Methodology

Overview

This site is a geopolitical analysis dashboard that summarizes public reporting into a single temperature score per tracked conflict. The score represents escalation pressure, not certainty, and is designed to be explainable through factor-level evidence.

Data Sources

Inputs come from curated public RSS feeds and query-based public news syndication endpoints. Sources are normalized, deduplicated, and link-checked so removed or inaccessible items can be marked without silently dropping historical traceability.

Source quality is weighted using domain/source profiles so a single low-credibility outlet cannot dominate model output. The pipeline favors corroboration across distinct publishers before allowing large score movement.

Signal Categories

Each conflict has a fixed set of weighted signal categories, such as force posture, strike tempo, official rhetoric, and diplomatic offset indicators. Category definitions live in configuration and are evaluated on each refresh run using keyword markers and recency weighting.

Temperature Calculation

Factor signals are scored on a 0-100 scale, multiplied by configured weights, and combined into a single weighted temperature percentage. Smoothing, corroboration damping, and low-evidence controls are applied to reduce false spikes caused by sparse or repetitive coverage.

Recent evidence can raise temperature quickly when multiple trusted sources align, while cooling is generally constrained to avoid unrealistic whiplash between refreshes. This keeps the dashboard responsive without behaving like a headline counter.

Conflict Status Categories

Status labels (`NO`, `KINDA`, `YES`) are separate from temperature. Status reflects editorial classification from public evidence thresholds, while temperature reflects model-estimated pressure for further escalation.

YES is used when reporting supports sustained foreign ground-force presence inside the target country. In plain terms, that means troops are physically operating on the ground and remain there beyond a brief raid or isolated incursion (often summarized as "boots on the ground").KINDA is used when meaningful hostile action occurs without that sustained ground presence, such as active strike exchanges or very brief ground incidents (for example, a short capture operation like the Venezuela president scenario).

That means a conflict can be KINDA and still show a cool temperature if follow-on escalation signals weaken.

Limitations

The model is constrained by public-source latency, reporting bias, and uneven source coverage across regions. It does not have classified inputs and should be treated as a monitoring aid, not as operational or personal safety guidance.