CalendarAll →

How Capi works

Three rules: only quality polls enter the main estimate, every uncertainty becomes a simulation and every source is cited.

The best polls

In the US, the average uses the five best-ranked pollsters in the Silver Bulletin accuracy ranking, over the last 45 days, corrected for each one's historical bias. Polls commissioned by parties or campaigns are excluded.

Visible uncertainty

Each forecast runs 10,000 simulations. A national error, shared by every state, is added to a state-specific error that grows when there are fewer polls. That is why we show ranges and chances alongside the median.

Each place against itself

On election night we compare each municipality or county with the previous election, so the order in which votes arrive stops misleading. In Brazil, our projection was within 0.1 point of the result with 59% counted.

Spain, province by province

The national average is spread across the 52 constituencies according to the 2023 result and turned into seats with the D'Hondt method, applying the 3% threshold in each province.

The top US pollsters

PollsterRanking positionGrade
NYT/Siena8A
Emerson College14A
Fox News (Beacon/Shaw)20A-
Suffolk University24A-
Marist29A-

Raters tracked: The Cook Political Report, Decision Desk HQ, The Economist, FiftyPlusOne, Fox News, Inside Elections, MS NOW, NBC News, RealClearPolitics, Sabato's Crystal Ball, Silver Bulletin, Split Ticket.

US model parameters

Polling window45 days
National error, Senate (std. dev.)3.0 points
National error, House3.5 points
State-level error4.0 points with 3+ pollsters; 4.5 with 2; 5.0 with 1; 6.0 with no top pollster
House district error6.0 points
Incumbency advantage3 points

Sources and licenses

TSE (Brazil); Interior Ministry (Spain); Israel's Central Elections Committee; Wikipedia (CC BY-SA) for polling and rating tables; University of Florida Election Lab (CC BY-NC-ND 4.0) for early voting; MEDSL (CC0) for results by vote mode; Silver Bulletin for the pollster ranking.