Probs
Five minutes from "what is this?" to your first scored forecast — and how to get the most out of everything else on the site.
01
Browse stocks, crypto and ETFs. Each market has a set of open questions about what its price will do.
02
Don't answer yes or no — answer how likely. 70% means "probably, but I could be wrong". The slider is your answer.
03
When the question's horizon arrives, the outcome is measured from real market data and your forecast is scored automatically.
Every question asks about one concrete, checkable outcome — "Will Apple be higher a month from now?", "Will Bitcoin swing more than 5% this week?" — and each shows two dates:
Questions come in two flavours, and the labels tell you which you're looking at:
The skill is calibration, not bravado. If you say 80% on ten questions, roughly eight should come true — that's what "good" means here, and it's exactly what the scoring rewards.
Forecasting skill compounds. Your logged-in home deals Today's calls — ten quick questions a day — and every day you make at least one forecast extends your streak. The full deck takes a couple of minutes.
To hear the moment your forecasts resolve, turn on email alerts from the notifications page or push alerts in settings. And install Probs as an app on your phone — ten seconds, no app store.
When a question resolves, your probability is compared with what actually happened (a Brier-style rule, the standard in forecasting science). Saying 90% and being right scores brilliantly; saying 90% and being wrong costs you far more than a cautious 60% would have. Scoring is also time-weighted: a good call held for weeks earns more than the same call sniped at the last minute. And beating the crowd pays extra — when your number diverges from the consensus and reality proves you right, a contrarian bonus (up to +50 points) lands on top. Riding the consensus earns no bonus, and a bold call that misses costs nothing beyond its normal score, so the best long-run strategy is simply to report what you truly believe, early.
Points measure skill above chance: a perfect early call earns 100, and answering 50% (a coin flip) earns exactly 0 — so there are no points for hedging, and no way to farm the leaderboard with volume. For the same reason, new forecasts are capped at 50 a day (revising your existing calls is always free). Scores accumulate into your accuracy, your calibration curve on your profile, and your Probs Score.
The leaderboard ranks everyone by one number — the Probs Score (marked ). It multiplies three things: your skill above chance (how far your scoring beats a coin flip — hedging counts for nothing), how many calls you've had resolve (more resolved calls carry more weight, with diminishing returns, so a hot streak on a handful of questions can't leapfrog a deep track record), and your consistency over time. A call that lands below a coin-flip counts as neutral, not a penalty— so your good calls only ever build your score, and one unlucky miss can't drag it to zero (you still need to beat a coin-flip on balance to score at all). That's why a 79% forecaster with 200 resolved calls can sit above an 85% one with only a few — and why accuracy alone is shown as a supporting stat, not the ranking.
Your profile also shows your edge — how much sharper your calls were than the crowd's consensus on the same questions (which automatically adjusts for question difficulty: matching an easy consensus earns nothing) — and your median lead time, how far before close you typically make your call. You need at least a handful of resolved calls to appear on the board at all, and the category and sector boards are scored the same way within each group.
Every call you make is also measured against the crowd. When you forecast, we note the consensus at that moment — your entry. If you disagreed with the crowd and the outcome proves you right, your track record shows it: each resolved call is treated as an equal-sized position taken at the crowd's odds, giving you a return per call and a beat-the-crowd rate on your profile. Update a call and it simply re-enters at the new consensus — there's no reward for piling on at the last minute.
It's hypothetical and equal-weighted — a clean measure of forecasting skill, not investment performance.
The percentage on every question is not a poll average. Each forecast is weighted by the forecaster's track record — proven accuracy, experience, and how fresh the forecast is — then pooled the mathematically-proper way. A reliable forecaster's 70% moves the number more than a newcomer's. The "agreement" label tells you whether forecasters broadly concur or are split.
Every market has a Data & analysis page — the ✨ tab — where the community's odds are turned into concrete numbers:
Live vs. snapshot. The projection numbers on the markets list, the screener and your home refresh a few times a day (a fresh snapshot morning, afternoon and after close). Open a market's Data & analysis deep dive and the projections are recomputed livefrom the crowd's current odds — so they move the moment forecasts do.
Basic data (price chart, stats, track record) is always free. The forward-looking analysis is a deep dive: you get 15 free per day, and Premium (£5/month) makes those live deep dives unlimited, lifts the 10-market watchlist cap, and adds a full readout of your own forecasting skill.
Every market, question and profile has a Share button. Post the link anywhere — X, Discord, WhatsApp — and it unfurls into a clean card showing the live odds or your track record. Sharing your profile is the quickest way to prove your accuracy and pull other forecasters in: the more sharp people forecasting, the sharper the consensus gets for everyone.
Everything on Probs is a measurement of expectations, not advice. Projections are model estimates built from community odds — treat them as one more input for your own thinking, never as a promise.