Twenty-five predictions you can check

10 October 2026 · The Law of the Board

Can game theory predict the future? Not the way a prophet claims to. It cannot tell you who wins an election. It can read the incentives already in place and say where they push. That is only worth something if the reading can be checked afterwards. The Rational Prophet, Book Four of the series, was published on 10 October 2026 to put that to the test.

Why most forecasts cannot be scored

Ask any forecast three questions: what would count as wrong, by when, and who decides? Most have no answer. That is how the game pays. A vague forecast can be reinterpreted to fit whatever happens, so it never loses; a precise one can, and nobody keeps the record anyway.

What makes a prediction checkable

Four parts, fixed before the event: a public source anyone can open, a threshold that decides the outcome, an expiry date, and a losing condition stated in advance. Every bet in the book has all four. There are twenty-five, from GP-01 to GP-26; GP-20 was withdrawn, and the withdrawal is printed rather than hidden. Each carries a baseline read and frozen on publication day, so the starting point cannot be moved later.

Three of the bets

GP-04 — Power becomes the binding constraint. Datacentre demand outruns the power available to carry it, read in the capacity auctions of the grid operator PJM. It wins if every auction in the window ends below the operator's own reliability requirement.

GP-18 — The dollar loses exclusivity, not first place. Its share of official reserves falls below half while staying at least fifteen points ahead of the runner-up, in the IMF's quarterly COFER data.

GP-21 — Restraint holds where the taboo is rejected. A state whose own doctrine rejects the nuclear taboo goes through an armed conflict without using the weapon. One detonation in conflict, by anyone, inside the window loses it.

None of them is a prediction about a person. Each is about what an arrangement of incentives produces, read in a number someone else publishes.

How the author is held to it

The author's probability for each bet is registered on the Ledger, not printed in the book, and will be judged with a Brier score as each bet closes.

The bets, their definitions, the baselines and the probabilities are pre-registered on the Open Science Framework, an independent archive the author cannot edit afterwards: osf.io/7ev8k, DOI 10.17605/OSF.IO/7EV8K.

The book also scores the predictions of the earlier volumes, failures first. The method comes from Professor Jiang Xueqin's lectures; the twenty-five predictions, and their errors, are the author's own.

Some bets will fail. That is the point of writing them down. A test for any forecast, these included: find the observation that would prove it wrong. If there is none, it was never a prediction. Why long horizons can be easier than short ones is set out in why thirty years is easier to predict than three, the Law of the Board applied to time.

Common questions

Can game theory predict the future?

Not specific events or personal decisions. It can read the incentives each player faces and say where an arrangement is heading. The honest test is whether such forecasts can be scored; The Rational Prophet makes twenty-five that can.

Why pre-register predictions on the Open Science Framework?

Because a forecaster who keeps his own record can edit it. Once the bets, baselines and probabilities are registered on the Open Science Framework, the author cannot change them, and anyone can compare the outcomes with what was written on publication day.