Every retirement calculator gives you the same comforting chart: a line that rises at 7% a year, a spend that comes out of it, and a balance that lasts to 95. The chart is not wrong on average. It is wrong in exactly the years that matter, because markets do not deliver 7% a year; they deliver −38% and then +26% and then 0%, and the order in which those arrive decides whether the money lasts.
This is sequence-of-returns risk, and the Retirement page in Oikonome is built to confront it rather than average it away.
Start with the deterministic model
The page first runs a plain projection from your actual data, not placeholders:
- Age from your birthdate. Target spend defaults to your household's real last-twelve-months spending. Social Security from the per-claim-age estimates on your SSA statement, if you have loaded them.
- Your accounts are grouped into four tax buckets by type: tax-deferred (401k, 403b, traditional IRA and the like), Roth, taxable (brokerage, crypto) and cash. Credit-card debt is netted out first.
- Everything is in today's dollars. Growth is the real return, and spending is a constant real amount.
- Each simulated year: investment buckets grow, cash keeps pace with inflation, required minimum distributions kick in at the age SECURE 2.0 sets for your birth year, and after retirement the year's spend is withdrawn tax-smart, cash first, then taxable, then tax-deferred, then Roth, each haircut by an effective tax rate.
A retirement age is feasible if the money lasts to your plan-to age. The headline is the earliest feasible age, and max sustainable spend is found by binary search: the largest constant spend that just barely survives. When no age works, the page says so and tells you the extra monthly saving that would make 67 work, and the smaller amount that makes 72 work.
Every input is a what-if you can override without saving. Nothing here is advice; it is arithmetic, and the same inputs always give the same answer.
Then replay it against history
The average-return chart above is the base case. The replay is the stress test.
It reruns your exact plan, same buckets, same draw order, same taxes, same RMDs, against every contiguous sequence of real annual returns since 1928: S&P 500 with dividends and 10-year Treasuries, inflation-adjusted, blended by your stock percentage. Sequences wrap at the end of the record so every start year gets a full run. The result is one plain fraction:
Survived 81 of 98 historical start years.
That is a number you can reason about. If the plan survives 98 of 98, the 7% chart was too pessimistic. If it survives 60, the chart was lying to you. Beside the fraction the card gives two spend levels: the spend with 90% historical success, and the spend that survived all of history, and the per-age table carries the same success percentage for every retirement age. Retiring into 1966 is the classic failure the replay catches: modest returns and high inflation for fifteen years, no single crash to point at, and the money runs out in the eighties.
A stress test block then overrides the first years of retirement with three named scenarios: a −35% crash in year one, a crash with slow recovery (−20%, −10%, 0%), and a lost decade of 0% real. The table shows the earliest feasible age and max spend under each, and the "crash-proof spend" is the worst of them, the number you could commit to no matter which of those you draw.
There is no random simulation in this. Monte Carlo draws returns from a distribution that was fitted to the same history and then shuffles them, which quietly assumes the bad years are independent of each other. They were not. Replaying the actual sequences keeps the correlation that made 1973–74 and 2000–02 hurt.
What to do with the number
Three moves change it most, in order of how much they usually help:
- Spend less in the first five years. The replay shows why: a plan that survives a bad start year survives everything. A flexible spend that drops 10% in a down year rescues most of the failing sequences.
- Retire a year later. Each year is one more of saving and one fewer of spending, and it shifts every sequence's worst years earlier in your life.
- Change the stock percentage and see what happens. A higher allocation raises the average and usually raises the survival count too, but not always, and the replay tells you which.
Type the change under adjust, hit Recalculate, watch the fraction move. That loop, more than any single projection, is what the page is for.
Caveats, stated plainly
The model is deterministic, so a plan that reads 98 of 98 is still a plan built on one set of tax rates, one Social Security estimate and your own spend history. It ignores healthcare shocks, pensions with cost-of-living quirks, and the fact that you might want to spend more at 66 than at 86. Cash is priced at 0% real by default, which is the kinder assumption. The Social Security estimate is only as good as the statement you loaded. Read the Assumptions line before you trust the headline.
The full description of the model, including the exact draw order and the tax haircuts, is in the Retirement guide. It is the same page a self-hoster reads, because the self-hosted and hosted products are the same software.