Location

server/worldmonitor/intelligence/v1/get-risk-scores.ts

const BASELINE_RISK: Record<string, number> = {
  US: 5, RU: 35, CN: 25, UA: 50, IR: 40, IL: 45, TW: 30, KP: 45,
  // ...
};

const EVENT_MULTIPLIER: Record<string, number> = {
  US: 0.3, RU: 2.0, CN: 2.5, UA: 0.8, IR: 2.0, IL: 0.7, TW: 1.5, KP: 3.0,
  // ...
};

These tables feed directly into the Combined Intelligence Index formula:

const baseline = BASELINE_RISK[code] || 20;
const multiplier = EVENT_MULTIPLIER[code] || 1.0;
// ...
const acledScore = Math.min(50, Math.round((d.battles * 3 + d.explosions * 4 + d.civilianViolence * 5) * multiplier));

Impact

The scores are not reproducible from the displayed data. A user who sees that North Korea scores 72/100 cannot independently verify it. The BASELINE_RISK: 45 and EVENT_MULTIPLIER: 3.0 for KP are invisible in the UI — they are editorial judgments hardcoded into the server.

Specific problems:

1. KP scores high with zero ACLED events. BASELINE_RISK: 45 × 0.4 baseline weight = 18 points before any live data is ingested. At full formula weight, KP can score above 50 with no current events.
2. IL scores are suppressed (0.7×). Events in Israel score 30% lower than identical events in Russia (2.0×). This is an editorial choice with no published rationale.
3. The multipliers are applied to ACLED fatality counts, which are already estimates from media reports. Multiplying contested fatality data by a hidden country-specific factor compounds the uncertainty without disclosing it.
4. The methodology note in the UI does not disclose these baselines or multipliers. Users comparing countries are not aware that a baseline political risk judgment is baked in.

Fix

• Document the baseline and multiplier methodology publicly (methodology page, API response metadata, or tooltip).
• Or move these values to the database / config layer so they are auditable and not embedded in server code.
• At minimum, include staticBaseline and the effective multiplier in the API response so third-party callers can reproduce the computation.