AI 中文总结
该研究指出比特币价格指标随时间退化，而减半时间结构稳定，发现自创世以来的时间幂律是唯一符号稳定的信号，预先注册了2026年底部及下次减半后525-546天的顶部窗口。
AI 中文摘要
每一个被广泛关注的比特币周期指标（Pi Cycle、MVRV、Mayer、Puell）在十年间都曾精准发出转向信号，随后按“精准、提前、失效”的序列退化，这是一种结构性现象。在四个减半周期（2011-2026年）中，五个顶级信号振荡器的周期峰值单调下降，而谷值则逐步升高，因此任何基于过往周期校准的阈值必然会停止触发；短期指标衰减至零，部分还会反转符号；但比特币的时间结构保持固定，成熟周期的顶部出现在对应减半后的525/546/534天，底部则出现在顶部后的406/364/366天。转向信号通过固定机械规则事后识别，而非实时记录。无时间偏差的零假设下，各类变体的联合聚类概率为5e-6至1e-3；更严格的经验零假设（相同规则下的块自举路径）从未复现其确定性构建的顶部聚类（10000次中0次）；底部聚类主要由回撤过程内在决定（路径占比31%-43%），因此证据集中于顶部相位对齐。在块高度（精确的210000块单位）下，顶部零假设仍为10000次中0次，仅出现部分底部结构；形状和波动率叠加无改进。自创世以来的时间因果幂律（指数约5.6）是唯一在成熟周期间符号稳定的信号，在第二个数据源和以太坊上可复现，且在当前周期中其战胜买入并持有策略的时间优势转为正向（仅一个保留样本，具暗示性而非决定性）。我们不依赖单周期显著性：轮换零假设显示HAC推断在此处过度拒绝（名义0.05水平下大小为0.33，p=0.21）。宏观驱动因素（M2、收益率曲线）表现出同样的不稳定性，在联合赛马中落败。我们预先注册了可证伪的窗口：2026年底部（10月5日-11月16日）和下一个顶部在下次减半后的525-546天。
英文摘要
Every widely followed Bitcoin cycle indicator (Pi Cycle, MVRV, Mayer, Puell) called turns precisely for a decade, then degraded in one sequence: precise, then early, then silent. This is one structural phenomenon. Across the four halving epochs (2011-2026), the per-cycle maxima of five top-calling oscillators decline monotonically while minima end higher, so any threshold calibrated on past cycles must stop firing; short-horizon indicators decay toward zero and several invert sign; yet Bitcoin's time structure stays fixed, with mature-cycle tops 525/546/534 days after their halvings and bottoms 406/364/366 days after their tops. Turns are identified retrospectively by a fixed mechanical rule, not a real-time record. Timing-free nulls put the joint clustering at 5e-6 to 1e-3 across every variant. A harder empirical null (block-bootstrapped paths under the identical rule) never reproduces the top cluster under its deterministic construction (0 of 10,000); the bottom cluster is largely intrinsic to the drawdown process (31-43% of paths), so the evidence concentrates in top phase-alignment. In block height (the exact 210,000-block unit) the top null stays 0 of 10,000 and partial bottom structure emerges; shape and volatility overlays do not improve. A causal power law in time-since-genesis (exponent near 5.6) is the only signal whose sign is stable across mature epochs, replicates on a second source and Ethereum, and whose timing edge over buy-and-hold turns positive in the current cycle (one holdout, suggestive not decisive). We rest nothing on per-epoch significance: a rotation null shows HAC inference over-rejects here (size 0.33 at nominal 0.05; p=0.21). Macro drivers (M2, yield curve) show the same instability and lose a joint horse race. We pre-register falsifiable windows: a 2026 bottom (Oct 5-Nov 16) and a next top 525-546 days after the following halving.
Comments32 pages, 11 figures. Code and data: https://github.com/BitcoinDaily/bitcoin-runs-on-a-clock
