Open methodology for Indian elections.
Per-constituency forecasts. Multi-cycle structural data. Bayesian and ensemble models. Methodology published before counting day, predictions unsealed during it. Researchers welcome.
Every forecast on this site is committed before the result is known. Methodology is the contract; accuracy is the test.
Choose a state
Tamil Nadu
Kerala
West Bengal
Puducherry
Assam
More states will be added as data is sourced and methodology validated against historical cycles. Contributions and tips welcome.
Every assumption documented.
Multi-cycle AC coefficients, Bayesian hierarchical models, Monte Carlo with calibrated mean reversion. Published in plain language before any prediction goes live.
Twenty-five years of voting.
Historical Assembly and Lok Sabha results per constituency, demographic features, candidate rosters. All available via REST API for researchers.
Predicted vs actual, transparently.
When the count finishes, every forecast is compared seat-by-seat against the result. Hits and misses are public. The model gets graded.
REST API. Free to query.
All public data — constituencies, historical voting, demographics, candidate rosters, methodology metadata — is queryable via a public REST API. Predictions and Monte Carlo distributions become queryable after the unlock date for each cycle.
API documentation · Read the methodology · About this project