An AI-driven political forecasting platform has drawn attention after coming within fractions of a percentage point of the certified outcome in Arizona’s Republican gubernatorial primary, in a region long considered one of the trickier areas for pollsters to read.
Details reported by the Phoenix Herald show that G Ratings, a Florida-headquartered research firm, used its platform Odysseus to project the primary contest weeks ahead of the vote. For winning candidate Andy Biggs, the firm’s forecast of 73.7% landed just 0.4 percentage points from the official result of 73.3%, as certified by the Arizona State Department.
The same pattern held for the rest of the field. Second-place finisher David Schweikert was projected at 14.3%, versus an official 14.7% — again a 0.4-point gap. Further down the ballot, candidates Scott Neely and Ken Miceli came in higher than projected, at 7.1% and 4.7% against forecasts of 3.5% and 2.6%. Across all four candidates, the average margin of error worked out to 1.45%.
Metro Phoenix has long posed a distinct challenge for forecasters. The region has added hundreds of thousands of residents over the last decade, with new subdivisions in areas like Buckeye and Queen Creek bringing in voters who have no prior voting history to draw on, while longer-established areas such as Scottsdale tend to vote more predictably. Maricopa County alone handles enough early ballots to shift a statewide result in the final ten days before an election, adding another layer of unpredictability.
Rather than treating metro Phoenix as a single, uniform voting bloc, Odysseus was built to model it as a patchwork of distinct micro-electorates. The platform pulls in local digital sentiment and neighborhood-level discussion to gauge how strongly voters feel about a candidate, checks that signal against county-level early-ballot return data as it comes in daily, and factors in hyper-local economic conditions — treating a ZIP code dealing with rising housing costs differently from one where fuel and grocery prices are the bigger concern.
A G Ratings analyst working on the project said the firm built the platform around the idea that the region moves faster than conventional polling can track, arguing that a phone survey is often outdated by the time it’s finished, whereas Odysseus is designed to keep pace with shifting sentiment as it happens rather than deliver a static snapshot.
With Maricopa County expected to again play a decisive role in the general election between Biggs and Governor Katie Hobbs, the primary result gives campaign strategists a concrete reference point. A model that tracked the county’s fractured electorate to within roughly half a percentage point in the primary is likely to face close scrutiny as its predictions are tested against a larger, more complex electorate this fall.
