bollinger reversion backtest

What the numbers mean

The win rate here is high by construction, and that is mechanical, not a compliment. Entry demands a close more than two standard deviations below the 20-day average; the exit asks only that price recover to that same average. The target sits close to the entry, and it drifts down toward price as the average follows the decline, so the bar each trade must clear keeps lowering itself. Expect many small, quickly resolved wins punctuated by fewer and larger losses: positions that never snap back have no stop to cut them, only the fifteen-day clock.

The 200-day filter is the load-bearing rule. It refuses the setup unless the stock is still above its long-term average, keeping the strategy out of genuine collapses — and standing it down during the stretches when two-sigma dislocations are most common. What is left spends most of the tested window in cash: at most five concurrent positions drawn from a ten-name list, each a tenth of the account, triggered only by a rare condition. Shallow drawdowns follow from that idleness rather than from any risk control in the rules.

Whatever the return line shows, it was produced by capital that sat out most of the window, so read it against exposure, not against holding the same ten names outright. Compare win rate with profit factor: together they describe how much larger the average loss is than the average win. Then check whether trades cluster into a few volatile stretches, because a decade of history built from a handful of busy months is thinner than the trade count suggests.