so a high R² might simply be an artifact of shared trend rather than evidence of a real underlying process. This is a well-known phenomenon in econometrics called spurious regression.
This is a legitimate concern that cannot be dismissed. However, several lines of evidence suggest the power law relationship is not spurious:
In a spurious regression, the fitted parameters drift continuously as new data is added—there is no "true" value for them to converge toward. The power law exponent (β) has remained essentially stable since approximately 2017, fluctuating within a narrow band despite Bitcoin's price moving through an entire cycle of boom, crash, and recovery. If the relationship were spurious, the parameters would wander. Instead, they converge—the behavior expected of a model capturing a real process. (Burger & Santostasi, 2024)
The mean squared error (MSE) of the model's residuals decreases over time. In a spurious regression, prediction errors do not systematically shrink—they fluctuate or grow as the spurious fit gradually breaks down. The power law model's residuals are doing the opposite: compressing. Each new year of data is closer to the model's prediction than the last, on average.
The spurious regression argument requires that both variables be non-stationary and non-cointegrated. But the log-residuals of the power law regression (the deviation of log-price from the model's predicted log-price) appear to be stationary—they oscillate around zero without trending. This is consistent with cointegration: even if the individual series trend, their relationship is stable. A model whose residuals mean-revert is, by definition, not spurious.
Purely statistical arguments cannot settle this debate alone. What distinguishes the power law from a spurious fit is that it has a plausible generative mechanism. Metcalfe's Law states that network value scales as the square of its users. Bitcoin's difficulty adjustment creates a feedback loop between price, hash rate, and security that reinforces adoption. The same power law patterns appear in internet adoption, mobile phone penetration, and social network growth—systems where adoption drives value and value drives adoption. The power law captures the mathematical signature of this process. Spurious regressions have no such mechanism; they are accidents of shared trend.
None of this constitutes proof that the power law will hold indefinitely. But the combination of parameter stability, decreasing prediction error, stationary residuals, and a plausible physical mechanism means the spurious regression critique does not invalidate the model on the evidence available. The strongest counter-evidence would be a sustained, multi-year deviation from the power law corridor—which has not occurred in 15+ years of data.
Intellectual honesty demands that we present what the model cannot do alongside what it can. No model is infallible, and overconfidence in any single framework is itself a risk.
Structural breaks: The power law could break down if Bitcoin's adoption curve is disrupted by regulation, technological failure, or a superior alternative.
In-sample fit: The R² is calculated on historical data. All models fit their training data well. The true test is out-of-sample performance.
Cycle timing: The model identifies where Bitcoin is relative to trend, not when the next cycle turn occurs. It is a valuation tool, not a timing tool.
Tail risk: In a catastrophic scenario (global ban, protocol failure), the floor could be violated. The model assumes continuity of the network.
We believe transparency about these limitations is what distinguishes a serious analytical framework from market hype. Our tools present projections as ranges with explicit uncertainty bands, not point predictions.
The power law framework is the engine behind everything we build. Our tools translate the model's mathematical properties into practical decision support:
Portfolio allocation: The model's expected return and volatility parameters feed directly into Kelly criterion position sizing, providing mathematically optimal allocation percentages based on where Bitcoin sits in its valuation range.
Retirement planning: The three-line framework (floor, fair value, ceiling) generates scenario projections for long-term financial planning—conservative, base, and optimistic cases grounded in the model rather than arbitrary assumptions.
Dynamic rebalancing: The multiplier drives buy/hold/trim signals based on deviation from fair value, replacing emotional decision-making with a systematic framework.
This work stands on the shoulders of researchers who established the theoretical and empirical foundations:
The Bitcoin Observatory's contribution is the application layer: translating these theoretical models into practical financial planning tools validated through real-world client consultations.
If you are an institutional allocator, fund manager, or financial advisor exploring Bitcoin allocation, we offer confidential analyses applying this framework to your specific portfolio context.
Disclaimer
This material is provided for educational and informational purposes only and does not constitute financial, investment, tax, or legal advice. The author is not a registered investment advisor, broker-dealer, or financial planner. All projections, models, and analyses are based on mathematical frameworks and historical data—past performance does not guarantee future results. Bitcoin and digital assets are highly volatile and speculative; you may lose your entire investment. The power law model, while demonstrating strong historical fit (R² > 0.95), is a statistical model subject to structural breaks and may not hold in the future. Always consult a qualified financial professional before making investment decisions.
© 2026 Alexander Mooij · Bitcoin Observatory · info@alexandermooij.com