What this shows: the rearranged asymmetric quadratic quantile regression fan for Bitcoin (Benjamin Cowen, “Asymmetric Tail Curvature in Bitcoin Price Quantiles”, May 2026), refitted live. In log–log space, log₁₀(P) = c + a·x + b·x², where x = ln(t) − μ is log-time (days since 2009-01-01, centered at the sample mean μ), so the log–time elasticity a + 2bx evolves and b < 0 bends the curve down over time. Curvature is shared within each tail group: lower tail (1%, 10%, 25%) gets one b, the median its own b, and the upper tail (75%, 95%, 99%) its own b — the upper-tail curvature is markedly more negative (the paper: b_U ≈ −0.326 vs b_L ≈ −0.024; asymmetry b_U − b_L ≈ −0.302, block-bootstrap p = 0.012). The seven quantile curves are made non-crossing by the rearrangement estimator (Chernozhukov–Fernández-Val–Galichon 2010): at every date the seven predictions are sorted into ascending order. The dashed gold lines are historical intraday-wick dislocation references: each is the Q1% curve scaled by (1 − k), where k is the live deviation of a major wick low (Aug 2010 $0.032; Aug 2015 $162; Mar 2020 $3,881; Nov 2022 FTX $15,474) from the contemporaneous Q1% estimate. The shaded gold zone spans the full historical dislocation range. Values after the last sample date are a mechanical extrapolation of the quadratic through the projection horizon — no literal long-horizon interpretation. Source: blockchain.info market-price (daily close). Refitted and updated daily by cron.