Power-Law Corridor

The Power-Law Corridor plots every Bitcoin daily close on a log-log axis — log price against log days-since-genesis — and overlays a single regression line. Two parallel lines bound it at ±1.5σ of the residual distribution; together they form the corridor. The shape is striking on a log-log axis: nearly fifteen years of price action that span more than five orders of magnitude collapse onto a single near-straight band.

Today's fit places the network at $78,199.03 against a median of $129,597.99, a residual of −39.7%. That puts spot in the Lower band band (25.8% of the way from support to resistance). The exponent n is currently 5.6756 with log₁₀(A) = -16.4616 and R² = 0.9612; coefficients refit overnight against the full daily-close history.

The input is a single series: every Bitcoin daily close (see data sources), paired with a deterministic day-count from the Bitcoin genesis block on 2009-01-03. The regression is:

log₁₀(price_i) = n · log₁₀(days_i) + log₁₀(A) + ε_i

Fit by ordinary least squares on every observation since the first traded daily close. The corridor's upper and lower bounds are the fitted median shifted by ±1.5σ in log space, where σ is the standard deviation of the residual log₁₀(price) − (n · log₁₀(days) + log₁₀(A)). Today's σ is 0.3026 in log-units, which corresponds to a corridor that bounds roughly 87% of the historical observations on a normal-residual approximation. Full derivation, including refit cadence, lives on the methodology page.

The two canonical primary sources offer different exponents — empirical and theoretical. Harold Christopher Burger published Bitcoin's natural long-term power-law corridor of growth on 3 September 2019 with the empirical OLS constants a = −17.01593313 and b = 5.84509376 (hcburger.com). Giovanni Santostasi argues separately for a theoretical exponent of n = 6, derived from a scaling-law argument (users ∝ t³, Metcalfe-style price ∝ users², hence price ∝ t⁶) in The Bitcoin Power Law Theory (Medium, 20 Mar 2024). The empirical fit and the theoretical derivation differ by roughly 3% on the exponent — a meaningful gap, not equivalence. Our nightly refit follows Burger's empirical method against the full daily-close history.

Locate spot on the price axis, read across to today's median, and note the log-residual. Negative residuals put Bitcoin below the corridor median — historical accumulation territory. Positive residuals push spot into the upper band, where past cycle tops have lived. The corridor's edges are not hard floors or ceilings — Bitcoin has briefly traded outside on both sides — but excursions outside have been brief in absolute days and small in absolute log-distance.

Sampling the canonical cycle anchors against today's fit shows the corridor's defining property: every cycle peak has lived in the upper band and every cycle trough has lived near support, even as the absolute prices that delivered those bands span four orders of magnitude. The 2024 pre-halving high is the most muted on the table — it cleared the median but did not reach the upper band of the current fit.

Santostasi's 2024 piece motivates the empirical fit by analogy to allometric scaling in biology and physics — Geoffrey West's work on the universal ¾-power scaling of metabolic rate with body mass, and the broader power-law families that show up in city-size distributions, river-network topologies, and cosmic structure. The framing is suggestive: scale-free networks of agents (a metabolism's cells, a city's people, a cryptocurrency's wallets) often grow at rates that compose into power laws.

What makes the analogy useful is what it disciplines. A power law fit through any growth process will look good on a log-log axis; the question is whether the generating process actually has the scale-free properties biological allometry relies on. Bitcoin's network does plausibly fit that description — Metcalfe-style adoption dynamics, hash-rate scaling with price, and address-count compounding. Santostasi runs the argument all the way to a theoretical exponent of n = 6 (users ∝ t³, Metcalfe-style price ∝ users²). Burger's empirical fit lands at n ≈ 5.84. Neither delivers the other: the empirical slope is what the regression spits out; the theoretical slope is what the scaling argument says it should be. A fit dressed up as physics is harder to reject than a fit honestly framed as a fit.

An honest reading of the chart includes the slow drift in the fit itself. Burger's 2019 publication used n = 5.84509376; our nightly refit currently sits at n = 5.6756. The exponent has flattened — not by much, but visibly. Each refit incorporates more low-volatility post-2018 time and proportionally less Mt. Gox-era doubling, which mechanically lowers the slope.

The drift is the model adapting honestly. It is also the seed of the second-largest criticism of the framework: a model whose parameters drift cycle by cycle and whose historical residuals are recomputed under the latest fit will always look better retrospectively than it did in real time. We try to surface both — the live coefficients on this page, the residuals at each cycle anchor under those coefficients — so a reader can audit the gap between forward and backward views of the model.

For a dollar-cost-averaging investor. Treat the lower half of the corridor as your accumulation range and the upper half as your hold range. Time below the median has historically averaged forward returns above the median — that is the model's main descriptive claim, and it is consistent with most mean-reversion-style indicators. There is no need to time around the chart.

For a cycle-timing trader. Pair the corridor with at least two of Pi Cycle, the 200-week MA, and MVRV‑Z. The corridor alone is a slow signal — it can spend months in any single decile of its band — and its edges are soft.

For a researcher. Coefficients refit every night. The OLS implementation, σ-multiplier choice, and the difference between Burger's empirical fit and Santostasi's theoretical exponent are documented on the methodology page.

The high R² is partly an artefact of regressing two trending series. Tim Stolte of Amdax laid out the formal critique in 2022 (Medium, 2 Sep 2022): “Logarithmically scaling time is possibly the weirdest thing I have ever seen in time series analysis” and “we are dealing with a so-called spurious regression… Making something look nice and familiar doesn't mean that it's useful or legit.” The point is technical: when both axes share a deterministic growth component (time on the x-axis, log-price tracking time roughly on the y-axis), classical OLS R² inflates without telling you the relationship has predictive content. The corridor's R² is high; that fact alone is not the proof of quality it appears to be.

Early-history prints carry disproportionate weight. Marty Kendall laid out the issue at mNAV Insights: “the earliest data has a massively more influential effect because each decade occupies the same width on a log axis” (mNAV Insights). On a log-time axis, the years 2010–2012 occupy the same horizontal width as 2015–2018 or 2018–2024. The OLS slope is anchored heavily by the Mt. Gox-era prints — a regime that no longer characterises the asset and that any honest model would weight down. We do not weight; we report the unweighted fit and flag the issue here.

Power laws don't survive forever. Bitcoin's adoption and hash-rate growth have plausibly fit a power law for fifteen years. There is no reason to expect the fit to persist as market cap approaches a meaningful share of global financial assets, sovereign adoption changes the order-flow regime, or regulatory shifts re-anchor the demand curve. Lyn Alden has framed the same observation as a shift from issuance-cycle dynamics to liquidity-cycle dynamics — a structural change the corridor cannot accommodate without a major refit. Treat the band as a long-run shape, not a forecast.

Canonical questions from Google's “People also ask” block for bitcoin power law, answered against the data on this page.