Methodology & Assumptions

How this forecast is built, updated, and constrained.

Purpose & Scope

This project estimates the probability that Vladimir Putin will not be in power by Dec 31, 2026. It is not a forecast of Russian state collapse. We track leadership exit via health, coup/elite bargain, forced resignation, or other removal.

Model Structure

1. Base Rate (P₀): Historical chance of an autocrat exiting within the horizon. Converted to odds O₀ = P₀/(1−P₀).
2. Bayes Factors (BF): Multiplicative evidence from six buckets:
   • Military — operational effectiveness, force generation, technological adaptation
   • Economy/Energy — production & exports, infrastructure integrity, sanctions/finance, domestic fuel prices, industrial/logistics, FX/reserves
   • Elite/Institutions — cohesion vs fracture, purges, succession signals
   • Domestic/Public — morale, shortages, protests, repression backfire
   • External/Diplomatic — allied cohesion, sanctions enforcement, arms flows
   • Symbolic/Context — prestige shocks, narrative fractures
3. Posterior: O = O₀ × Π(BFᵢ);   P = O/(1+O).

Evidence Standards

• BF > 1.15 requires ≥2 independent reputable sources or direct data (imagery, filings, official notices).
• Single-source/provisional items enter smaller (≈1.05–1.10) and are upgraded/downgraded on confirmation.
• We distinguish signals (operational/economic) from noise (rhetoric without action).

Controls Against Double Counting

• One BF per bucket per update cycle (weekly cadence).
• Correlation penalty 0.90–0.95 when multiple items share a cause (e.g., refinery damage & export loss).
• Caps on Economy/Energy (typically 2.0–2.5 per cycle) to prevent runaway compounding.

Update Cadence

Manual weekly updates. Emergency updates for major inflections (e.g., elite defection, nationwide rationing, large territorial shifts, strategic weapons use).

What Moves the Needle

• Up: sustained refinery/grid outages inside Russia; elite fracture/purge; long-range strikes reducing production capacity; Asian oil demand reductions; widespread domestic shortages.
• Down: Russian multi-front gains >30 days; effective export workarounds; easing domestic shortages; alliance fatigue.

Interpretation

Numbers are probabilities, not certainties. A 70–90% probability still allows meaningful downside scenarios. The model is disciplined: transparent inputs, explicit penalties, and dated entries.

Limitations

• Open-source bias and reporting lags.
• Unobservable elite dynamics may move faster than public data.
• Exogenous shocks (e.g., sudden health events) modeled as residual risk, not forecastable specifics.

Change Log & Versioning

Each weekly update appends a dated entry to the Dashboard “Recent Triggers.” Major methodology changes are noted here and in the repo commit history.

Contact

Questions, corrections, or data contributions: open an issue in the repo or email the maintainer.