The nature and novelty of crypto markets have given rise to speculative bubbles, which have permeated almost all cryptocurrencies. This paper shows that the log-periodic model with conditional heteroscedasticity structures has predictive capabilities to estimate the most likely crash date of cryptocurrency bubbles. We use the 2017 bitcoin bubble to perform the primary analysis and date a potential crash just four days before the price peak. We detect the crash date a month before the Bitcoin prices reach their highest value. The bitcoin price fell 30% two weeks after reaching its maximum value. Robustness exercises include the Ether bubble in 2021 and others in Bitcoin’s history to show that the model can be helpful to crypto investors.