About Predictive Politics
Predictive Politics is a Canadian federal election forecast model built to provide transparent, riding-level projections of vote shares and win probabilities across all 343 electoral districts.
What This Site Does
The site publishes interactive forecasts showing — for each of Canada’s 343 ridings — which party is projected to win, the estimated vote share for each party, and the uncertainty around those estimates. Forecasts are derived from a poll aggregation and uniform swing model described in detail on the Methodology page.
Results are presented in two formats:
- An interactive choropleth map on the home page, where each riding is coloured by the projected winning party and the shade reflects win probability.
- Riding-by-riding tables on the regional pages, showing the forecast bar chart for every district in each province and territory.
Why Build This?
Canada has a number of excellent election trackers, but many are opaque about exactly how they work or present only national or provincial seat totals. Predictive Politics aims to be explicit about its assumptions and limitations, and to visualize uncertainty at the riding level — so readers can see not just who is projected to win a seat, but how confident the model actually is.
The 2025 Federal Election
The current forecasts cover the 2025 Canadian federal election, held on April 28, 2025. The model was updated regularly in the weeks leading up to election day as new polling data became available.
Limitations
No forecast model is a crystal ball. Predictive Politics uses a national uniform swing approach, which does not account for local candidate effects, incumbency, or regional trends that deviate from the national picture. Uncertainty estimates are based on polling margin of error and do not capture the possibility of correlated, systematic polling error. Treat all projections — especially in competitive ridings — as probabilistic estimates, not predictions.
For the full technical description of how the model works, see the Methodology page.