Methodology

We publish ordinary news with an extra layer: the historical set-ups it most resembles, and what those set-ups produced. This page explains every number you see, so you can judge it yourself.

1. The story

Each article starts with the day's reporting. We draw on wire services, primary documents (central bank statements, court filings, official results) and established outlets, and we link every source at the end of the piece. Nothing in the story section is speculation; the analysis lives in the sections that follow.

2. The pattern library

A pattern is a cluster of past events that shared a recognisable set-up: the conditions, the actors' incentives and the constraints they faced. Each pattern page lists its cases, what happened in each, and the signals that mark the pattern. The library is curated by the desk and grows as stories demand it. We prefer fewer, well-documented cases to long lists.

3. The rhyme score

The rhyme score (0–100) says how closely today's set-up matches a pattern. It is a structured judgement, not a model output. The desk scores five dimensions, each 0–20: the trigger (is the precipitating event of the same kind?), the conditions (economic, political or technical backdrop), the actors' incentives, the constraints they face, and the information environment. We publish the "where it breaks" note alongside every match because the divergences usually matter more than the similarities.

Score	Reading
80–100	Close rhyme. Same trigger, conditions and incentives; divergences are second-order.
60–79	Clear rhyme with a material divergence noted.
40–59	Partial rhyme. Useful as context, weak as a base rate.
<40	Not published as a match.

4. The outlook indicator

The outlook lists the plausible resolutions of the story over a stated horizon, each with a probability. The starting point is the base rate: across the matched patterns' cases, how often did each kind of outcome occur? Cases from higher-scoring patterns carry more weight. The desk then adjusts for the divergences, and states the adjustment in the "basis" line under each scenario. Probabilities always sum to 100.

Confidence reflects the evidence, not the probability. High means many cases, clean matches and a short horizon. Moderate means a reasonable sample with meaningful divergences. Low means few cases or a long horizon; treat the number as a structured guess.

5. What the numbers are not

They are not forecasts, trading signals or advice. History gives base rates, and base rates are a good place to start thinking, but every story also has features with no precedent. Our job is to make the precedent visible and the uncertainty honest. When a story resolves, we record the outcome on its page and in the pattern library, so our track record is public.

6. Corrections