How it works.

A purpose-built reasoning pipeline.

Built for one job: calibrated probabilistic forecasting. Where generalist AI gives you a paragraph of hedged language, Prophiq gives you a ranked, calibrated answer - backed by an ensemble of reasoning systems and a consensus engine purpose-built for the task.

Every Prophiq call runs through a multi-stage analytical pipeline. Real-time research is pulled from across the globe, then evaluated independently by an ensemble of frontier reasoning systems. Prophiq's consensus engine merges their answers using a calibrated weighted Borda count - the same scoring method used in academic forecasting research.

Two layers of intelligence, one calibrated answer.

Prophiq runs on two layers of intelligence in parallel. Real-time research pulled from across the globe - market signals, historical patterns, expert commentary, breaking news. And our own calibration record - a proprietary dataset of every call Prophiq has made and every outcome that resolved it, across all four domains.

Both feed every forecast. Each resolved event becomes signal in the next call, sharpening calibration across the system.

We give you probabilities, not predictions.

When Prophiq says "60% Verstappen wins Monaco," it is not predicting that Verstappen wins. It is saying: more likely than not, but far from certain. The right way to read a probability isn't "will this happen?" - it's "how confident should I be?"

This makes calibration matter more than hit rate. A perfectly calibrated forecaster who says 60% should be right 60% of the time - and wrong 40% of the time, by design. Those misses aren't failures. They're the math working as intended.

What the numbers mean.

A probability is a measurement of confidence, not a prediction of outcome. Here's how to read them:

How it's played out.

Recent forecasts and how they resolved.