Methodology
Every calculator on this site that touches investing puts a probability distribution in front of you. This page says which one, how it is built, and where it is wrong.
How the market distribution works
A modelled pot is treated as being multiplied by a LOG-NORMAL factor: normally distributed in log-return space, not in the return itself. Two consequences follow directly from that choice. First, the pot can shrink a lot in a bad period but can never go negative, a real constraint that a plain normal distribution doesn't respect. Second, a full year's return is the PRODUCT of the factors compounding together, not their sum. Most calculators draw twelve of these factors a month at a time and compound them together; a few instead draw one factor per year.
Two numbers steer the whole distribution, both set in the Settings menu: Market Return (the compounded median the distribution is centred on) and Market Volatility (σ, how spread out it is). Turn volatility to 0 and the whole distribution collapses onto the plain deterministic line: the spread disappears, and mean and median become the same number. Almost every calculator computes this distribution directly from its mathematics (a closed form or a probability grid, not a sampled approximation of one), so the bands and histograms are identical every time the page loads; see "The four exceptions" below for the handful of pages that still work differently.
Mean vs median: why the solid line coincides with the dashed P50
On every fan chart the solid line is the median (P50) path. It is the same deterministic calculation the site would show with volatility off, and it coincides with the dashed line in the middle of the fan: half of the modelled distribution's mass sits above it, half below. The arithmetic mean is a DIFFERENT, higher number that isn't drawn on the chart. A lognormal distribution is right-skewed: a small number of very good outcomes pull the mean well above the midpoint of outcomes, so the mean sits above the median. Read the solid/dashed line as "the outcome the typical future actually lands near," and the mean as "the average across every possible future," which is not shown. They diverge more the longer the horizon and the higher the volatility.
The four exceptions
Four calculators carry a seed and a Reroll button: each draws from its own seeded, reproducible random-number generator, not from true randomness. The seed is visible in the URL and in the Reroll button, which simply moves to the next seed. The same seed always produces the same run, which is why a shared link reproduces the exact chart the sender saw, rather than a fresh random draw every time the page loads. Every other calculator on the site has no seed and no Reroll button, because it has nothing random left to redraw. Three of those four (the exploration strategy comparison, the repeated-bets calculator and the portfolio mix, a bootstrap over its stylised 30-year return table) run that simulation AS the model: their headline result is the Monte-Carlo distribution itself, not an approximation of a closed-form one. The fourth, the compound-interest illustration, uses its own seeded draw only to demonstrate the arithmetic/geometric gap on one example path; its actual headline result carries no randomness at all.
Everything is in today's money
Every figure this site produces, whether a 30-year net worth, a monthly surplus or a lifetime advantage, is in today's purchasing power, not in the inflated currency units a calendar would show you decades from now. That keeps every number comparable to a price you'd recognise today.
For that to hold, EVERY rate is a real rate: the growth or interest ABOVE inflation. Market Return in Settings is real, and so is every rate field on every calculator: rent increases, property appreciation, salary raises, loan and mortgage rates, energy price growth, cash interest. Each such field is marked "real". To convert a figure you know in ordinary (nominal) terms, subtract inflation: a 3.5% mortgage with 2% inflation is 1.5% real, a 3% raise is 1% real. Amounts you type in (rent, costs, salaries) are today's amounts; a cost held flat therefore keeps pace with inflation.
What this gives up: the nominal amount that will actually print on a future bank statement is not shown. And a fixed-rate loan is only approximated. Its payment is fixed in nominal money, so in reality its burden shrinks a little each year with inflation, while the model keeps it constant in today's money.
Where this model is wrong
The central objection to models like this one: real markets have FAT TAILS. Crashes far larger than a normal (or lognormal) distribution allows happen far more often than it predicts, and it is the rare extreme run, not the average, that ends up dominating a long-run outcome. A model built on independent, normally-distributed monthly shocks structurally cannot produce a crash as severe or as frequent as history actually has.
Two further gaps compound that one. Real returns are not independent from month to month: volatility clusters (calm periods and turbulent periods each tend to persist), and there is some mean reversion after large moves. Neither is represented here, where every month's draw is independent of the last. And the historical mean this site calibrates to is itself uncertain, estimated from a relatively short run of a survivorship-biased sample of markets that happened to succeed, rather than the full population of markets that could have happened.
The consequence, stated plainly: our P10, the "bad but not catastrophic" line on every fan chart, is probably too optimistic. Treat the fan as a map of ordinary variation around a plausible average, not as a bound on how bad the worst case can actually get.
What this site ignores, in general
- Taxes beyond what an individual calculator models explicitly. Every jurisdiction differs, and this site does not attempt to be a tax advisor.
- Transaction costs and spreads: buying, selling and rebalancing are all treated as frictionless.
- Sequence-of-returns risk clustered around a single retirement date. A bad run of returns right before you need the money hurts more than the same bad run spread evenly across decades, and the modelled distribution here does not single that moment out.
- Personal circumstances such as illness, divorce, job loss or moving country, any of which can matter more than the financial math on this page.
- The fact that every input you type is itself a guess about your own future behaviour, income and spending, and the model can only be as good as that guess.