Ridings ranked by forecast gap between the favourite and the runner-up. Selecting a row or a map cell updates the riding panel.
Official Elections BC results by electoral district (2017, 2020 and 2024, DataBC open data), provincial polls published since 2017 (Wikipedia), regional breakdowns of 48 polls taken from the pollsters' reports, 34 of them since the 2024 election, and the official 2026 candidate list. The 2020 results are transposed onto the 2023 map: the election-day votes of each voting area are split between the 2023 districts according to the 2024 registered voters in the overlap of voting areas, and advance, mail-in and special votes, known only by district, follow for each party the distribution of its election-day votes. Boundaries of the 93 districts of the 2023 redistribution (Elections BC), simplified for display.
The poll average is a Gaussian process on the isometric log-ratios of the shares, fitted to the polls published since the 2024 election (each firm's last poll per week), with changepoints at leadership changes and party-system realignments, and predicted at election day. On the eve of the 2020 and 2024 elections, the mean absolute error of this average is 1.2 and 1.0 points. A party measured province-wide but running in only some districts has its share scaled to p·c, c being the share of the 2024 votes cast in the districts where it has a candidate.
The 2024 results are carried to the provincial composition by two swing forms, each in half of the draws. By transfer, a declining party loses the same proportion of its votes everywhere and the freed votes go to the growing parties, in proportion to their gains; by additive swing, each party gains or loses the same number of points everywhere. In both, a party with no candidate in a district receives no votes there. Evaluated at known provincial shares, the transfer gives a root mean square error of district shares of 6.2 points (2017→2020) and 6.5 points (2020→2024), the additive swing 5.7 and 7.4, a proportional swing 6.1 and 17.3; the mixture of the first two forms has, on average, the best log-likelihood of the winners on the eve of both elections. A regional shift taken from the breakdowns (Metro Vancouver, Vancouver Island, rest of the province) and an incumbency effect (open seat, incumbent running under another banner), evaluated the same way, do not improve the forecast: their weight is zero.
Each draw combines a draw of the poll average, a provincial shock in centred log-ratios with variance τ²/s for a party of share s, τ being estimated on the eve-of-election poll errors of 2020 and 2024, and, for each district, a local deviation drawn from the errors of the 2017→2020 and 2020→2024 transitions under the same swing form, centred by region. On the eve of the 2020 and 2024 elections, each evaluated with parameters estimated without it, all 6 of the 90% intervals of seats by party contain the result; the absolute seat error of the median scenario (medoid) is 0 and 6, and the winner is correct in 73 of 87 and 79 of 93 districts.
Each draw is assigned to a single government scenario: majority when the first party reaches 47 seats, minority otherwise. For a minority government, the parties whose seats, added to those of the first party, reach a majority are listed, excluding the Official Opposition (second party in seats): these parties hold the balance of power. In the certainty chart, each riding is assigned to its favourite party and classed by that party's probability of winning: safe (95% or more), likely (75 to 95%), lean (60 to 75%), toss-up (under 60%).