ProjectMoosed Research Lab · World Forecasts
Every forecast here is written in plain language, given a probability, a deadline, and a named source that settles it — then locked. No vague verses, no retrofitting after the fact, no silent edits. When the deadline passes, we grade ourselves in public, right or wrong.
Eighteen world predictions for 2027, across politics, geopolitics, the economy, climate, space, technology, health, and culture. The scope of this lab is unlimited — we predict anything our prediction radar picks up, even war — and the categories are labels, not gates. Each prediction states what would make it come true, how likely we judge it, when it settles, and who settles it. The probabilities are estimates, not guarantees — and this page is where you can hold us to them.
The founding register settles through 2027 — too long to wait for the first grade. Batch 002 is the radar's near-term sweep: 22 predictions that resolve between December 2026 and April 2027, locked the same day the site launched, before any of these outcomes was known — including the House call below, locked on Election Day itself, before polls closed. The founding 18 above are unchanged; this batch stands next to them under the same rules: plain language, one probability, a deadline, a named settling source, no edits after lock.
Nothing to grade yet — honestly stated, the way it should be. The first prediction settles on December 9, 2026 (the Federal Reserve's December decision), Batch 002 resolves through April 30, 2027, and the founding register settles through 2027, most of it on December 31, 2027. As each one resolves, it moves here with its outcome, its source link, and its Brier contribution. Our calibration line will read: "When we say 70%, it has happened X% of the time" — computed only from settled predictions, with the sample size printed next to it.
dfcb0b4aa35fcd6e48e4e52cc19a3bceca2c4a671244664468767291c11ee11e
ec499ca76903f4be784ab368531542a91a9c2db75df2b3871990595d0e5ab502
Good forecasting is a discipline, not a gift. Ours follows the rules that separate a forecast from a horoscope:
Start with how often this kind of thing usually happens — how often a hurricane of that strength lands, how often a government shuts down — then adjust for the specific evidence. Never start from a feeling.
If a prediction can't be settled yes-or-no by a named source on a named date, it isn't a prediction. It doesn't go in the register.
"Might happen" is unfalsifiable. A number can be scored. Our probabilities are graded with the Brier score, which punishes confident wrongness hardest.
Publication is the lock. If events change our mind, we publish a new dated prediction — the old one stays and gets graded as written.
The scoreboard shows everything, forever. A forecasting record you can cherry-pick is marketing, not research.
These are research estimates about the world, published to test our own accuracy. They are not betting, investment, or safety advice.
We build on the foundations of Nostradamus, rebuilt as a real and effective technique. What we keep from him: long-horizon attention to recurring human patterns — wars, plagues, succession crises, fires, floods; the study of cycles, comparing today's configurations with documented past ones (his "comparative horoscopy," rebuilt as reference-class and historical-cycle analysis); and cataloguing many calls instead of a few safe ones. What we discard is everything that made his calls ungradeable: the deliberate vagueness, the anagrams, the missing dates. Learn from him whatever is useful — then lock it, price it, and grade it.
Scope is unlimited. A prediction enters the register when the radar — the signals, trends, and scheduled events of the world — produces something worth a falsifiable call: an election, a launch window, a treaty deadline, a storm season, a war. If it can be settled, it can be predicted.
This lab exists because of a 16th-century lesson. Michel de Nostredame published 942 deliberately obscure verses in 1555, and for 470 years readers have matched them to events after the events happened — a process called postdiction, which can "confirm" almost anything. His one famous quatrain with a real date in it, 1999, failed on schedule. Our companion site, nostradamus.projectmoosed.com, grades his famous claims one by one against the original text: what is proven, what is documented, and what is hype. This register is built to be everything his wasn't: clear before the fact, and checkable after it.