[Paper Review] Dissection of Bitcoin's Multiscale Bubble History from January 2012 to February 2018

Jan-Christian Gerlach, Guilherme Demos+1 more|arXiv (Cornell University)|Apr 17, 2018

Complex Systems and Time Series Analysis4 citations

TL;DR

This paper proposes a multiscale bubble detection framework for Bitcoin from January 2012 to February 2018, combining automated peak detection, Lagrange-regularized regime change detection, and LPPLS modeling with k-means clustering of critical time predictions. It identifies 3 major and 10 minor bubbles, with LPPLS confidence indicators and clustered critical times providing early warnings of imminent crashes.

ABSTRACT

We present a detailed bubble analysis of the Bitcoin to US Dollar price dynamics from January 2012 to February 2018. We introduce a robust automatic peak detection method that classifies price time series into periods of uninterrupted market growth (drawups) and regimes of uninterrupted market decrease (drawdowns). In combination with the Lagrange Regularisation Method for detecting the beginning of a new market regime, we identify 3 major peaks and 10 additional smaller peaks, that have punctuated the dynamics of Bitcoin price during the analyzed time period. We explain this classification of long and short bubbles by a number of quantitative metrics and graphs to understand the main socio-economic drivers behind the ascent of Bitcoin over this period. Then, a detailed analysis of the growing risks associated with the three long bubbles using the Log-Periodic Power Law Singularity (LPPLS) model is based on the LPPLS Confidence Indicators, defined as the fraction of qualified fits of the LPPLS model over multiple time windows. Furthermore, for various fictitious 'present' times $t_2$ before the crashes, we employ a clustering method to group the predicted critical times $t_c$ of the LPPLS fits over different time scales, where $t_c$ is the most probable time for the ending of the bubble. Each cluster is proposed as a plausible scenario for the subsequent Bitcoin price evolution. We present these predictions for the three long bubbles and the four short bubbles that our time scale of analysis was able to resolve. Overall, our predictive scheme provides useful information to warn of an imminent crash risk.

Motivation & Objective

• To systematically identify and classify long- and short-term price bubbles in Bitcoin from January 2012 to February 2018.
• To develop a robust, automated method for detecting drawups and drawdowns in Bitcoin price time series using Lagrange regularization.
• To assess the predictability of major bubbles using the Log-Periodic Power Law Singularity (LPPLS) model and confidence indicators.
• To forecast critical crash times by clustering LPPLS-predicted critical times across multiple time scales.
• To link bubble dynamics to socio-economic drivers such as search volume, user activity, and speculative behavior.

Proposed method

• Applies an automated peak detection algorithm to segment Bitcoin price time series into uninterrupted drawups and drawdowns.
• Uses Lagrange Regularization to detect the start of new market regimes, ensuring robustness against noise and false positives.
• Employs the LPPLS model to fit bubble dynamics, with confidence indicators defined as the fraction of qualified fits across sliding time windows.
• Applies k-means clustering to group predicted critical times $t_c$ from LPPLS fits across different time scales to identify plausible crash scenarios.
• Uses the Silhouette metric to determine the optimal number of clusters in the k-means algorithm, ensuring meaningful grouping of predicted critical times.
• Integrates socio-economic data (e.g., search queries, network users) to interpret the drivers behind observed bubble formations.

Experimental results

Research questions

• RQ1 What are the key structural regimes (drawups and drawdowns) in Bitcoin’s price evolution from January 2012 to February 2018?
• RQ2 How accurately can the LPPLS model detect and predict the end of major and minor bubbles in Bitcoin’s price history?
• RQ3 What is the predictive power of clustered critical times $t_c$ derived from multiscale LPPLS fits for forecasting imminent market crashes?
• RQ4 How do socio-economic indicators such as search volume and user activity correlate with the onset and collapse of Bitcoin bubbles?
• RQ5 To what extent is the start time of a bubble ($t_1^*$) stable across different analysis windows, supporting the reliability of early detection?

Key findings

• The study identifies 3 major bubbles and 10 minor bubbles in Bitcoin’s price history from January 2012 to February 2018, with the largest bubble peaking in mid-December 2017.
• The LPPLS confidence indicator reached high values (over 80%) during the lead-up to the December 2017 crash, indicating strong model fit and high predictability.
• For the December 2017 bubble, the k-means clustering of predicted critical times $t_c$ revealed two dominant clusters, suggesting two plausible crash scenarios with critical times within a 10-day window.
• The method successfully detected the start of the 2017 bubble with high stability, as evidenced by consistent $t_1^*$ estimates across varying analysis windows.
• The Silhouette metric identified optimal cluster counts of 2–3 for most bubbles, validating the robustness of the clustering approach in grouping critical time predictions.
• The analysis confirms that Bitcoin’s price dynamics are strongly influenced by speculative behavior, with strong feedback loops between price surges, search volume, and user growth.