Game Systems/The Model

Polling

The polls you see are not the truth. They are a deliberately blurred view of the real model, and learning to read them through that blur is part of the game.

What a poll actually is#

A reported poll is the true bloc-aggregated share, plus two distortions:

Sampling noise, scaled by the poll's sample size.
A house effect — a persistent, per-pollster lean.

The pollsters#

Pollster	House effect	Sample
Quinnipiac	+0.5 D	1,100
Marist	−0.4 D	1,000
Siena/NYT	+0.2 D	900
Trafalgar	−1.2 D	1,050
Monmouth	+0.3 D	800

Trafalgar is the strongest house effect on the board at 1.2 points, and it leans Republican. A single Trafalgar poll showing your Democratic candidate down a point is consistent with a tied race.

Sample sizes vary, so margin of error varies with them. Monmouth's 800 is noisier than Quinnipiac's 1,100.

Determinism#

Polls are deterministic for a given game seed, turn, state and pollster. The same poll does not reshuffle when the UI re-renders, and reloading a save does not reroll it.

That determinism is a fairness guarantee, not a convenience: you cannot save-scum a better poll, and a poll you saw is a poll that was really generated for that week.

Reading through the blur#

Average across pollsters. House effects roughly cancel across the five; a single poll does not.
Correct for known leans. Trafalgar at −1.2 and Quinnipiac at +0.5 are 1.7 points apart before anyone has been sampled.
Trends beat levels. Noise is per-poll; a direction sustained across several weeks and several pollsters is signal.
Small samples move more. An 800-sample poll swinging three points may be nothing at all.
Your own projection is not a poll. The projected electoral vote in the recap comes from the model directly — that is ground truth.

What this means in play#

The gap between the polls and the model is a genuine information problem. You are making resource decisions — where to send the candidate, where to buy ads — on evidence that is systematically distorted in known ways.

Chasing a single poll is how campaigns waste weeks. The recap and the timeline are truthful; the polls are the fog.