A live experiment in forecasting elections by how people react.
What we want
Forecasting that is independent, transparent and built on human reactions: how the news moves different groups of voters. Polls can be biased or inaccurate; we show where and why our engine sees something different.
How we do it
Voter personas built from real survey answers, one for each voter group in each state, react to the day's news. Every race is then played out 40,000 times.
How you can check
It is a live experiment, and you can follow its progress on our map, updated every day from October 12, race by race. Every forecast is scored in public every Monday from October 19, against the same benchmarks, right or wrong.
Fair to every side
We show what the engine sees, and why. You judge.
- No campaign, party or political group funds, directs or reviews this work.
- No paid ads or boosted posts, and no coordination with any campaign.
- We never take sides. Where our numbers differ from others, we show where and why.
- Prediction markets are shown beside every forecast, and never feed into it.
- When a part of the engine doesn’t work, we say so, and say what we changed.
Who we are
A small, independent lab.
Our work is built on long-term research focused on social dynamics. NotAPoll.org is published by Scalia Studio. The name says what this is: not a poll. We don’t ask people what they think today; we model how they react.
Corrections, questions and press: hello@notapoll.org