Market Time Machine: Historical Portfolio Simulator

What the Market Time Machine shows

This tool takes a portfolio you design and runs it through the actual returns of the last hundred years — the Great Depression, the war years, the 1970s inflation, the dot-com collapse, 2008, the pandemic. You are not looking at a projection. You are looking at what your allocation would have done to real money through the worst decades on record, next to the benchmarks it is usually compared against.

How it works

You set the percentage of your portfolio in five asset classes — stocks, bonds, real estate, gold and cash — and pick a start year. For every year from then to 2025, the last full year, the tool takes that year's actual return for each asset class, weights them by your allocation, and applies the combined figure to your running balance. It does the same for five reference portfolios at once: a classic 60/40, Ray Dalio's All Weather, Harry Browne's Permanent Portfolio, and an aggressive and a conservative mix.

Because your target weights are re-applied to every single year, the model is doing something specific: it rebalances back to your target allocation annually. That is not a neutral detail. Rebalancing is what forces you to sell whatever just went up and buy whatever just fell, and over a century it is responsible for a meaningful share of the difference between a mixed portfolio and the sum of its parts. The events overlaid on the chart are there so a number has a story attached to it — a 43% fall is abstract until it is labeled 1931.

You can also add a yearly contribution. It goes in at the start of every year, before that year's return, and the five reference portfolios get exactly the same amounts on the same dates. Once money is being added, final value divided by starting capital is no longer a return, so the cards switch to the time-weighted return — the portfolio's own annual growth, whenever the money arrived — and the panel under them adds the money-weighted return, which is what your money actually earned, and what the same total would have become had it all gone in on day one.

What it assumes

• Returns are nominal unless you switch to the Real view, which takes out each year's US consumer price inflation (CPI-U, December to December). The gap is widest in the 1940s and the 1970s: a portfolio that grew 8% in 1979 lost about 5% of its purchasing power that year, because prices rose 13.3%.
• Rebalancing back to target happens every year, free and instantly. In a taxable account, rebalancing realizes gains and costs money.
• No fund fees, no trading costs, no taxes and no dividend withholding. A single percentage point of annual cost compounds to roughly a quarter of the final balance over forty years.
• The asset classes are broad indices. "Stocks" is a whole market, not the shares you actually own, and no index existed to buy for most of the period being modeled.
• The data are US calendar-year total returns from 1920 to 2025: the S&P 500 with dividends, the 10-year Treasury bond, 3-month Treasury bills and gold (Aswath Damodaran's data set at NYU Stern, with Shiller and NBER figures for the 1920s), and the FTSE Nareit All Equity REITs index from 1972. There was no REIT index before 1972, so earlier years use US home prices, which leave out rent and understate what property returned. Returns are in US dollars and are applied to whatever amount you type, whatever the currency symbol: the figures are illustrative, not a currency conversion.

History is a sample of one. That a portfolio survived every crisis in this dataset is evidence about its character, not a guarantee about the next crisis, which will be shaped differently.

A worked example

Put 100% in stocks and start in 1929. By the end of 1932 the balance has lost almost two thirds of its value. It is back above where it started at the end of 1936, falls below it again in the 1937 crash, and only stays above it from the end of 1944 — sixteen years after the start. Put the same money in the Permanent Portfolio — a quarter each in stocks, Treasury bonds, gold and cash — and its worst point in those years is about 15% down.

Then run both to the end of 2025. The all-stock portfolio ends far ahead. Both facts are true at once, and the useful question is not which number is bigger but whether you would have held the first one through 1932 — because an allocation you abandon at the bottom returns nothing at all. That is what the tool is really testing.

Common questions

Are the returns adjusted for inflation?

Only if you ask. The default view shows nominal total returns; the Real view takes out each year's US consumer price inflation, so every value is in money of constant purchasing power. Over long windows the gap is large: read any nominal result that spans the 1940s or the 1970s as much bigger than it was in what it could buy.

Does it rebalance?

Yes, once a year, back to the exact weights you set. If you want to know what buy-and-never-touch would have done instead, set a single asset class to 100% — that is the only allocation rebalancing cannot change.

Why does the start year change the answer so much?

Because the order of returns matters as much as their average. Starting a year before a crash and a year after it produces very different balances from identical average returns — which is the single strongest argument for not putting a lifetime of savings in on one day.