利用多重分形尺度识别比特币市场状态
文章 arXiv papers · 作者: Josselin Garnier et al.
总结
论文利用多重分形特性估计工具，研究比特币收益在多个时间尺度上的结构。文中指出，收益方差与时间增量之间遵循幂律关系，并以波动率和赫斯特指数表征。由于这些参数会随时间变化，作者引入广义赫斯特指数，以评估是否充分捕捉了多重分形特性。
作者利用局部幂律参数估计和拟合优度识别状态转换，并报告称该方法能自动检测与已知比特币市场事件相关的日期。论文还报告称，尽管价格大幅波动且非平稳，比特币在整个观察历史期间仍保有有序的相关结构。摘要没有提供样本日期、事件列表、参数估计或比较验证，因此这些结论仅适用于所分析的时期和所述方法。
核心观点
- 比特币收益呈现多尺度相关性，可用方差与时间增量之间的幂律缩放关系描述。
- 波动率和赫斯特指数用于表征缩放行为，二者都可能随时间变化。
- 研究提出广义赫斯特指数，用于评估多重分形特性的拟合效果。
- 研究使用局部幂律估计和拟合优度检测状态转换。
- 报告的分析发现了与市场事件相关的状态变化，以及研究历史期间有序的长期相关结构。
标签
全文
# Chaos and Order in the Bitcoin Market # Chaos and Order in the Bitcoin Market The bitcoin price has surged in recent years and it has also exhibited phases of rapid decay. In this paper we address the question to what extent this novel cryptocurrency market can be viewed as a classic or semi-efficient market. Novel and robust tools for estimation of multi-fractal properties are used to show that the bitcoin price exhibits a very interesting multi-scale correlation structure. This structure can be described by a power-law behavior of the variances of the returns as functions of time increments and it can be characterized by two parameters, the volatility and the Hurst exponent. These power-law parameters, however, vary in time. A new notion of generalized Hurst exponent is introduced which allows us to check if the multi-fractal character of the underlying signal is well captured. It is moreover shown how the monitoring of the power-law parameters can be used to identify regime shifts for the bitcoin price. A novel technique for identifying the regimes switches based on a goodness of fit of the local power-law parameters is presented. It automatically detects dates associated with some known events in the bitcoin market place. A very surprising result is moreover that, despite the wild ride of the bitcoin price in recent years and its multi-fractal and non-stationary character, this price has both local power-law behaviors and a very orderly correlation structure when it is observed on its entire period of existence.
在遵守原作品许可的前提下，附作者信息全文展示。 许可协议: abstract CC0
此摘要由 Stratmill 研究智能体根据原文撰写，并非原文副本。