Track Record · Quebec election, October 5, 2026
Quebec 2026: how we did
A PQ minority, the Liberals as official opposition, the Conservatives third: that was our most likely outcome. Here is, number by number, what we got right and where we went wrong.
Seats
Our two models read the same polls. Harfang des neiges is our reference projection; Orbit, our experimental model, runs alongside it in our Labs. Each number is the median of thousands of simulated elections, with the range where 8 simulations in 10 fall.
| Party | Elected | Harfang, median | Harfang, leading | Orbit, median | Orbit, leading |
|---|---|---|---|---|---|
| PQ | 59 | 61 52–67 | 62 | 60 45–73 | 68 |
| QLP | 40 | 37 33–44 | 38 | 36 29–46 | 36 |
| PCQ | 19 | 19 15–23 | 18 | 17 11–25 | 15 |
| QS | 9 | 6 5–8 | 6 | 8 5–11 | 8 |
| CAQ | 0 | 3 1–7 | 3 | 4 0–12 | 0 |
Harfang: 2.2 seats off on average, Conservatives exact. Orbit: 2.4. Our ranges held the result for 3 of 5 parties (Harfang) and 5 of 5 (Orbit). Harfang's two misses are QS (9 elected, 6 projected) and the CAQ, wiped out: we gave it a 3% chance of ending at zero.
The median or the map?
On our map, each riding is coloured by its favourite. Adding them up gives PQ 62, QLP 38, PCQ 18, QS 6, CAQ 3 (2.4 seats) for Harfang, slightly less accurate than the median: local misses partly cancel out in the total. Our probabilities announced in advance about 18 ridings where the favourite would lose; there were 17. The blend of both engines gives PQ 65, QLP 38, PCQ 16, QS 8, CAQ 0 (2.4).
The vote
| Party | Result | Harfang | Orbit | Poll average |
|---|---|---|---|---|
| PQ | 28.0 | 28.2 27.2–29.2 | 29.5 25.7–33.3 | 28.6 |
| QLP | 24.2 | 24.1 22.7–25.5 | 24.1 20.5–27.6 | 23.9 |
| PCQ | 20.7 | 20.1 18.8–21.5 | 19.8 16.6–23.1 | 19.7 |
| QS | 12.4 | 10.8 9.8–11.9 | 10.2 7.8–12.8 | 10.9 |
| CAQ | 13.3 | 15.4 13.9–16.8 | 15.5 12.5–18.6 | 15.7 |
Harfang was within 0.9 point on average, better than the average of the final polls (1.2). PQ, Liberals and Conservatives are within a point. Our two misses are the polls' misses: the CAQ 2 points too high, QS 1.6 too low. Orbit misses a bit more (1.4), but its wider ranges held all five results; Harfang's held three of five.
The 127 ridings
Number of ridings where each engine's favourite won.
Harfang got the number of right winners it announced: its probabilities were sound. Orbit got more right than expected: it was too cautious. The 50/50 blend of the two, which we publish in the Labs, beat both.
When we said…, we were right…
| Stated probability | Harfang | Orbit | Blend |
|---|---|---|---|
| 0–60% | 41% | 57% | 65% |
| 60–80% | 79% | 86% | 85% |
| 80–90% | 90% | 96% | 92% |
| 90–99% | 96% | 100% | 97% |
| ≥ 99% | 100% | 100% | 100% |
None of the 45 ridings Harfang gave at 99% or more flipped.
Our 17 misses
Ridings where Harfang's favourite lost, from the closest miss to the clearest: at the top, ridings where we already gave the winner a good chance.
| Riding | Winner | Its chance per Harfang | Harfang's favourite | Orbit | Blend | Final margin |
|---|---|---|---|---|---|---|
| Sherbrooke | QS | 44% | PQ 56% | QS 75% | QS 60% | 1.3 |
| Nicolet-Bécancour | PCQ | 41% | PQ 58% | PQ 59% | PQ 58% | 12.4 |
| Ungava | PQ | 39% | QLP 57% | QLP 56% | QLP 57% | 12.0 |
| Granby | PQ | 33% | CAQ 54% | PQ 60% | PQ 47% | 12.4 |
| Maurice-Richard | QLP | 33% | PQ 38% | QS 60% | QS 45% | 1.0 |
| Brome-Missisquoi | QLP | 32% | PQ 55% | PQ 57% | PQ 56% | 5.3 |
| Charlevoix–Côte-de-Beaupré | PCQ | 31% | PQ 68% | PQ 67% | PQ 68% | 13.4 |
| Bellefeuille | PQ | 31% | CAQ 33% | PQ 85% | PQ 58% | 18.1 |
| Huntingdon | PQ | 31% | QLP 45% | QLP 53% | QLP 49% | 5.0 |
| Rivière-du-Loup–Témiscouata–Les Basques | PQ | 28% | CAQ 41% | PQ 89% | PQ 58% | 12.3 |
| Rosemont | QS | 23% | PQ 77% | PQ 53% | PQ 65% | 11.3 |
| Louis-Hébert | PCQ | 20% | PQ 77% | PQ 66% | PQ 71% | 2.4 |
| Marie-Lacoste-Gérin-Lajoie | PQ | 15% | PCQ 66% | PQ 69% | PCQ 45% | 8.8 |
| Montarville | QLP | 12% | PQ 86% | PQ 76% | PQ 81% | 0.8 |
| Daniel-Johnson | PQ | 11% | PCQ 51% | PQ 74% | PQ 42% | 9.8 |
| Jean-Talon | QLP | 10% | PQ 89% | PQ 72% | PQ 81% | 2.5 |
| Taschereau | QS | 7% | PQ 94% | PQ 87% | PQ 90% | 16.5 |
Where we went wrong
Average gap between our projection and the result, per riding, in points. Positive: we had the party too high.
