This paper investigates risk measures based on drawdown and maximum drawdown during the rising phase of speculative bubbles that end in large endogenous crashes. Building on prior work that identifies such bubbles via log-periodic power laws (LPPL), the authors analyze the duration and size of drawdowns (consecutive price declines) in 17 historical bubble episodes across global stock indices. They define pure drawdown (DD) as the cumulative loss from a local peak to the next local minimum, and measure drawdown duration as the number of days in that decline. The study finds that drawdown duration during bubble rising phases follows a power law distribution with an exponent close to 1, indicating a heavy-tailed behavior. Using the generalized Pareto distribution (GPD), they fit the cumulative probability of drawdown size conditioned on duration, revealing that bubble phases exhibit lower scale parameters (β) and higher shape parameters (ξ) compared to the entire time series, with non-overlapping confidence intervals. The expected maximum drawdown (E[M]) and the worst-case drawdown (D̄) are also computed, showing that bubble rising phases have narrower ranges and lower values than the full datasets, though some overlap exists for indices with severe internal drops. The authors argue that incorporating drawdown duration into risk measures provides a more nuanced view of market risk during bubbles, as long sequences of declines can trigger investor withdrawal. The results suggest that risk profiles during speculative bubbles are distinct from normal periods, offering potential for embedding these measures into predictive frameworks for crash timing. This work contributes to quantitative finance by emphasizing the role of time in drawdown risk and providing empirical benchmarks for bubble-specific risk assessment.