Using Predictly, step by step
How to write a question Predictly can answer, what it gives back, and how to read each part of the answer.
In four moves
- You ask the questionIn English or French, in the field on the home page. Between 10 and 300 characters. A free account is needed to start the analysis.
- Predictly researchesIt identifies the event, runs its research across several providers, then sorts and weighs what it finds. This takes a few tens of seconds.
- You get the answerThe most likely outcome, its percentage, a confidence level, what argues each way, and the sources consulted.
- You find it againEvery forecast is saved automatically and stays in your history. You can also share its public link.
What makes a good question
- It is about something not yet settled. Predictly declines to call an event that has already happened.
- It is about one subject. Two questions in one give an answer that properly answers neither.
- It implies a deadline, explicitly or not: “this season”, “before the end of the year”, “in 2027”.
- The name in it means one thing. If the name is ambiguous, Predictly will ask you to narrow it rather than choose for you.
- It fits in a sentence. Between 10 and 300 characters.
Examples
- Works well
- “Will PSG win the Champions League this season?” — one subject, one deadline, a checkable outcome.
- Works well
- “Will the next iPhone ship before next October?” — a clear resolution date.
- Rephrase it
- “Who won the last World Cup?” — that is a question about the past. Predictly forecasts; it does not recount.
- Rephrase it
- “Is AI dangerous?” — that is an opinion, not an event anyone will one day know happened or not.
- Narrow it
- A question whose name means two things at once: Predictly will name the readings and hand it back to you.
Reading the answer
- The answer
- The most likely outcome, in as few words as possible. Sometimes “yes” or “no”, sometimes a name.
- The percentage
- The estimated probability of that outcome, rounded to the nearest 5 points.
- The confidence level
- How well documented the answer is — not the same thing as the probability.
- What argues for / against
- The points kept on each side, tied to what was read.
- The sources
- The documents actually consulted, most-cited first, with their dates where known.
- The resolution note
- What will determine whether the forecast held, and by when.
What the percentage and the confidence actually mean is covered in understanding probabilities.
When Predictly does not answer
- The event is already settled: Predictly says so rather than presenting a known fact as a forecast.
- The subject is ambiguous: the possible readings are named and the question comes back to you.
- The evidence is not there: no percentage is shown, and Predictly says it has not enough to go on.
In all three cases the refusal is deliberate. An engine that produces a number whatever happens mostly produces numbers that mean nothing.
The account, the forecasts and the credits
A free account is needed to start an analysis: a forecast costs a real research run and several calls to the engine, and the account is what keeps that spending from being open to the world.
The first forecast on each account is free. The ones after it use a credit. Credits can be bought singly, in packs or through a subscription; the current amounts are shown at the point of purchase, and the full rules are in the terms.
English or French
The language changes the presentation, not the forecast. Predictly identifies the event before researching it, so the same question asked in either language gives the same answer, written in yours.
How that identification works is described in the methodology.