Own the whole US market. A three-check trend test, run once a month, decides how much of it you hold; the rest sits in short Treasuries. Built for a ten-year horizon, where surviving a deep fall matters more than winning a good year.
Loading…
Enter what your account is worth. Nothing you type is sent anywhere — it stays in this browser.
SPY, 1994–2026 — a window that contains both the dot-com bust and 2008. Decisions are monthly; the falls below are measured daily, which is how your account is marked. Costs and taxes are not modelled; the rule traded roughly once every two years.
| Approach | Return / year | Worst fall | Years to recover |
|---|---|---|---|
| Buy and hold the index | 10.74% | −55.2% | 8.4 |
| This ladder | 10.43% | −26.8% | 3.3 |
“Years to recover” is arithmetic, not a forecast: a 55.2% fall needs a 123% gain to get back to even, which takes 8.4 years at 10% a year. That is why the drawdown column matters more than the return column when your horizon is ten years — the deep hole spends most of them digging out.
On the numbers themselves: these falls were previously published as −50.8% and −17.0%. Those came from an equity curve measured only at month ends, which cannot see a drop that starts and finishes inside one month, and it flattered both lines by roughly ten points. The figures above are the daily ones. The conclusion did not change — the ladder still roughly halves the fall for about 0.3 points of annual return — but the number you would actually have watched is the bigger one.
It does not predict returns. Across 1994–2026 the next month’s return was not ordered by rung — the 67% rung actually beat the 100% rung, and the correlation between allocation and next-month return was +0.083, which is statistically indistinguishable from zero. What the rungs do order is the share of months that finished positive (48% / 53% / 71% / 69%). The rule separates fragile tape from healthy tape; the drawdown benefit is mechanical, not clairvoyant.
It will lag in a straight bull market. It sits partly out of the market about 30% of the time. In a decade with no crash, buy-and-hold wins and this looks like dead weight. You pay that cost for years to collect in one bad year — which is what insurance feels like.
A 32-year average is not your ten years. Any single decade can look nothing like the average. The −26.8% worst fall is what happened, not a limit on what can happen.
Treasuries are not risk-free either. The parked sleeve is short-dated precisely to keep that risk small, but even short Treasuries lose value when rates rise — in 2022 stocks and bonds fell together, so the defensive half offered no shelter. Anything longer-dated would have been considerably worse.
The hard part is not the maths. A rule that says “hold a third” during a crash is easy to backtest and genuinely difficult to follow. Educational information only — not financial advice.