Bitcoin · power law · methodology
In an attempt to mount evidence going forward, here's how powerlawfloor approaches this:
01 / FROZEN AT LAUNCH
The exponent, the coefficient and the floor multiplier this site uses were fit on Bitcoin price data only through 6 Jun 2026. They have not been adjusted since.
Any future change to these constants will be disclosed as a new, dated model version. It will never be a silent revision. The changelog at the bottom of this page is where a change like that gets recorded, with its date and its reason.
02 / OUT-OF-SAMPLE
The price-versus-floor chart on this site, from 6 Jun 2026 forward, is the model's live, unedited predictive record. It is not a backtest.
A backtest is fit to data it already knows the answer to. An out-of-sample record is not, the parameters were locked before most of that price history existed. That distinction is the entire reason the parameters were frozen in the first place, and it only holds if the freeze actually holds.
03 / EVIDENCE, NOT PROOF
A multi-year track record strengthens or weakens confidence in the model. It does not settle the question, and it was never going to.
Even a correct model will have periods of deviation. Price can sit above or below the trend for a stretch and still be consistent with a real, underlying regularity. Even a flawed model can track well for a while by chance alone, especially over a short window or a single market cycle. Neither a good run nor a bad run, taken on its own, proves anything.
04 / THE FALSIFICATION CONDITION
a sustained close below the frozen floor for six consecutive months, with the log-log R-squared degrading and the slope forced materially lower on each refit.
This is the line that would cause us to say the model has failed, stated in advance rather than defined after the fact once a candidate breach has already happened. A model that only tells you it was wrong once you go looking isn't falsifiable. It's just unfalsifiable with extra steps.
05 / MECHANISM, NOT JUST CURVE-FIT
The historical fit is one kind of evidence. The numbers have tracked closely for over a decade. The mechanism is a different kind. Metcalfe-style network growth, where user adoption scales roughly with the cube of time (N proportional to t3) and price follows from network value, gives a causal story for why a power-law shape might hold at all, not just a description of the shape it happened to take.
These two are not the same evidence counted twice. A tight historical fit with no plausible mechanism is just curve-fitting. A plausible mechanism with no supporting fit is just a story. Treating them as one combined, doubly-confirmed thing would overstate the case, so this site tries to keep them separate. The fit itself is checked on its own terms in What the Floor Is, and the mechanism is stress-tested on its own terms in Disprove the Math, where the causal chain gets attacked directly instead of assumed.
This page describes a method, not a result. It is not investment advice, and nothing here guarantees an outcome.
CHANGELOG
Last updated: 6 Jun 2026 (launch).
No parameter changes have been recorded since launch.