ProbabilityGrid combines explainable estimates without hiding where the numbers come from. The current release is an educational forecasting platform, not a real-money exchange.
Probability describes the estimated chance of an outcome. Confidence describes how much evidence supports that estimate. A 70% forecast with low confidence is visibly different from a 70% forecast supported by fresh, independent sources.
The platform model, demonstration crowd sample and virtual-market simulation remain individually visible. The first release uses curated sample consensus data rather than claiming a live public crowd or real order book.
Available signals are clamped to a safe range, converted to log odds, weighted by source role and confidence, and converted back to a probability. Missing signals redistribute their weight across the evidence that is actually available.
Factor controls let you test evidence, timing and execution assumptions without rewriting the published baseline. Your local forecast and virtual positions are stored only on this device when browser storage is available.
Every public question has an unambiguous close time, resolution rule and named reference source. A question without a usable resolution path should remain a draft rather than appear as a public market.
General factor models cover broad questions. Domain-specific calculators can replace the model signal when stronger structure exists. The World Cup simulator is the first specialist adapter.
These are starting weights, not a promise that every signal is equally trustworthy. Confidence, freshness and availability change effective weight. Specialist adapters may provide the model estimate directly.
Virtual points are a local learning tool. They have no monetary value, cannot be transferred and do not represent external liquidity or a tradable price.