Maximum Drawdown
Maximum drawdown is the largest peak-to-trough percentage decline in a portfolio's value before a new peak is reached, capturing the worst loss an investor would have endured.
Maximum drawdown (MDD) measures the deepest fall from a historical high point to a subsequent low point over a given period, expressed as a percentage of the peak value. To compute it, you track the running maximum of the equity curve, measure the percentage drop from that running maximum at every point, and take the largest such drop. A maximum drawdown of -40% means that at the worst moment, the portfolio had lost 40% of the value it had reached at its prior high.
Unlike volatility or standard deviation, which treat upside and downside swings symmetrically, maximum drawdown focuses only on downside pain and on the path of returns, not just their dispersion. This makes it one of the most psychologically and practically relevant risk measures: it answers "how bad did it actually get?" and approximates the loss that might have forced an investor to capitulate or a leveraged account to be liquidated.
Drawdown also has a time dimension worth watching alongside the depth: the drawdown duration is how long it took to recover back to the prior peak. A strategy can have a modest maximum drawdown but take years to recover, which carries real opportunity cost. Investors use MDD to size positions, set risk limits, and judge whether a strategy's headline returns are worth the worst-case experience required to capture them.
hedgewing.ai reports maximum drawdown as part of the institutional risk analytics it generates from nightly walk-forward backtesting, so the figure reflects out-of-sample behavior rather than an idealized fit. It feeds directly into the Calmar ratio (return divided by maximum drawdown) and complements Value-at-Risk and the Sharpe and Sortino ratios in describing the full downside profile of the ensemble's signals.
Related terms
Calmar Ratio · Value at Risk (VaR) · Volatility · Walk-Forward Backtesting
Back to the hedgewing.ai glossary · See AI stock forecasts