Two retirees earn the same average return and one runs out of money: why the order of returns dominates retirement outcomes, which historical cohorts it broke, and how to test a plan against it honestly.
Two retirees each average 6% a year over thirty years. One dies wealthy; the other runs out at 82. The difference isn't return, cost or spending — it's order. Losses in the first years of drawdown are lethal because each withdrawal in a down market sells more units, and those units never recover. The same losses in year twenty barely matter. Accumulators experience the average; decumulators experience the sequence.
Run every rolling retirement window through market history since 1928 and the failures cluster tightly. The 1929 cohort you'd guess. The worse one you might not: the mid-1960s — average returns over their thirty years were respectable, but the 1966–81 inflation grind forced inflation-swollen withdrawals into a stagnant market for fifteen straight years. Sequence risk is usually told as a crash story; history says it's just as much an inflation story.
Averages can't see this, so honest testing takes two forms. Simulation: thousands of market paths with fat-tailed shocks and inflation that persists once it takes hold — the features that made 1966 lethal — giving a success rate rather than a single line. Replay: your actual plan walked through every historical start year, so you can point at exactly which pasts it survives. Hundred Summers' Stress Test tab does both on your own plan — the backtest runs every window since 1928 — and the deterministic mitigations (a guaranteed floor via the State Pension, deferral or an annuity; a cash buffer; flexible spending in bad years) can each be modelled and re-tested rather than taken on faith.