How AllOdds works

To forecast a matchup: assign a probability (0–100%) that the Democrat wins. The complement is the Republican's probability. You don't need to forecast every pairing — just the ones you have a view on. Your forecasts are combined with other users' estimates to form a community aggregate for each cell in the grid.

AllOdds does not collect user forecasts for presidential primary nominations or the overall general-election winner question. Those odds are sourced automatically from Polymarket, a real-money prediction market where traders put capital behind their predictions.

This is an intentional design choice. Primary odds and overall winner markets are things Polymarket already does well — real money creates accurate incentives and the markets are liquid enough to trust. AllOdds's unique contribution is the matchup grid, which no single market can produce efficiently. Keeping the site focused there means your forecast energy goes where it actually adds new information.

Polymarket data is displayed transparently on each primary and overall winner question page. If a candidate appears on Polymarket but hasn't been added to AllOdds yet, the site flags it for review.

The matchup grid is the central product of AllOdds. It shows every Democrat vs. every Republican head-to-head — each pairing as its own independently-forecasted question — in a single view.

The key insight the grid surfaces: the best candidate for each party depends on who they're running against. A populist Republican may be competitive against an establishment Democrat and collapse against a populist one. That interaction is invisible in any single market number — it only shows up when you look at the full matrix.

A second insight: the nomination frontrunner isn't always the most electable. The grid lets you compare each candidate's win probabilities across all opponents alongside their nomination odds, making visible where the likely nominee leaves wins on the table.

Each cell shows the community's estimated probability for that specific pairing. The estimates are only as good as the forecasters behind them — which is why participating matters. The more people forecast each cell, the more informative the surface becomes.

Open the Matchup Grid →

Questions are scored after they resolve (i.e., once the winner is known). Your score on a question has three components that are blended together.

Your leaderboard total is the sum of blended scores across all resolved questions you forecasted. You are only scored on questions you participated in — skipping a question never hurts you.

Because scoring is time-weighted, it matters whether your forecast reflects what you actively believe right now — or just a prediction you made once and forgot about. AllOdds tracks whether your forecast has gone stale.

A forecast becomes stale when it hasn't been updated or reaffirmed in at least 10% of the question's total lifetime (minimum 14 days). For example, on a question that runs two years, your forecast goes stale if you haven't touched it in about 73 days.

When a forecast is stale, you'll see an amber notice on the question page with a Reaffirm forecast button. Clicking it tells the scoring system you've reviewed your probabilities and still stand by them — without changing anything. It resets the staleness clock and logs a new entry in your forecast history.

The leaderboard ranks users by their total blended score across all resolved questions. Both an all-time ranking and a monthly ranking are shown once questions start resolving.

There are currently no prizes, but we are actively seeking funding and partnerships to introduce a reward structure. Forecasters who perform well early will have a documented track record once prizes come online.

Monthly rankings reset on the first of each month. They reflect only blended scores from questions that resolved during that calendar month.