How ElectionPlus.AI Works
The game. The AI. The scoring.
ElectionPlus.AI turns Arizona local elections into prediction games. Our algorithm sets the baseline — you try to beat it with local knowledge. Here's everything you need to know.
The prediction game
How the game works.
Every Arizona election on ElectionPlus.AI is a prediction game. Each game is sponsored by a local business. Here's how you play.
Browse open games
Each prediction game corresponds to an Arizona election — mayoral races, city council, school boards, county races, and ballot initiatives. Every game shows its sponsor and the AI's current prediction.
See the AI's prediction
Before you make your call, you can see what ElectionPlus.AI's algorithm predicts. The AI shows its confidence percentage for each outcome — this is the benchmark you're trying to beat.
Submit your forecast
Pick the outcome you predict and set your confidence level (1%–99%). Your forecast locks in when polls close. Be honest with your confidence — calibration matters as much as the call.
Get scored and ranked
After results come in, both you and the AI are scored using the Brier score. Beat the AI's accuracy on a market and your points double. Top forecasters climb the leaderboard.
Sponsored games
Every game has a sponsor.
Each prediction game on ElectionPlus.AI is sponsored by an Arizona business. Sponsors help fund the platform and get visibility in front of civically engaged Arizonans.
Sponsor-branded markets
Each prediction game displays "Sponsored by [Business Name]" at the top. Sponsors choose which markets they want to be associated with — local races in their community.
Civic engagement
Sponsors support an informed electorate by making prediction games available. The platform drives awareness and engagement around local elections that often get overlooked.
Current Arizona sponsors
Arizona Federal Credit Union, Postino Wine Cafe, Press Coffee, Four Peaks Brewing, Changing Hands Bookstore, Barrio Brewing, Arizona Wilderness, and more.
Scoring system
The math behind the leaderboard.
ElectionPlus.AI uses the Brier score — the same accuracy metric trusted by professional forecasting organizations like FiveThirtyEight, IARPA, and the Good Judgment Project.
Confident + right = max points
Predicted 90% Yes and Yes happened? Near-perfect score. Calibration matters — your confidence level is just as important as getting the outcome right.
Confident + wrong = max penalty
Predicted 90% Yes but No happened? Maximum penalty. Overconfidence is the costliest mistake in the Brier scoring system.
2x bonus for beating the AI
If your Brier score on a market is better than the AI's, your points for that market are doubled. This is the core mechanic — humans vs. machine.
Streak multipliers
Beat the AI on 5 consecutive markets at 70%+ accuracy and unlock a 1.25x scoring multiplier for the next 24 hours.
The ElectionPlus.AI algorithm
How our AI makes its predictions.
Understanding the algorithm helps you find the gaps where local knowledge gives you an edge. Here's a transparent look at what powers our predictions.
Data sources
The model ingests publicly available polling data, voter registration numbers, historical turnout by precinct, campaign finance filings (FEC + state), and demographic profiles from US Census data.
Weighting engine
Each data source gets a weight based on recency, reliability, and historical predictive power for the specific race type. Mayoral, council, and school board races use different weight profiles tuned on historical Arizona results.
Historical calibration
Trained on thousands of resolved local elections across the US. The model adjusts for incumbent advantage, off-cycle turnout depression, regional voting tendencies, and party registration shifts since the last election.
What the AI can't see
Yard signs. Community sentiment. Local endorsements. Group chat energy. The vibe at a town hall. These are exactly the signals where humans have the edge over any data-driven model — and why local forecasters consistently beat the algorithm.
Confidence output
For each market the AI outputs a calibrated probability for every outcome. If the AI says 70% across a hundred markets, roughly 70 of them should happen. This calibration is what makes the Brier score meaningful.
Current win rates
Across Arizona markets this season: the AI leads in 58% of all markets. But top-100 human forecasters outperform it in 71% of markets they enter. Local knowledge consistently makes the difference.
In plain English
The ElectionPlus.AI algorithm is a probabilistic forecasting model. It collects every publicly available data point about a race — polls, money, registrations, demographics, history — and uses machine learning to estimate the probability of each outcome. It's trained on thousands of past local elections so it knows patterns like "incumbents in off-cycle municipal races win X% of the time." But it's fundamentally limited to data that exists in databases. It can't knock on doors, and it can't tell if a candidate just had a terrible debate performance. That gap between data and ground truth is the whole game.
FAQ
Questions answered.
Is ElectionPlus.AI free?
Yes — completely free to play. No entry fee, no subscription, no credit card required. The platform is supported by local business sponsors.
How does the AI make its predictions?
The AI model uses publicly available data: polling, voter registration, historical turnout, fundraising filings, and census demographics. Each input is weighted by recency and reliability, and the model is trained on thousands of resolved local elections. It outputs calibrated probability estimates for each outcome.
What does 'Sponsored by' mean on each prediction game?
Each prediction game is sponsored by an Arizona business. The sponsor's name appears at the top of the game card. Sponsors support the platform and get visibility with civically engaged Arizonans. Sponsorship has no influence on AI predictions or scoring.
How is my prediction scored?
We use the Brier score — a continuous, calibration-aware scoring rule used by professional forecasters. It measures the squared difference between your predicted probability and the actual outcome. Lower Brier scores are better; we convert them to points so higher is better on the leaderboard. Beat the AI's score on a market and your points double.
What elections are available right now?
We're starting with Arizona as our test market. Currently covering Phoenix, Tucson, Scottsdale, Mesa, Tempe, Chandler, Flagstaff, Gilbert, and Maricopa and Pima counties. More Arizona races are added regularly, and we plan to expand to other states soon.
Can the AI really be beaten?
Absolutely. Top-100 human forecasters beat the AI in 71% of the markets they enter. The AI is strong on data — polling, demographics, historical patterns — but it can't see the on-the-ground signals that locals pick up: yard signs, community energy, local endorsements, town hall turnout. That gap is the whole game.
What's a Brier score?
The Brier score is a proper scoring rule that measures how close your predicted probabilities are to what actually happened. If you predict 80% for an outcome and it happens, your Brier score for that prediction is (1 - 0.8)² = 0.04 (very good). If it doesn't happen, it's (0 - 0.8)² = 0.64 (bad). The key insight: you can't game it by being overconfident — honest calibration is always the best strategy.
Why Arizona first?
Arizona has a rich landscape of local elections, competitive races across the political spectrum, and an engaged electorate. It's the ideal test market for ElectionPlus.AI before expanding nationally. The diversity of race types — from urban Phoenix council seats to rural ballot initiatives — lets us validate the algorithm across many contexts.
How does the leaderboard work?
The leaderboard ranks all forecasters by total points accumulated across resolved markets. Points come from Brier scores (converted so higher is better), with a 2x multiplier for markets where you beat the AI. We show all-time, monthly, and weekly views. The AI itself is also ranked — it's sitting at #3 this season.
Will ElectionPlus.AI expand beyond Arizona?
Yes — Arizona is our test market. We're planning to expand to other states and metro areas once we've validated the platform and algorithm here. Cities like Austin, Denver, Seattle, Atlanta, and more are on our roadmap.