turtle breakout backtest

What the numbers mean

Entry requires a close above the highest high of the prior forty days; the exit fires on a close below the lowest low of the prior twenty. Qualifying takes twice as long as losing the position, and that asymmetry is the design. There is no profit target and no holding limit, so a trade that keeps working is held until the trend breaks its own twenty-day floor. The rule set accepts more losing trades than winning ones; it only adds up if the average winner is a multiple of the average loser.

Ten ETFs soften that shape in both directions. ETFs are baskets already: they trend more smoothly than single names, and they truncate the right tail the design depends on. Six of the ten are the same equity bet in different wrappers, so a twenty-percent position size with a five-position cap can leave the account fully committed to one macro exposure in the week everything breaks out together, then flat again a few months later. GLD, TLT and EEM are the only genuine diversifiers here.

One deviation from the original matters: the two-ATR stop sets risk per trade but not position size, which stays a flat slice of notional. A semiconductor position and a Treasury position therefore carry identical dollars and very different risk. It is long-only as well, so the record contains long stretches holding nothing. Check whether the worst drawdown is genuinely shallower than owning the basket outright, and what the smoother ride cost in return.