Cowen model finds Bitcoin upside compressing, floor steady

Benjamin Cowen, a quantitative analyst, published a paper in May 2026 using 16 years of daily Bitcoin data through May 2026. The analysis replaces a single power-law line with a fan of quantile regression bands and finds the upper-price quantiles bending inward while lower bands remain approximately straight.

The paper, titled Asymmetric Tail Curvature in Bitcoin Price Quantiles, reports an upper-tail curvature of about -0.33 that is statistically distinct from zero and a lower-tail curvature near -0.02 that is not distinguishable from flat. A bootstrap test run over 27 expanding historical windows keeps the upper-tail curvature clustered near -0.3 and rejects symmetry in every window.

Cowen describes a mechanism he calls diminishing reflexivity. The paper notes that when Bitcoin’s market capitalization was small, relatively modest capital inflows produced large percentage price moves. As market capitalization reached the trillions, the same percentage moves require much larger capital flows, the paper states. In the model, structural demand follows a long-run power-law path while speculative amplitude declines, creating a fan of bands that compresses at the top across cycles.

The paper benchmarks three well-known prior models against the 2019–2026 price record and reports consistent overshoots. An original power-law fit calibrated through 2018 overshot actual prices on 77.2% of trading days with an average error about 32.1% above actual. The stock-to-flow model attributed to PlanB overshot on 94.9% of days with an average error of 294.5%. The S2FX cross-asset model implied an average overshoot of 1,699%, which would project prices above $5 million under simple extrapolation.

Cowen lists several caveats. The effective sample spans four halving cycles, so short sub-windows are weakly identified. The results are sensitive to the model start date: shifting the anchor from January 2009 to January 2010 reduces the upper-tail curvature toward zero. The chosen functional form implies mathematical behavior at infinity that Cowen does not interpret as a real-world price target. The paper also notes that the lower band has been breached in past stress episodes, including events from 2010–2015 and the FTX collapse in November 2022.

The paper does not present the model as a short-term market-timing tool. It frames the result as a single estimated parameter and confidence interval that quantifies the rate at which speculative upside may compress. Cowen retains the four-year cycle framework while reporting a narrower range of plausible peak heights compared with earlier cycles.

Market context in the paper and its public statements notes Bitcoin trading just under $70,000 with a market capitalization above $1.4 trillion. The asset reached a cycle high near $126,080 in October 2025 and has since fallen roughly 44% from that peak and about 33% over the past year. Cowen projects the next cyclical low should arrive in the fourth quarter of 2026, with a base case in October.

In public remarks Cowen wrote: “I have no interest in personal attacks or tribalism. My goal has always been to approach markets…”