The deepest hole a strategy has dug you into, peak to bottom.
Maximum drawdown is the biggest drop from a high point to the low that follows it on your equity curve. It's the single number that tells you how deep the worst streak got, in R, percent, or dollars. Run your account up to +120R, watch it sink to +75R before it recovers, and your max drawdown was 45R. It decides if you can actually live with a system, because an edge only pays if you're still sitting there when it comes back.
Find the highest point the curve reached, then the lowest point after it before a new high. The gap is your max drawdown. But depth is only one part. A 30R drawdown that heals in 40 trades is a different beast from a 30R drawdown that drags on for 300, and the slow one is the account killer, because that's the one you quit in the middle of the period.
Because the backtest never felt anything. Live, you get slippage, missed fills, and the very human urge to switch the thing off two trades before the bottom. A clean backtest also hides the drawdown you haven't met yet. The worst loss in your sample is almost never the worst loss possible... just the worst one that fit inside your data. Plan for a hole deeper than your backtest's deepest, because give it long enough and you'll dig one.
It depends on your edge and how much you risk per trade, not a magic number that works for everyone. A system risking 1% a trade shows a very different drawdown from one risking 3% on the same trades. Plenty of retail systems run from 2x to 4x their average yearly return in R. If your worst drawdown is deeper than a whole year of gains, the strategy is making you suffer more than it pays.