Monte Carlo retirement calculator
Your chance that every year’s spending is paid, from thousands of seeded simulations on US market history or on random returns, with the spread of outcomes and when failures happen. It runs the Glideworth planning engine in your browser; nothing you type is sent anywhere.
In 989 of 1,000 simulations (98.9%), every year’s spending of $40,000 was paid for 30 years. The sampling margin on that figure is about ±0.6 points.
- Chance of success
- 98.9%±0.6 points (95% sampling margin)
- Simulations that fell short
- 11Half of them by year 29
- Median portfolio after 30 years
- $1,650,055Today’s money
- 10th percentile after that
- $524,789One simulation in ten ends lower
Portfolio across the simulations, in today’s money
- 10th to 90th percentile (lighter band)
- 25th to 75th percentile (darker band)
- Median (50th percentile)
Show the numbers as a table
| Year | 10th | 25th | Median | 75th | 90th |
|---|---|---|---|---|---|
| 0 | $1,000,000 | $1,000,000 | $1,000,000 | $1,000,000 | $1,000,000 |
| 1 | $865,993 | $933,346 | $1,037,212 | $1,096,963 | $1,164,389 |
| 2 | $793,409 | $884,195 | $1,052,834 | $1,150,413 | $1,240,926 |
| 3 | $825,929 | $876,359 | $1,048,585 | $1,165,064 | $1,331,773 |
| 4 | $754,372 | $874,338 | $1,027,673 | $1,240,757 | $1,355,582 |
| 5 | $772,319 | $853,515 | $1,033,508 | $1,270,890 | $1,453,388 |
| 6 | $739,713 | $886,804 | $1,045,322 | $1,333,235 | $1,565,496 |
| 7 | $720,350 | $845,075 | $1,057,282 | $1,357,449 | $1,661,963 |
| 8 | $657,441 | $862,297 | $1,113,656 | $1,468,697 | $1,762,588 |
| 9 | $605,623 | $858,228 | $1,147,083 | $1,520,541 | $1,791,967 |
| 10 | $622,461 | $893,886 | $1,150,978 | $1,574,782 | $1,924,714 |
| 11 | $549,614 | $844,982 | $1,219,360 | $1,596,377 | $1,963,558 |
| 12 | $545,649 | $831,941 | $1,218,228 | $1,633,163 | $2,135,300 |
| 13 | $520,352 | $840,286 | $1,232,782 | $1,603,690 | $2,168,564 |
| 14 | $521,967 | $817,395 | $1,296,882 | $1,685,477 | $2,256,905 |
| 15 | $585,624 | $804,619 | $1,252,094 | $1,710,323 | $2,185,298 |
| 16 | $525,381 | $855,446 | $1,266,918 | $1,782,903 | $2,562,395 |
| 17 | $528,861 | $830,642 | $1,310,546 | $1,851,595 | $2,380,506 |
| 18 | $551,566 | $781,484 | $1,366,812 | $1,899,668 | $2,587,277 |
| 19 | $491,935 | $809,425 | $1,413,097 | $2,016,053 | $2,644,957 |
| 20 | $476,514 | $827,405 | $1,445,391 | $2,065,371 | $2,708,404 |
| 21 | $478,814 | $808,238 | $1,464,131 | $2,177,719 | $2,920,824 |
| 22 | $465,529 | $809,640 | $1,468,431 | $2,313,861 | $2,853,379 |
| 23 | $469,168 | $802,098 | $1,489,294 | $2,370,244 | $3,090,796 |
| 24 | $464,108 | $723,962 | $1,544,956 | $2,370,264 | $3,177,144 |
| 25 | $465,726 | $778,550 | $1,650,845 | $2,472,577 | $3,271,149 |
| 26 | $452,524 | $818,995 | $1,550,199 | $2,499,998 | $3,497,381 |
| 27 | $513,653 | $884,982 | $1,646,729 | $2,568,720 | $3,494,783 |
| 28 | $513,598 | $922,954 | $1,624,263 | $2,699,769 | $3,498,777 |
| 29 | $498,121 | $922,586 | $1,668,411 | $2,902,648 | $3,814,489 |
| 30 | $524,789 | $967,450 | $1,650,055 | $2,828,850 | $3,844,930 |
Share of simulations that have had a year not fully paid
- Fell short by this year
Show the numbers as a table
| Year | Share |
|---|---|
| 1 | 0.0% |
| 2 | 0.0% |
| 3 | 0.0% |
| 4 | 0.0% |
| 5 | 0.0% |
| 6 | 0.0% |
| 7 | 0.0% |
| 8 | 0.0% |
| 9 | 0.0% |
| 10 | 0.0% |
| 11 | 0.0% |
| 12 | 0.0% |
| 13 | 0.0% |
| 14 | 0.0% |
| 15 | 0.0% |
| 16 | 0.0% |
| 17 | 0.0% |
| 18 | 0.0% |
| 19 | 0.0% |
| 20 | 0.0% |
| 21 | 0.0% |
| 22 | 0.0% |
| 23 | 0.0% |
| 24 | 0.0% |
| 25 | 0.0% |
| 26 | 0.0% |
| 27 | 0.0% |
| 28 | 0.0% |
| 29 | 1.1% |
| 30 | 1.1% |
How this was calculated
- US history: a random historical start year, then real consecutive years, restarting at a random year when the data runs out, from US 1928-2025 (98 years): S&P 500, 10-year Treasuries, T-bills, CPI-U. Portfolio 60% stocks and 40% bonds, rebalanced yearly; real returns and inflation come from the data and results are converted back to today’s money.
