METHODOLOGY

Data and model methodology

Sources, transformations, current coefficients and the validation gates for MP-26.

MODEL PIPELINE

From raw observation to published probability

Polling, official results, ratings, demographics and campaign-finance data.

Preserve field dates, sample, population, mode, sponsor and source URL.

Weight the national environment and run 50,000 seeded, correlated simulations.

The probability engine is backtested on 2018 and 2022; the benchmark itself cannot be replayed, so the model stays experimental.

Use AI for retrieval and synthesis while keeping every numeric claim grounded.

SOURCE REGISTER