Deep dives into our methodology, feature explanations, and prediction market research.
Most AI analytics tools send one prompt to one model and call it a day. We run six frontier models in parallel, quantify their disagreement, and apply calibration corrections. Here's exactly how it works.
Win rate is a terrible metric for evaluating forecasters. A prediction of 51% on an event that happens counts as a 'win' — but it's barely better than a coin flip. CRPS holds us to a higher standard.
When a large player loads up on one side of a market, the order book tells a story. Our analyser detects these imbalances in real-time and cross-references them with our AI forecast.