The 4% rule was tested on 30-year retirements. Yours might run 50. So I pointed my calculator's engine at every half-century window in market history: 48 sequences, real returns, no shortcuts. The rule failed one start in seven. Here's what didn't fail.
The 4% rule is the most quoted number in retirement planning, and it was never tested for the retirement most FIRE people are actually planning. Bengen's 1994 paper and the Trinity Study both used 30-year windows, the length of a traditional retirement starting at 65. Retire at 40 and you need 50. So I ran the test nobody runs: my calculator's stress-test engine, pointed at every full 50-year retirement that has existed since 1928, real inflation-adjusted S&P 500 returns, year by year. Forty-eight complete windows, 1928 through 1976 starts. Here's what survived.
| Withdrawal rate | 30-yr windows survived | 40-yr | 50-yr |
|---|---|---|---|
| 3.0% | 68/68 | 58/58 | 48/48 |
| 3.5% | 68/68 | 58/58 | 47/48 |
| 4.0% | 65/68 | 54/58 | 41/48 |
| 4.5% | 61/68 | 47/58 | 35/48 |
| 5.0% | 55/68 | 44/58 | 31/48 |
Portfolio-only, no cash buffer. All-equity inflation-adjusted S&P 500 returns 1928–2024, constant real-dollar withdrawals, no fees. Full windows only.
Read the 50-year column. The 4% rule, the one every article quotes, failed 7 of the 48 historical half-century retirements, about one in seven. And the failures aren't the ancient history you'd guess. Starting years 1929 and 1930 fail, sure. But so do 1965, 1966, 1968, 1969 and 1973: ordinary-looking years where a decade of flat markets plus inflation quietly bled the portfolio before the 1980s bull could save it. Stretch the horizon from 30 to 50 years and the rule's failure rate goes from about 4% of starts to about 15%.
I searched for the maximum rate that survived every window, the SAFEMAX, in 0.05% steps, run two ways: the bare portfolio, and the way my calculator actually models a plan, with a cash buffer of 2.5 years of spending carved out of the same starting wealth. The buffer strategy is the calculator's published one: spend from cash in down years so you're not selling shares at the bottom, refill it from gains in strong years.
| Horizon | SAFEMAX, portfolio only | SAFEMAX, with 2.5-yr cash buffer |
|---|---|---|
| 30 years | 3.75% | 3.95% |
| 40 years | 3.60% | 3.75% |
| 50 years | 3.45% | 3.55% |
Two things in that table are the whole reason this site defaults to what it defaults to. First, the horizon penalty is real but smaller than the doom version: going from 30 to 50 years costs you about 0.3 to 0.4 points of withdrawal rate, not half your lifestyle. Second, the buffer buys you measurable safety at conservative rates: at 3.5% over 50 years, the bare portfolio fails exactly one starting year, 1929, and the buffered version fails none. Forty-eight out of forty-eight. That single failing year is the entire gap between 3.5% and a perfect record, and 2.5 years of cash closes it. (How big the buffer should be is its own question; I tested every size from zero to five years.)
Because the buffer is carved out of the same starting wealth, it's cash that isn't compounding, and at aggressive withdrawal rates that drag starts to show. At 5% over 50 years the buffered plan actually survived fewer windows than the bare portfolio, 29 versus 31, because plans failing on raw withdrawal size need every dollar invested, not parked. The buffer is sequence-risk insurance: it shines at 3.5 to 4%, where bad early years are what kill you, and it stops helping when the rate itself is the problem. I'd rather show you that result than hide it.
The simulation is the same runHistoricalSequence function the calculator runs on your own numbers, not a separate model built for a blog post. Data: inflation-adjusted (“real”) S&P 500 total returns, 1928 through 2024, the same 97-year series in the calculator's source. Withdrawals: a constant real amount, set as a percentage of starting wealth, taken annually. Buffer runs: cash equal to 2.5 years of withdrawals carved from starting wealth, drawn in negative-return years, refilled from gains in years above +10%. Windows: full horizons only, so 68 thirty-year, 58 forty-year and 48 fifty-year sequences; no partial windows, no survivorship shortcuts. No fees, no taxes, all-equity, which is the convention of the original studies and makes the rates here ceilings rather than promises. Taxes are exactly why the scenario pages land below these numbers for real spending. Worth saying plainly: 48 historical windows is everything that exists, and they overlap heavily, so this is the complete record rather than 48 independent trials.
Where this lands against the literature: Bengen's original SAFEMAX was about 4.1% for 30 years on a stock-bond mix, and later reassessments have moved it around as the data grew. My 30-year all-equity figure of 3.75% bare, 3.95% buffered, sits right in that neighborhood, which is the sanity check. The 50-year numbers are the part the literature mostly doesn't cover, and they're the reason my calculator defaults to 3.5% with a cash buffer: that exact configuration is the highest one with a perfect historical record over half a century.
Run your own plan against all 48 sequencesNot financial advice. Consult a fee-only fiduciary CFP for personalized guidance. Tax figures use 2026 brackets.