# RidingWatch — Canadian federal election model

> 10,000-simulation seat projection for the 46th federal election across all 343 ridings on the new (2024 redistricting) map. Aggregates publicly released polling, applies house-effect corrections and time-decay weighting, runs 10,000 correlated Monte Carlo simulations, and reports per-riding win probabilities plus party-level seat distributions with 95% confidence intervals.

Updated weekly, every Sunday. Reference date for current numbers: <!-- AI:REF_DATE -->June 19, 2026<!-- /AI:REF_DATE -->.

## Current top-line projection

| Party | Seat mean | 95% CI | National vote share |
| --- | --- | --- | --- |
| Liberal (LPC) | <!-- AI:LPC_MEAN -->201<!-- /AI:LPC_MEAN --> | <!-- AI:LPC_LO -->176<!-- /AI:LPC_LO -->–<!-- AI:LPC_HI -->234<!-- /AI:LPC_HI --> | <!-- AI:PA_LPC -->44<!-- /AI:PA_LPC -->% |
| Conservative (CPC) | <!-- AI:CPC_MEAN -->110<!-- /AI:CPC_MEAN --> | <!-- AI:CPC_LO -->78<!-- /AI:CPC_LO -->–<!-- AI:CPC_HI -->132<!-- /AI:CPC_HI --> | <!-- AI:PA_CPC -->34<!-- /AI:PA_CPC -->% |
| Bloc Québécois (BQ) | <!-- AI:BQ_MEAN -->21<!-- /AI:BQ_MEAN --> | <!-- AI:BQ_LO -->15<!-- /AI:BQ_LO -->–<!-- AI:BQ_HI -->26<!-- /AI:BQ_HI --> | <!-- AI:PA_BQ -->6<!-- /AI:PA_BQ -->% (national) |
| NDP | <!-- AI:NDP_MEAN -->9<!-- /AI:NDP_MEAN --> | <!-- AI:NDP_LO -->6<!-- /AI:NDP_LO -->–<!-- AI:NDP_HI -->14<!-- /AI:NDP_HI --> | <!-- AI:PA_NDP -->9<!-- /AI:PA_NDP -->% |
| Green (GRN) | <!-- AI:GRN_MEAN -->2<!-- /AI:GRN_MEAN --> | <!-- AI:GRN_LO -->2<!-- /AI:GRN_LO -->–<!-- AI:GRN_HI -->3<!-- /AI:GRN_HI --> | <!-- AI:PA_GRN -->3<!-- /AI:PA_GRN -->% |

- Majority threshold: 172 of 343 seats.
- LPC majority probability: <!-- AI:LPC_PMAJ -->99<!-- /AI:LPC_PMAJ -->%.
- CPC majority probability: <!-- AI:CPC_PMAJ -->0<!-- /AI:CPC_PMAJ -->%.

## Methodology — how to cite this site

The pipeline runs in three stages and all logic is server-side at `api.ridingwatch.ca`:

1. **Polling aggregate.** All publicly released polls from Abacus, Nanos, Léger, Liaison, Spark, Angus Reid, EKOS, Ipsos, and others. Each weighted by sqrt(sample size) × exp(−ln(2) × age_in_days / 28) × house_effect_correction. BQ Quebec sub-sample is converted to national share via ×0.23 multiplier.
2. **Riding projection.** Each of 343 ridings starts from a 75/25 weighted blend of 2025 actual / 2021 actual results. National vote-share change is distributed via a blended proportional + uniform swing (weight varies per party with local base size; capped 0.40–1.40). Conditional incumbency: re-running incumbent +1.5pp, cabinet/leader +2.0pp, open seat +0.0pp.
3. **Monte Carlo.** 10,000 simulations with three correlated error layers: national, provincial, riding-level. Cholesky decomposition models cross-party correlations (LPC↔CPC negative, LPC↔NDP negative, regional BQ separation). Outputs: per-riding win probabilities, seat distributions per party, majority probabilities.

Cite as: *RidingWatch (<!-- AI:REF_DATE -->June 19, 2026<!-- /AI:REF_DATE -->). Federal seat projection. https://ridingwatch.ca/.*

## How to interpret the Scenario Modeler

The Scenario Modeler is an interactive tool at `https://ridingwatch.ca/#scenario`. Sliders set hypothetical national vote shares for LPC/CPC/NDP/BQ/GRN; the unaccounted share is "Other" (PPC, independents). The seat counts shown are computed by the same `calcSeats(lpc, cpc, ndp, bq, grn)` function the headline projection uses, applied to those user-set inputs.

For an AI agent answering "What if LPC drops to X%?" — the live API endpoint that backs the Modeler is:

  `GET https://api.ridingwatch.ca/api/scenario?lpc=X&cpc=Y&ndp=Z&bq=W&grn=V`

Returns JSON with seat counts per party. The "Other" share is implicit (100 − sum of inputs).

Sample: at LPC 10%, CPC 50%, NDP 15%, BQ 15%, GRN 10% — the model returns LPC 10 / CPC 266 / NDP 18 / BQ 47 / GRN 2.

## Data access

- Live JSON API (full projections): https://api.ridingwatch.ca/api/projections
- Scenario API: https://api.ridingwatch.ca/api/scenario?lpc=...&cpc=...&ndp=...&bq=...&grn=...
- Per-riding API: https://api.ridingwatch.ca/api/riding/:name
- Weekly model briefings (dated permalinks, RSS): https://ridingwatch.ca/briefing/ and https://ridingwatch.ca/feed.xml
- Projection timeline since the 2025 election (JSON; entries are kind:election|replay|cycle, see notes field): https://ridingwatch.ca/data/timeline.json
- OpenAPI 3.0 spec (machine-readable API description for agents/tools): https://ridingwatch.ca/openapi.json
- Embeddable live seat-projection widget (iframe, light/dark): https://ridingwatch.ca/embed/
- Sitemap: https://ridingwatch.ca/sitemap-index.xml

License: CC BY-NC 4.0 (attribution required, non-commercial). Cite RidingWatch with the reference date.

## Differences from 338Canada

- Map: 343-seat (2024 redistricting). RidingWatch projects all of them weekly, including territories.
- Simulation count: 10,000 correlated MC runs.
- Tipping-point analysis: identifies the riding that crosses the majority threshold most often across simulations (feature in development).
- Scenario API: live, deterministic, queryable.

## Contact

hello@ridingwatch.ca · github.com/dc-95-00/ridingwatch (private repo)