Every trading bot, no matter how good, will eventually hit a drawdown: a period where losses pile up and the account falls from its peak. Drawdowns are normal. What separates traders who survive them from traders who blow up is having a plan decided in advance for what to do when they arrive. Without a plan, a drawdown triggers panic, and panic triggers the worst decisions in trading: increasing size to recover, disabling safety limits, or abandoning a good strategy at its lowest point.

A drawdown is measured from the account's highest point to its lowest point after that peak, expressed as a percentage. A ten percent drawdown means the account is ten percent below its best value. Small drawdowns, under ten percent, are routine noise for most strategies. Moderate drawdowns of ten to twenty percent demand attention but are not necessarily a sign of failure. Drawdowns beyond twenty percent, or beyond the maximum the strategy showed in backtesting, are a serious warning that something may have changed.

The first job during a drawdown is diagnosis, not action. Ask whether the market regime has changed: a trend-following bot will struggle in a sideways market, and a mean-reversion bot will struggle in a strong trend. Check whether the bot is behaving as designed, or whether it has developed a fault, such as trading too frequently, ignoring its stop losses, or trading pairs it should not. Check external factors too: a change in exchange fees, a period of extreme volatility, or a strategy that has become crowded can all degrade performance without any fault in the bot itself.

Pausing is the right response to a normal drawdown within the strategy's historical range. Stop live trading, switch the bot to paper mode, and observe whether performance recovers as market conditions normalise. Pausing costs nothing except opportunity, and it prevents the most common drawdown mistake, which is making changes in the middle of a losing streak based on emotion rather than evidence.

Retuning is appropriate when you have identified a specific, fixable cause. Perhaps the bot's stop losses are too tight for current volatility, or a particular pair is dragging down results, or the strategy needs a filter for news events. Make one change at a time, test it in paper trading first, and compare results against the original configuration over a meaningful sample, not three trades. Retuning based on a handful of recent losses is curve-fitting in disguise: it optimises for the past week and usually fails the next one.

Quitting is the right response when the drawdown exceeds the strategy's worst historical case by a wide margin, when the cause cannot be identified, or when you no longer trust the bot's behaviour. Quitting is not failure. Keeping a broken strategy running out of stubbornness is failure. A trader who quits a failing bot preserves capital for the next opportunity. A trader who rides it to zero preserves nothing.

The key to all three responses is deciding your limits before the drawdown happens. Set a soft limit, for example a fifteen percent drawdown, at which you pause and investigate. Set a hard limit, for example a twenty-five percent drawdown, at which you stop the bot entirely and do not restart it without a full review. Write these limits down and, if your bot platform supports it, configure them as automatic kill switches so that the decision is enforced even when you are asleep or away from your desk.

Also set a time-based rule: if a bot has not recovered to a new high within a set period, such as three months, review it regardless of the drawdown size. A strategy that grinds sideways for months while charging subscription fees is losing money slowly, and slow losses deserve the same scrutiny as fast ones.

The hardest part of drawdown management is emotional. Watching an account fall triggers a powerful urge to act, and most actions taken in that state make things worse. Two habits help. First, review the bot's performance weekly, not hourly. Constant monitoring amplifies every fluctuation and invites impulsive changes. Second, keep a written log of decisions made during drawdowns, including the reason for each decision. Reading that log later teaches you which of your instincts were useful and which were costly.

Drawdowns are not a sign that automation has failed you. They are a feature of every trading approach ever devised. The traders who last are the ones who planned for drawdowns before they arrived, responded with a process rather than a panic, and knew when to pause, when to retune, and when to walk away.