Are cryptocurrencies real financial bubbles? Evidence from quantitative analyses

Battle Count: 50

Highly relevant for risk management and bubble detection in cryptocurrency trading. The combined JLS+PSY framework provides early warning signals for crashes (e.g., Bitcoin mid-Dec 2017, Ether mid-Jun 2017). However, the signals are probabilistic and not deterministic trading rules. The methodology is more suited for risk assessment and position sizing than for generating direct buy/sell signals. The general applicability to any financial time series adds value for multi-asset quantitative strategies. The paper does not provide backtested trading strategies or performance metrics.

High complexity due to: (1) Multiple model implementations (JLS with OLS/GLS/MLE, PSY with BSADF/BSADF*), (2) Non-linear optimization of LPPL requiring Genetic Algorithms for global optimization, (3) Multiple time-window calibration strategy with 20 sub-windows per date, (4) Monte Carlo simulations for PSY confidence levels, (5) Sornette bounds validation, (6) Significant computational resources required for repeated calibrations across all date-window combinations. The authors note the numerical effort is 'relevant and requires huge computational resources.'

The methodology is well-described with specific parameters (window lengths: 72 for Bitcoin, 69 for Ether; 20 sub-windows; Sornette bounds). However, no code or repository is provided. Data sources (Bloomberg XBTUSD Curncy, Etherscan, Coinmarketcap) are commercial or require access. The Genetic Algorithm implementation details and Monte Carlo simulation parameters are not fully specified. The authors reference prior work (Bianchetti et al., 2016) for methodology validation.