1
Collect official evidence
We prioritise structured data and first-party institutional sources, then record the evidence used by the forecasting system.
2
Convert evidence into probability
Signals are combined into a probability rather than a vague prediction. A neutral evidence set can legitimately remain near 50%.
3
Define the resolution rule
Every public forecast needs a clear outcome rule and a deadline so the question can later be judged YES or NO.
4
Resolve against official outcomes
Machine-verifiable outcomes can resolve automatically. Ambiguous outcomes are not guessed; they are held for explicit review.
5
Lock the historical forecast
The probability at resolution, source evidence and outcome are preserved in an immutable resolution ledger.
6
Score the forecast
We use Brier scoring so calibrated confidence matters. A confident wrong forecast is penalised more than a cautious one.
Why this matters: the goal is not to publish dramatic predictions. It is to build a measurable forecasting record that decision-makers can inspect over time.