geekLections Forecasting Engine · Arm 1 (PyMC v5) + Arm 2 (Svelte 5) Gamma-Quotient Simplex + Booth GP

Bayesian Multi-Party Polling Aggregation & Lower-House Booth Model

Gamma-Quotient Bayesian State-Space Polling Fan Chart (50% & 90% Credible Ribbons)

Loading posterior polling trajectory...

Pollster House Effects Forest Plot (CLR-Anchored Percentage-Point Bias)

Loading pollster house effects...

Interactive Gamma-Quotient (GQD) & Division 2CP Posterior Simulator (Outputs A–E)

Live Browser Monte Carlo (N = 2,500)

Adjust individual party primary shares and party-specific standard deviations (σ_p) below. The browser computes the exact Gamma Quotient Distribution parameters (α_p, β_p), simulates preference transfers, and outputs (A) 2PP Beta, (B) First-Preference GQD, (C) 2CP Matchup Probabilities, (D) Conditional 2CP Betas, and (E) Candidate Win Probabilities.

Output (E) — Overall Candidate Win Probabilities (Σ P(Win_c) = 100%):

ALP

93.0%

LNP

7.0%

GRN

0.0%

IND

0.0%

ONP

0.0%

Output (B) — Division First-Preference Gamma Quotient Distribution GQD(α, β):

Party	Simplex μ_p	σ_p	GQD α_p	GQD β_p
ALP Labor (ALP)	31.5%	±2.2%	96.2	304.4
LNP Coalition (LNP)	34.5%	±2.4%	88.7	256.0
GRN The Greens (GRN)	14.5%	±1.8%	47.4	326.2
IND Independent / Teal	12.5%	±2.6%	17.7	140.6
ONP One Nation (ONP)	7.0%	±1.4%	21.6	307.9

Output (A) — Classic 2PP Beta (ALP vs LNP)

θ_2PP ~ Beta(169.0, 153.5)

ALP 2PP Mean: 52.4% (±2.8%)

Outputs (C) & (D) — 2CP Matchup Probabilities & Conditional Betas

Final-Two Pair	Matchup Prob (C)	Conditional Beta(a, b) (D)	Cand 1 Mean	P(Win₁ \| Pair)
ALP vs LNP	100.0%	`Beta(142.8, 121.0)`	54.1%	93.0%

Mathematical Summary & Manual CLI Pipeline

1. Why Gamma Quotients (Scaled Dirichlet) Instead of Standard Dirichlet?

A standard n-party Dirichlet distribution Dir(α₁, …, α_n) has only one global concentration parameter α₀ = Σα_i, locking all n party variances to Var(X_i) = μ_i(1 − μ_i) / (α₀ + 1). By drawing independent Gamma random variables Y_i ~ Gamma(α_i, β_i) with party-specific rate parameters β_i and normalizing P_i = Y_i / ΣY_j, we gain 2n − 1 degrees of freedom on the simplex—allowing major parties (ALP, L/NP) and minor/micro parties (GRN, ONP, Ind/Teals) to exhibit realistic, independent volatility and pollster house effects while remaining strictly order-invariant.

2. Lower-House Booth Covariance, Redistributions & Declaration Votes (Outputs A–E)

On election night, ordinary polling booths report early while Pre-Poll Voting Centres (PPVCs), Postal Votes (modeled via daily AEC Issued PVA counts × return rate ρ_d and arrival curve F(t)), and Absent votes arrive over subsequent days with systematic late-Coalition / Green-Absent tilts. Our spatial-demographic Gaussian Process updates unobserved booths and non-ordinary vote pools in real time, producing for every electoral division:

Output (A): Marginal Two-Party Preferred (2PP) Beta distribution Beta(α_2PP, β_2PP).
Output (B): First-Preference Gamma-Quotient Distribution GQD(α_d, β_d).
Output (C): Final Two-Candidate Preferred (2CP) matchup probabilities Pr(2CP = {i, j}) across Classic, Green, Teal/Independent, and One Nation contests.
Output (D): Conditional Beta distribution Beta(α_ij, β_ij) for every plausible 2CP pair.
Output (E): Total candidate win probability W_i ∈ [0, 100%] marginalizing over all 2CP matchups.

3. Running the Python Pipeline Locally

Raw AEC CSVs, boundary shapefiles, and PyMC NetCDF chains are stored outside Git in your cloud-synced GL_DATA_DIR directory, while distilled JSON summaries with SHA-256 provenance manifests are written to static/data/:

uv run python -m pipeline.cli check-env
uv run python -m pipeline.cli generate-demo-data