Probs
Probs is a forecasting site for markets. You make probability calls on specific questions, like whether NVIDIA closes higher this week or Bitcoin falls 10% this month, and when the answer arrives your call is scored against it. From thousands of scored calls the site builds two things: a live consensus on where markets are heading, and a public record of who actually predicts them well.
Waze doesn't know the roads. Its drivers do, and the app's job is to collect what they see and trust the reliable ones more. Probs runs on the same idea. The site holds no view on Apple or gold. It gathers forecasts from anyone willing to be scored, weights each one by that forecaster's record of accuracy, calibration and consistency, and publishes the blend. A brand-new account counts. Someone who has been right for two years counts for more.
We write the questions ourselves rather than crowdsourcing them, and each one resolves on an objective fact: a price on a date, a threshold crossed or not. Scoring only means something when the question has a clean answer, and the rankings only mean something when the scoring does.
The questions also come in ladders. Will it be higher, will it gain 5%, will it fall 12%? Each answer pins one point on a curve, and together they let the site rebuild the full distribution of returns the crowd expects for an asset: projected prices with an explicit 80% range, implied volatility, skew. Option markets use the same mathematics to read expectations out of prices; we run it on forecasts. Everything shown is a measurement of what forecasters expect, labelled as an estimate. None of it is advice.
Every call is stamped with the consensus at the moment you made it. When the question resolves, the call settles as if you had taken an equal-sized position against the crowd at those odds, which gives your profile a return per call and a rate of beating the consensus. Matching the crowd earns nothing. The record is built by being right where the crowd was wrong, and after enough resolved calls it becomes hard to fake and impossible to buy.
You can follow forecasters whose record you rate. Their calls land on your dashboard, with a flag when one of them takes the opposite side of a call you've made, because one of you is about to be proven wrong. The site also publishes its own calibration, meaning how often its 70% calls actually happen, so you can check the engine instead of taking our word for it.
Forecasting is free and stays free: the consensus, the rankings, and 15 full market deep-dives a day, with basic market data never limited. Premium costs £5 a month and removes the deep-dive limit, lifts the watchlist cap, and adds a complete readout of your own skill, including your calibration curve and accuracy by category. Premium pays for the site. It doesn't gate the core of it.
Questions, ideas, or something broken? probsquestions@gmail.com