The answer
What to buy
Six things. Every one is a fund, bought once, held for years.
What US$4.5 million would have done
What the worst stretch looked like, in dollars
The number that matters more than the allocation
Or fund the company for a few years, then stop
The honest part
The portfolio, in detail
Four versions: the answer, the same thing without the trend fund in case you cannot buy it, one steadier and one bolder. Pick one to see its numbers.
Growth of US$4.5 million
A ratio scale makes an equal percentage move the same height everywhere on the chart, so the 1987 crash and the 2008 crash look as big as they felt. On a straight scale early history looks flat.
How far below its own high it ever fell
Your rule is the red line: never more than 30% below the highest value the account has ever reached.
Year by year
Every crisis since 1987
What this mix did, next to what 100% world shares did.
If you had started at the worst possible moment
The annual return over every five-year stretch. The lowest point is what matters: it is the worst luck the mix has ever handed anyone.
Decade by decade — what you are actually buying
This is the most honest way to look at it. In most decades, simply holding all the world’s shares beat this mix. The reason to own it is what happens in the decade that does not go well.
How much rests on the gold
Twenty per cent in a metal that pays no interest deserves a challenge. Here is the same portfolio with the gold weight moved into world shares.
If you would rather hold fewer funds
Simplicity is nearly free here. A portfolio you actually hold beats a better one you fiddle with, so this is a real option, not a consolation prize.
Next to the famous recipes
The well-known fixed portfolios, each measured over the longest history it can reach, against the same two rules. Green passes, red breaks.
Twenty thousand alternative histories
Each holding
What each fund is, what it costs, and its own history — as far back as honest data goes.
Why not 10%
The trade-off, measured
For each loss limit you might set, the best annual return that has actually been achievable under it over 1980–2026. Every point is a real portfolio of the funds on this site, searched and refined separately — not a theoretical curve. The fourth column converts the month-end figure to what you would actually have seen on a daily screen, which is about 3 points deeper.
The three ways to reach 10%, and what each costs
The same search, on four different slices of history
The most revealing table on this site. Identical funds, identical rule, four different starting years.
Looking forward, not backward
What if I am wrong about the trend fund?
The managed-futures sleeve is the piece with the least real history: the fund is eighteen months old, and everything before 2007 is a model. So here is the same portfolio holding 0%, 10%, 12%, 15% and 20% of it.
How to buy it
Which account, which order, in which order.
The shopping list
The order to do it in
If you want to pick your own shares
You said you do. Here is how to do it without handing your family a tax bill they have to sell assets to pay.
All at once, or spread out?
What it costs to run
Rules that do not change
Can you trust this
Where the numbers come from, what was checked, and what is still a guess.
The blind test
Choose the weights using only data up to a date, then score them on the years that came after. It is the one test an optimiser cannot game, and it is brutal.
And the fair comparison — what the same years paid to someone who never optimised anything:
If you had started at a different moment
You are investing one lump sum on one day. The annual return over every ten-year stretch this mix has ever seen, by starting month.
How often to rebalance
Every figure on this site assumes rebalancing once a month, which nobody actually does. Here is what really happens at each frequency, after trading costs.