donchian breakout backtest

What the numbers mean

The Donchian rule is reactive rather than predictive, and the shape of the returns follows directly from that. Entry only fires after the close has already cleared the highest high of the prior twenty sessions, so every position starts late. The exit only fires after the close breaks the prior ten-day low, so every position ends late too. The distance between a trade's peak and the ten-day low beneath it is handed back on every single winner, by design. That giveback is most of what the drawdown figure is measuring here — less a run of bad trades than the structural cost of never selling anywhere near a top.

The two clocks are deliberately asymmetric: twenty days to get in, ten to get out. Being quicker to leave than to arrive keeps the losing tail short, which is why the hit rate lands closer to a coin flip than the usual breakout caricature of many tiny losses and one rare enormous win. The payoff still comes from size rather than frequency. The 8% stop sits underneath names that can travel that far on a single earnings gap, so it frequently resolves a trade before the channel gets a vote at all, converting what would have been a channel exit into a fixed-size loss.

Concentration is the other defining parameter. Twenty percent per position across five slots means a fully committed book of five correlated mega-caps whose breakouts tend to fire in the same week. Read the win rate and the profit factor together — the gap between them is the whole asymmetry, and either number alone misleads. Then read drawdown against the annualised figure, and keep in mind there is no take profit and no holding limit: the right tail did the heavy lifting, and it was paid for with long flat stretches in cash when nothing was making new highs.