A thousand market histories
No market has ever returned 5%, 5%, 5%, year after year, yet that is exactly what a deterministic projection assumes. The difference is not cosmetic. Two portfolios averaging identical returns can end decades apart depending purely on the order those returns arrived in.
This simulator swaps the single line for many. Each run draws a different random sequence of monthly returns around your expected average, with annual volatility fixed at 15%, then plays your entire plan through it: contributions, retirement, withdrawals, the lot. A thousand runs later you have a distribution instead of a guess.
The draws themselves are the model's weakest joint, so hold that thought from the start. Randomly generated returns are not real markets, which have fat tails, momentum and correlated crashes that a simple random draw misses; bad years cluster in ways this simulation does not fully capture, as one scan down the year-by-year historical record makes plain, and the historical backtest is the better check for exactly that. A second quirk is mechanical rather than statistical: the generator is seeded, so the same inputs always give the same chart. Intentional, because a success rate that jiggles on every reload cannot be reasoned about, but it does mean you are looking at one thousand-run sample rather than a fresh one each time.
The 15% is a deliberate choice rather than a slider on this page: it approximates a globally diversified equity portfolio. Middle of the road rather than cautious, mind, and below the 19.4% standard deviation of the historical series this site backtests against. If you want to test that assumption instead of accepting it, the FIRE calculator exposes both the volatility and the run count once its chart is switched to Monte Carlo. A heavy bond allocation swings less than the model assumes and a concentrated or single-country portfolio more, so read the answer as an equity-shaped one.
The fan, the tails and the number on top
The chart shows percentile bands, not individual runs: the middle band is where the bulk of outcomes landed, the outer edges are the lucky and unlucky tails. Width is the thing to read. A narrow fan means the plan behaves similarly whatever markets do, usually because contributions dominate, or the horizon is short. A wide fan means the outcome depends heavily on luck, and the median line in the middle deserves much less confidence than its crispness suggests.
Above the fan sits the success rate: the share of runs that reached your target and still had money at 95. The horizon is deliberately long, and how plausible 95 is depends on where you live: the ONS life expectancy tables cover the UK, and the UN's World Population Prospects carries the same figures for every other country. The rate is reported alongside the fan rather than alone, because a 90% success rate with a catastrophic 10% tail is a different proposition from 90% where the failures are near-misses.
Be precise about what the percentage means, too. An 87% success rate does not mean you have an 87% chance of a comfortable retirement. It means: under this model, with this volatility assumption and this average return, 87% of simulated sequences left money at the end. Fixed withdrawals are assumed here, and the FIRE calculator is where you can switch strategy; nobody actually withdraws rigidly through a downturn anyway: trimming spending in a bad year lifts real-world outcomes well above what a rigid model predicts. Redundancy, inheritance, health, divorce and the state pension all move outcomes more than a few percentage points of simulated success. Mortality sits outside the model as well, so failure at 93 is weighted the same as failure at 70, which overstates risk; Rich, Broke or Dead corrects for that. And there is no tax or fees, as everywhere on this site.
So treat anything above roughly 85% as "this plan has genuine margin", and stop optimising for the last few points. Not as a pass mark: the figure moves several points on the volatility you assume and on the strategy you pick, so read it as a band rather than a grade. A 100% rate is certainly not the goal. Reached by saving, it almost always means you over-saved or worked too long, and chasing 99% usually means working years longer to insure against futures that mostly involve you being dead. Reached by switching to a flexible withdrawal rule, it means something else entirely: those rules cannot run the pot to zero, so they score near 100% by construction and pay for it with sharp cuts in bad years. The sensitivity analysis shows what those extra years actually bought, and reading a Monte Carlo run goes further into which number to aim at and what it does not promise.
Fourteen points of Nadia's luck
Nadia is 44, has £340,000, adds £1,600 a month, expects 5% real and wants £33,000 a year at 3.75%. Her deterministic projection says FIRE at 55, comfortable to 95. Through the simulator, the success rate is about 72%: the average future is fine, and more than a quarter of the simulated ones are not, which is exactly the gap between the two kinds of answer.
Push the same plan through the engine at 20% volatility, the kind a more concentrated, equity-heavy portfolio carries, and success falls to around 58% while the fan widens markedly at the top and bottom. Same expected return, same contributions. The only change is how bumpy the road is assumed to be, and it costs fourteen points of success. The tool holds volatility at 15% precisely because an assumption you cannot see is doing that much work. Fixed rather than left to optimism, though 15% is a moderate setting and not a cautious one, so 20% is closer to a correction than to a stress test.
Bumpiness cuts the other way as well, which surprises people who add bonds expecting safety. Over very long horizons, low-return assets can raise failure risk rather than lower it: the portfolio grows too slowly to outpace inflation and withdrawals. Volatility protection and longevity protection pull in opposite directions.
A thousand random futures is one way to stress a plan. The other is to pick the single worst plausible future and stare at it. The crash test does that: a crash landing in year one, which is the scenario that does the most damage.