The method
This methodology explains how CassandrIA measures country pressure and forecasts conflict escalation. It also explains how each published forecast is checked against a dated record when its window closes. Six layers per country: one descriptive identity layer and five pressure readings. A country score from 0 to 100. Four bands. A signed reading every eleven days, and a monthly outlook scored against the record. Nothing here is a recommended action.
How we work
Four steps in a fixed order, each with its own rules.
Six layers per country: one descriptive identity layer and five pressure readings.
Stress followed from each country to routes and markets, with forty five questions tracked daily.
A monthly probability for each country and event class, with a fixed deadline.
Each estimate checked against its named source when the window closes.
In detail
Each step has one job. Its limits are part of the method.
Two hundred and six countries are read on six layers: identity, economy, politics, events, information and society. Every eleven days the reading is reviewed, signed and frozen. The number on a country page is this measurement.
A reading describes the present. It does not say what will happen.
Open data on trade routes and commodity markets show where stress in one country travels to others. Forty five standing questions are tracked every day, and each one lights up when enough independent signals agree.
Disagreeing signals stay visible. They are not averaged away.
Read the forty five questions.
From 25 Oct 2026, estimates combine each country’s historical escalation rate at comparable recent death levels with rates pooled across countries. The pooled rates account for the average monthly deaths over the latest three months and how many of those months had deaths. They also account for the change from the preceding three months. A fixed smoothing rule gives pooled rates more influence when country history is limited. The method was selected using previously examined 2018–2025 data, so this retrospective comparison does not establish prospective accuracy or calibration.
When a window closes, each estimate is checked against the sources named in its rule and marked as happened or not happened. When the sources cannot settle a case, that is recorded too. The record only grows. Nothing is rewritten and nothing is removed.
What is measured
L1 describes identity. L2 to L6 are five pressure readings on a 0 to 100 scale where data are available. The figure under each layer counts what moved since the first signed reading.
Change over time
Fields changed between consecutive signed readings. Layer totals are shown separately above.
Loading cycle history
Readings without a published series are marked on the axis.
Reading the colour
The band follows the published score, with thresholds at 55, 65 and 80. Phase 1 to 4 moves with the band. Historical charts run to 2025.
Source: CassandrIA country overview.
The rhythm
A cycle is reviewed, signed and frozen. Each dot is a signed cycle with the fields it changed.
A recent page load does not make an old observation current. Observation dates differ between sources.
Standards
Scope
The overview and the detailed readings have different scope. Check the country page and its source notice before citing a figure.
Band, score and phase for every entry on the map and in the atlas.
Layer by layer, with the film of events and the pressure points it touches.
Readers
Country readings and a monthly outlook, free and the same for every reader.
Words
The terms used in these readings, in one line each.