Where the market numbers come from
Every forecast here rests on assumptions about returns and inflation. Most calculators ask you to supply them. This one does not, so it owes you an explanation of where they come from instead.
Why there is no rate of return field
Section titled “Why there is no rate of return field”Ask someone for their expected return and they will type a single number, usually a good one, and the calculator will then apply it every year for thirty years. That future does not exist. Markets do not deliver 7% annually; they deliver 22%, then -18%, then 9%, in an order nobody controls.
The order matters enormously when you are withdrawing. Two futures with identical average returns can end very differently depending on whether the bad years land early or late. Take money out of a portfolio that just dropped 30% and you sell more shares to fund the same spending, leaving less to recover with. That is sequence of returns risk, and it is most of what separates a retirement plan that works from one that does not. A single average return hides it completely.
So the tool simulates instead, and the assumption you are trusting is the model rather than a number you guessed.
What a simulation contains
Section titled “What a simulation contains”Each of the 1,000 runs is a complete life of your plan against one possible market. For every year it produces six figures:
| Rate | What it covers |
|---|---|
| Stock return | Price appreciation of the stock sleeve |
| Stock dividend | Dividend yield on the stock sleeve, taxed as qualified dividends |
| Bond return | Price change of the bond sleeve as rates move |
| Bond yield | Interest from the bond sleeve, taxed as ordinary income |
| Cash yield | Yield on cash. Cash does not appreciate, it only earns |
| Inflation | The CPI figure for that simulated year |
These move together rather than independently, because in reality they do. A year with a spike in inflation is not free to have whatever bond return it likes.
Investment assumptions
Section titled “Investment assumptions”Each sleeve is modeled as a diversified fund, not as individual holdings.
- Stocks follow a broad US index, the S&P 500. A total-market index fund is close enough. Individual stocks, sector bets, and a separate international mix are not modeled.
- Bonds follow a broad US intermediate-term bond index: Treasuries mixed with investment-grade corporates. Not individual bonds, not high-yield, not TIPS as their own sleeve.
- Cash earns a money-market yield based on 3-month Treasury bills. The model does not assume anyone parks significant savings in a regular bank account that pays almost nothing. If you do, that slice of the forecast is a little optimistic.
Where the raw material comes from
Section titled “Where the raw material comes from”Two ingredients.
Starting levels come from market conditions as of August 2026. Short-term rates, bond yields, dividend yield, and inflation are the jumping-off point, so a projection starts from close to where things actually were then rather than from a long-run average that stopped being true years ago. These starting levels are a dated snapshot that we review and update periodically, not a live feed, so a fast-moving market can get ahead of them.
The year-to-year movement comes from history, specifically US market history from 1928 through 2025 as collected in the widely used Damodaran dataset. Drawing from the real record rather than from a bell curve preserves the things that matter and that tidy statistical distributions get wrong: crashes are deeper than a normal distribution predicts, bad years cluster into multi-year bears, and assets move together in ways a random draw would break apart.
How the result compares with that history: after inflation, simulated stocks earn a little less than US stocks did from 1928 to 2025, about 6.5% a year against 6.8%. Inflation itself averages what it did over that period, about 3.1% a year. The model assumes stocks beat cash by less than they did, because the US record is unusually good by world standards and assuming it repeats is optimistic. It also pays cash a little more than the near-zero return after inflation that cash earned over that period, which offsets most of the difference.
Where the model is deliberately cautious is the bad outcomes. Every simulated year borrows its surprises from a real year, but the years are recombined, so a simulation can string together a 30-year stretch worse than any the US has actually lived through. That is why a plan usually scores a little lower here than when you replay history.
The four market conditions
Section titled “The four market conditions”The Projections tab shows one path at a time. Which one you get is your choice.
| Condition | What it is |
|---|---|
| Average | The 50th percentile outcome. Half the simulations did better. |
| Below-average | The 25th percentile. Three in four did better. |
| Significantly below-average | The 10th percentile. Nine in ten did better. |
| Historical | Not random. A replay of real calendar years. |
An important detail: each of the first three is one real simulation, picked out of the 1,000 by where it finished. It is not an average of anything. That is what lets the Details table and the Cash Flow diagram describe a coherent year-by-year story, with taxes and withdrawals that are consistent with each other, rather than a blend of figures from different runs that never happened together.
The four demo plans all ship on Below-average, which is a good default to spend most of your time in.
Historical playback
Section titled “Historical playback”Instead of simulating, this replays actual calendar years in order starting from a year you pick. Retiring into 1929, 1973, or 2000 are the instructive ones. When it reaches the end of the record it wraps around to a loopback year you also choose.
If the stretch of history you picked did better than more than half of the 1,000 simulations, the Projections tab says so. It is easy to accidentally choose a flattering run of years.
The same 1,000 futures for every plan
Section titled “The same 1,000 futures for every plan”Every plan is simulated against the same set of market histories. Three consequences, all of them useful:
- A cloned plan charts identically to the plan it came from.
- Two unrelated plans are genuinely comparable, because neither got luckier draws.
- When you edit an input, the change you see came from your edit. The market did not move underneath you.
The refresh control on Projections draws a fresh set if you want to see how much the particular sample matters. It lasts for the session only.
The limits of all this
Section titled “The limits of all this”Past performance does not guarantee or indicate future results, and the future is under no obligation to stay inside anything in the historical record. Each simulated year’s surprise is one that actually happened between 1928 and 2025, so the model never invents a shock larger than the largest on record. It rearranges those years into sequences history never produced, some of them worse than anything the US has lived through, but a real future year can still be worse than every past one. Treat a high chance of success as evidence that a plan is robust across a wide range of outcomes, not as a promise, and see what’s modeled and what isn’t for the rest of the picture.