Your backtest shows one ordering of outcomes out of an astronomical number. If the five worst trades happened to cluster at the start rather than scattered through year three, your drawdown would look completely different and you might have stopped trading.
Shuffling the order 5,000 times and recording the maximum drawdown each time gives a realistic risk picture. It is common for a strategy reporting a 12% historical max drawdown to show a 95th-percentile shuffled drawdown above 20%.
The exercise also kills a bad habit: quoting the historical drawdown as the number your position size is built around. It is a single sample from a wide distribution, and it is usually on the kind side.
Related: monte-carlo-simulation, max-drawdown, risk-of-ruin, equity-curve