The Alternate Historian
Decades that never happened, reported as if they had
What this is
A simulator of economic decades that did not happen. You describe a world in a sentence -- a stagflation, a deflationary bust -- and it builds ten years of monthly returns for equities, bonds, credit, commodities, property and private funds inside that world.
Then you live through the decade as the investment office of an institution. The news arrives month by month. At set windows you make a handful of real decisions -- what to hold, what to commit to private funds, when to rebalance. At the end you find out whether those decisions beat leaving the portfolio alone.
Two things make it more than a quiz. Private holdings keep two sets of books, as they do in life: the reported marks are appraised and smoothed, the true ones are not, and the gap between them is widest exactly when markets fall. And every decade is reproducible to the last bit -- the same seed replays the same ten years, so any number on any screen can be traced back and recomputed.
What it is not
It is not a forecast and it is not advice.
There are two kinds of world, and they are made differently. The hand-built ones come from a small parametric engine: you state the shape of the decade and it draws returns to fit. The generated ones are cut out of real market history -- the months between 1953 and 2020, sliced into strips and spliced into new sequences -- so what they contain is rearranged truth.
A third method was tried and rejected. A generator that paints decades from scratch, rather than rearranging them, was put against the splice three times under rules sealed before each contest. It won the first two. On the third, run over the full 1953 to 2020 span, its edge came out smaller than its own run-to-run noise, which under the sealed rule is not a win. It was shelved, and the project stayed with rearranged truth.
What that leaves is stated plainly in the project's own record: this is not a convincing model of history. Both world-makers undercall how long a regime lasts, and both understate crashes by roughly half. Nothing produced here supports a real investment decision. It is a rehearsal room, not a crystal ball.
The simulator that runs the decade is deliberately simple. That is a choice, not a limitation being hidden: a small engine is one whose every number can be checked by hand, and the project keeps a standing register of the places where it is faithful to its own design but not to an allocator's expectations. It touches no real money, holds no client data and places no trades.
Read before you ask
- The plain-English manual What the platform is, what it honestly is not, and every task walked through without assuming you know a terminal.
- Build summary -- what exists The system map: every capability with its status and its evidence, and the register of known gaps, which is the point of it.
Both are also in the library.
Asking for a seat
Playing a decade is by invitation, because each reader's session is scored and kept. If you would like one, say who you are and the editor will read it by hand.
Already invited? Open the invitation link the editor sent you. It works once, and signs your browser in for the three months after that. If yours has been used already, ask for another.
The editor signs in here.