A strategy has a Sharpe ratio of 1.0 at 12% volatility and a worst drawdown of 8% in two years live. What drawdown should investors expect over ten years, and is a 20% fall in year three evidence it is broken?

Investors should expect a worst drawdown of around 18% over ten years, and a 20% fall is well inside the normal range. About 34% of simulated ten-year paths for a Sharpe 1 strategy at 12% volatility fall 20% or more. The 8% record reflects a short, lucky window: even over two years the median worst fall is 10.8%. A drawdown alone says little; changes in the signal's behaviour say more.

Think of the worst traffic jam you have sat in. After one month of commuting it might be 40 minutes; after ten years it is probably two hours, not because the roads got worse but because you sampled more days. The maximum drawdown is the worst stretch in the record, so it can only grow as the record lengthens. Seravane's 8% is the worst of two years. Even for those two years it was on the kind side: the median worst fall over two years for this strategy is 10.8%, and about 79% of two-year paths go beyond 8%.

Simulate it. Draw 20,000 ten-year paths of weekly returns with a 12% excess return and 12% volatility, and record each path's worst peak-to-trough fall. The median is 17.8%, one path in ten is worse than 26.1%, and one in twenty is worse than 29.2%. In units of annual volatility, the median ten-year drawdown is about 1.5 vols, which is a useful rule to carry: for a Sharpe 1 strategy over a decade, expect the worst fall to be around one and a half times the annual volatility, and plan for twice that. The maximum drawdownThe largest fall from a running peak to a later trough, as a percentage of the peak. It depends on the path and on how long you watch. in the marketing should be replaced by this range.

Not on its own. A 20% fall is 1.67 annual vols, and 34% of healthy ten-year paths contain one. Hitting a drawdown that roughly a third of working strategies hit cannot tell you this one has stopped working. The test that has power looks inside the returns: has the signal's correlation with next-period returns collapsed, has turnover or slippage jumped, has the strategy started losing in the market conditions it used to win in? A drawdown with an unchanged hit rate and unchanged costs is most likely bad luck; one with a changed signal is evidence.

Two limits belong in the answer. The simulation assumes the Sharpe really is 1.0, but two years of data estimate a Sharpe with a standard error of about 1 over the square root of 2, roughly 0.7, so the true figure could be much lower, and a lower Sharpe means deeper drawdowns. It also assumes normal returns; fat tails and volatility clustering make deep drawdowns more common, not less. Both errors point the same way, so the 18% median is a floor for planning, not a ceiling.