Polls & Odds

Methodology. The polling average is a weighted average of the most recent national polls of the 46th federal election: each poll's weight halves every 14 days and grows with sample size, though it stops growing past 5,000 respondents. Because pollsters publish at very different rates, a firm releasing several polls within 14 days has each of those polls weighted down, and no single pollster contributes more than 35% of the total weight — so a weekly tracker cannot dominate the average. The seat projection applies each party's regional swing — our current regional standing minus its 2025 result — to all 343 ridings' official 45th general election results (Elections Canada), then runs 10,000 simulations with a ±3-point correlated regional error and ±4.5-point riding-level noise. It is deliberately simpler than dedicated seat models (no incumbency, candidate or riding effects): read the ranges, not the point estimates. The trend line applies the same average at weekly steps; each dot is one published poll. Poll figures are compiled into Open Data Canada's own poll registry directly from each polling firm's published releases — Léger, Abacus Data, Nanos Research, Liaison Strategies, EKOS, Mainstreet Research and others; where the release is public, the pollster's name in the table links to it. Provincial shading uses a dedicated provincial poll where we have a recent one; where we don't, we fall back to the provincial crosstabs printed inside national releases — pooled across at least two releases, or taken from a single release when the pollster states a regional sample of at least 400 respondents. Crosstab subsamples carry larger error than full provincial polls, and each province's source is named on the map. Prediction-market probabilities are live prices from Polymarket — the share price of a contract paying $1 if the outcome happens, read as the market's implied probability; markets under US$25,000 traded are flagged as thin, since a handful of bettors can move them. A poll measures support today; a market prices the final outcome — they answer different questions, which is exactly why comparing them is informative. Leader favourability is a separate weighted average of net positive-minus-negative ratings for each federal leader, built the same way as the vote-intention average, from releases by Abacus Data, Angus Reid Institute and Liaison Strategies; Nanos Research's "preferred prime minister" horse-race question measures something different and is excluded. This page is not affiliated with any pollster, party, or Polymarket.