The power-law model relates Bitcoin's price to its age. But the recovery forecasts in a Nov. 10 preprint also assume that deviations from the long-run curve will reverse, a tendency the paper's historical analysis does not resolve.
Authors Carlos Baquero and Raquel Menezes argued that Bitcoin's long-run growth rate has slowed under the conditions their test permits.
The study's price data end on Sept. 19, so its findings do not account for November's selling.
In a June 5 analysis, model advocate Stephen Perrenod made the attractor argument explicit. He said tests of price deviations supported an attractor around which bubbles and bear markets occur.
He also cautioned that the tests neither predicted next month's price nor proved the relationship would continue indefinitely.
Reports on power-law models have likewise described an attractor framework while leaving unresolved what sustains it.
Baquero and Menezes analyzed 820 weekly observations from January 2011 through September 2026. For a specified family of possible price deviations, their test found evidence that the long-run growth rate was slowing.
At a nominal significance level of 5%, the test rejected constant drift in expected log price, a result compatible with a power-law trend but not unique to it.
Their main model family allowed shocks to persist as much as a random walk, in which a shock need not fade. When a prespecified sensitivity check allowed changes to persist even longer, the rejection disappeared, showing that the conclusion depended on the persistence assumption.
The paper also found that a Gompertz curve, which approaches a ceiling, fit the historical record almost as well as the power law.
The recovery is built into the forecast
The authors said their fitted models for price deviations revert toward their trends. As a result, most forecasts project a rise from the September endpoint of approximately $81,000.
The selected models build in a tendency to return toward the trend, even though the historical analysis cannot establish that tendency within the approach it evaluates.
For Dec. 28, 2030, the power-law model gave a median of $383,000 and a 90% interval from $150,000 to $1,015,000. The medians across all five trend models ranged from $85,000 to $1,011,000.
The authors cautioned that their predictive intervals are probably too narrow and are not calibrated. Each forecast combines the fitted distribution with simulated and refitted alternatives, but those refits explore only nearby choices for how deviations behave.
As a result, broader uncertainty about persistence is only partly captured.
The model assumes a pull toward the trend, so its projected rebound cannot serve as evidence that such a pull exists.
The paper calls itself a preregistered reanalysis of a record the authors had already studied. They said they froze the protocol and posted its hash on Oct. 6, applied the historical test once on Oct. 7, then registered the results and forecasts later that morning.
The forecasts cover the last Saturday of each year from 2027 through 2030. They will be assessed jointly with a log-density score, which measures how well the predicted distributions match the eventual prices.
Better performance would demonstrate better prediction for those four observations. The distinction matters for the case for a Bitcoin recovery: a historical growth curve can frame a forecast, but an assumed return toward it remains a proposition to be tested against future prices.