| Region | Ridings | Right | PQ | QLP | PCQ | QS | CAQ |
|---|---|---|---|---|---|---|---|
| Montreal and Laval | 33 | 31 | +2.4 | +1.0 | +3.9 | −7.2 | −1.1 |
| Montreal suburbs | 38 | 35 | −1.9 | −0.6 | +1.5 | −0.1 | +1.3 |
| Quebec City region | 9 | 6 | +3.6 | −1.2 | −3.9 | −2.4 | +4.0 |
| Chaudière-Appalaches | 7 | 7 | +0.9 | +0.4 | −8.7 | +0.9 | +6.5 |
| Eastern Townships | 7 | 4 | +1.6 | −5.0 | +2.0 | +0.6 | +1.2 |
| Rest of Quebec | 31 | 26 | −0.6 | −0.2 | +0.2 | +0.2 | +0.4 |
Two misses stand out. Québec solidaire was underestimated by 7 points in Montreal ridings: its vote there is more concentrated than we thought, and it took Rosemont, Taschereau and Sherbrooke. And the CAQ collapsed mostly in its former strongholds of the Beauce and Quebec City, to the Conservatives' benefit, while we spread its drop more evenly.
Harfang versus Orbit
The two engines picked different favourites in 10 ridings. Orbit was right 6 times, Harfang 3 times.
| Riding | Harfang | Orbit | Winner |
|---|---|---|---|
| Sherbrooke | PQ 56% | QS 75% | QS |
| Orford | QLP 75% | PQ 56% | QLP |
| Daniel-Johnson | PCQ 51% | PQ 74% | PQ |
| Marie-Lacoste-Gérin-Lajoie | PCQ 66% | PQ 69% | PQ |
| Granby | CAQ 54% | PQ 60% | PQ |
| Maurice-Richard | PQ 38% | QS 60% | QLP |
| Bellefeuille | CAQ 33% | PQ 85% | PQ |
| Chapleau | QLP 73% | PQ 34% | QLP |
| Champlain | PCQ 48% | PQ 44% | PCQ |
| Rivière-du-Loup–Témiscouata–Les Basques | CAQ 41% | PQ 89% | PQ |
Orbit on a regular computer and on a quantum chip
On the eve of the vote, Orbit drew 65,536 elections on IBM's Fez quantum chip, in 80 seconds of machine time. Same model, two machines: here is what each gave.
| Regular computer | Fez chip | Result | |
|---|---|---|---|
| PQ (seats) | 59.8 | 58.8 | 59 |
| QLP (seats) | 37.4 | 37.1 | 40 |
| PCQ (seats) | 17.0 | 17.3 | 19 |
| QS (seats) | 7.9 | 7.9 | 9 |
| CAQ (seats) | 4.8 | 5.8 | 0 |
| Right winners | 112 | 112 | 127 |
| Chance of a PQ majority | 37% | 32% | — |
The chip reproduces the regular calculation almost exactly: the same number of right winners, and a single riding where the favourite changes (Jean-Lesage, won by QS, which both machines miss). The chip's noise moves about one seat, mostly toward the CAQ. Today's quantum machines do not make the projection more accurate; they redo it, with a little noise.
Fastest Call, election night
We called the winner in all 127 ridings, and the count overturned none of the calls. 55 were called automatically, when both our engines gave the green light; 72 after a check, including 4 against the model's view, all right. The PQ government was called at 8:27 p.m., its minority at 9:22 p.m..
127 calls · a median 39 minutes after polls closed
Our calls versus the major networks
Two calls mattered on Monday night: who would form the government, then whether it would be a majority or a minority.
| PQ to form government | Minority government | |
|---|---|---|
| Vote-Scope | 8:27 p.m. | 9:22 p.m. |
| CTV News | about 8:18 p.m. | 9:38 p.m. |
| Radio-Canada | 8:37 p.m. | 10:00 p.m. |
| TVA Nouvelles | about 8:45 p.m. | about 10 p.m. |
CTV was first to call the PQ government, nine minutes before us. On the minority, the call that settled the night, we were first: 16 minutes ahead of CTV and about 38 minutes ahead of Radio-Canada and TVA. If you followed the night on Vote-Scope, you could go to bed early with no surprise the next morning: none of our 127 calls was overturned.
Network times from their own reports and live blogs; "about" where they do not give the minute.
What we take away
- Our vote ranges were too narrow when every pollster misses in the same direction, as with the CAQ and QS. We will widen them to account for that shared error.
- Our map must better track where a party gains or loses: a vote concentrated in Montreal for QS, a drop concentrated in its strongholds for the CAQ.
- The blend of our two engines picked more right winners than either. We will give it a bigger role.
- On election night, show the seat median and the count of ridings each party leads side by side, saying in advance how many ridings may flip: that is the most honest picture.
Preliminary results from Élections Québec, all polls counted; this review will be updated with the official results. Projections as published on the eve of the vote.