Live Election-Night Modeling

Live results modeling, licensed for your site.

We build county-level Bayesian election-night models that update live as votes are counted, projecting the final margin, a win probability, and a full distribution of outcomes minute by minute. License the build for your own results page, or run it privately for your team.

20,000
simulations per update
60s
refresh against the county feed
County
level, not statewide averages
What You're Licensing

A live results page, not a static forecast.

Our pre-election models tell you what's likely to happen before a single vote is counted. This is the other half: a page that watches the count happen and tells your readers what it means, race by race, county by county, as the numbers come in.

Every build starts from the same core: a correlated Bayesian model that holds counted votes fixed and projects only what's left, drawing on a county's known partisan lean, its reporting history, and how much of its vote is still outstanding. As returns land, the model narrows in on a final margin and a win probability instead of just repeating whatever the running tally says.

The page itself is built to survive election night: no dependency on a mapping library or a CDN that might go down at the worst possible moment, a live margin gauge, county-level maps of what's counted and what's still out, and a live posterior distribution so readers can see the uncertainty, not just a single number.

What ships

A self-contained build for your race

  • Live margin gauge, county maps, and a margin-over-time chart
  • Backend that polls the results feed and re-runs the model on a fixed interval
  • Your colors, your typography, your domain
  • A written methodology section explaining the model in plain terms
Methodology

How the model stays honest while the count is still moving.

The same approach that powers our own live race pages, adapted to whichever state or district you need covered.

01

Counted votes held fixed

Only the uncounted remainder is projected, and it's projected at the margin appropriate to whichever vote mode (early, mail, Election Day) is actually left in that county, not a single blended average.

02

Correlated county-level error

A surprise in one county moves its region more than it moves the rest of the state, decomposed into a statewide shift, a regional shift, and demographic similarity, with a cap on how far any single county can move the topline.

03

20,000 simulations, every update

Each refresh re-runs a full Monte Carlo across every county to produce a win probability, a projected margin, and 50/90 intervals, not just a point estimate.

Licensing

Built for your site, or kept entirely private.

Every model is a custom build for the specific race you need. The license just determines who sees the results.

Public / White-Label

Display it on your results site

We build the live model to your race and embed it on your own domain under your own branding, ready to run on election night.

  • Your colors, fonts, and layout applied to the build
  • Attribution to SoCal Strategies as the modeling partner
  • Support during election night while the model is live
Private

Run it for your team only

The same live model, delivered as a private page or feed for internal use: a campaign's war room, a newsroom's backend, or an organization tracking a race without publishing it.

  • No public branding or attribution required
  • Same underlying methodology as our public builds

Pricing depends on the race's complexity and how much lead time we have before election night. Reach out early so we can plan the county-level model around your specific results feed.

Email us for a quote