- 1,000 simulations with seed 2026: the same seed and inputs always give the same result (engine runMonteCarlo, one derived seed per trial).
- Success: every year’s spending paid in full (the engine’s “all funded” definition). A simulation that is short by a little in its last year counts as a failure.
- Money moves in the middle of each year (half a year’s growth), the planner’s default.
- No tax. The engine’s flat tax mode is set to 0%; the full planner estimates tax for 12 countries.
Results update as you type (simulations run in the background and can be cancelled). The first result shown is for the default inputs, calculated when this page was built.
How the calculator works
Each simulation does three things, using the planning engine’s own Monte Carlo runner:
- Draw a path of returns (and, for history, inflation) for every year of the plan.
- Run the plan along that path: take out the spending (and add any income), grow the portfolio.
- Record whether every year’s spending was paid in full.
The chance of success is the share of simulations that succeeded. The market models differ only in step 1:
- US history, random start years (default): each simulation starts in a random year of US 1928-2025 (98 years): S&P 500, 10-year Treasuries, T-bills, CPI-U and walks forward through real consecutive years, restarting at a random year when the data runs out. It keeps the real relationship between stocks, bonds and inflation within a year and real runs of years.
- US history, 5-year blocks: random blocks of five consecutive years stitched together (a block bootstrap, after Künsch, Annals of Statistics, 1989), which can produce sequences history never did while keeping short-term patterns.
- Random returns: each year’s real return drawn independently from a log-normal distribution with the average and volatility you set. Easy to vary; no runs of bad years unless they happen by chance.
Every simulation is seeded: the run’s seed and the simulation number give each one its own seed, so the same inputs and seed always reproduce the same answer. The seed is shown with the result and can be changed under “More options”.
Reading the result
margin ≈ 1.96 × √(p × (1 − p) ÷ n)
A simulated success rate p from n simulations estimates the model’s true answer with sampling noise. Near 87%, the 95% margin is about ±2.1 points at 1,000 simulations and ±0.7 at 10,000. Two plans that differ by a point are only comparable with the same seed and count.
Worked example
Computed by the calculator’s own code when this page was built, with the plan from the Monte Carlo article: 1,000,000 invested, 40,000 a year of spending in today’s money taken at the start of each year, 30 years, no tax or other income, real returns log-normal with an average of 5% and a volatility of 12%. With 10,000 simulations and seed 2026 the engine finds every year paid in 86.5% of them. The article’s independent script, with its own random numbers, found 86.7%; the difference is within the sampling noise.
| Average real return | Volatility 8% | Volatility 12% | Volatility 16% |
|---|---|---|---|
| 4% | 89.6% | 75.6% | 64.3% |
| 5% | 96.5% | 84.5% | 73.6% |
| 6% | 99.0% | 91.9% | 81.6% |
The same plan under each market model, with the calculator’s defaults (60% stocks, money moving mid-year):
| Market model | Chance of success |
|---|---|
| US history, random start years | 98.9% |
| US history, 5-year blocks | 94.2% |
| Random returns (log-normal 5%, 12%) | 85.9% |
What this assumes and leaves out
- Fixed spending. Every simulation spends the same amount in today’s money whatever happens. A simulation that is short by a little in its last year fails just as one that runs out early does.
- Only the market varies. Lifespan, health costs, tax law and your own behaviour are fixed.
- One country’s history. The history models use US 1928-2025 (98 years): S&P 500, 10-year Treasuries, T-bills, CPI-U. Data: Aswath Damodaran,Historical Returns on Stocks, Bonds and Bills (NYU Stern), and US Bureau of Labor Statistics CPI-U.
- No fees, one portfolio rebalanced yearly, and no tax unless you set the flat rate.
- Timing. Money moves in the middle of each year by default. In the planner’s own check against an established historical calculator (60/40, 4% inflation-adjusted, 30 years, US data from 1928), moving withdrawals from the start to the middle of the year moved the answer from 94.2% to 95.7%, and to the end of the year 98.6%.
Questions
What does an 87% chance of success mean?
That in 87% of the simulated market sequences every year’s spending was paid in full. It describes how a plan with fixed spending behaves under the model’s assumptions. It is not a forecast, and not the chance that you will run out: real people cut spending, delay purchases or earn income when markets go badly.
What chance of success is good enough?
There is no standard. Studies use different targets, and the full planner’s risk-based guardrails default to a target of 80% inside a band of 70% to 95%. A higher figure also means spending you may never use.
How many simulations do I need?
More simulations reduce sampling noise only. Near 87%, 1,000 simulations carry a margin of about ±2.1 points and 10,000 about ±0.7. The model’s assumptions matter far more than the count.
Why does the same seed always give the same answer?
A seed fixes the sequence of random numbers. The engine derives one seed per simulation from the run’s seed, so the same inputs and seed reproduce the result exactly, and comparing two plans with the same seed compares like with like. Change the seed to see how much of a difference is noise.