Who is sharpest this quarter?

Will September inflation show prices up less than 3% from a year ago?

Will the #1 film on the Oct 9 to 11 weekend gross over $40 million in the US?

Best Track Record

The most accurate record on the quarter’s Editors’ Set among contestants, with enough answered questions behind it.

Prediction of the Quarter

A standout successful contrarian prediction. Among answered questions with at least 30 members predicting, the winner is the highest-scoring prediction first entered in the first half of the question’s open period, when the crowd average (not counting the member’s own prediction) at the freeze, or just before an early close, was on the other side of 50%, and whose scored entry stayed on the same side of 50% as the first. Ties go to the earliest entry. If nothing qualifies, it is not awarded. Open to every ranked member.

Rising Predictor

The best Editors’ Set score among contestants whose record is under 12 months old.

Best Forecasting Team or Model

The top team or AI model, with the real people behind it. First awarded when our team and model track opens, which is later.

Live scoreboard

The best predictions are named as questions are answered.

Quarterly awards

The best scores and standout predictions of the quarter.

Annual awards

The best scores and standout predictions of the year.

1 · We publish the question

Every question has an exact outcome, a deadline, a named public place where we check the answer, a plain definition of its key terms, a line on what does not count, and a rule for what happens if the event is cancelled. Two editors test each question before it goes live: each decides some made-up cases on their own, and if they disagree the wording is rewritten. Every deadline is 11:59 PM Eastern Time and is also shown in your local time. Each question has a deadline for predictions, an event date and a resolution date, the last day the answer can come in. We write questions, and members can suggest them. Every question is reviewed on the same standards.

2 · You make a prediction

Pick a chance from 1% to 99%. You can update it at no cost until 24 hours before the deadline, and your last update before then is the one that counts. New predictions close at the same moment, so everyone has the same information. If the named source publishes the answer before the deadline, the question closes at that moment, and your last update before then (or before the freeze, if that came first) is the one scored. A prediction entered after that moment is not counted, even if we record the answer later.

3 · We record the time

Our own server stamps every prediction with the date and time. We do not copy time stamps or records from anywhere else. A prediction counts only if our stamp is earlier than the answer.

4 · The answer comes in

When the event is decided, we check the named source and post a proposed result, Yes, No or Cancelled, with the evidence linked. For 24 hours any member can flag it. After that the result is final, and scores are set. If the source corrects itself later, we correct the result in public and rescore.

5 · We score it and name the best

Every prediction gets a score. The best predictions are announced with the member’s name. Other predictions stay private unless their owner shares them.

Every answered question counts

Your rank rests on all the questions you predicted that have been answered in the ranking window, and at least 10 are needed (5 for a market scoreboard). A new ranking starts each quarter, and your record stays on your page.

Market scoreboards

Each market has its own scoreboard, built only from the answered questions in that market.

Questions from the community

Questions suggested by members are scored and shown like any other. They count for everyone except the member who suggested them, whose own main rank leaves them out. Sponsor suggestions work the same way.

One deadline for everyone

Every question has one deadline for all players, so nobody gets a later look at the news than anyone else.

Followers are not scores

We record follower counts for information. They do not affect a score, a rank or an award.

Quarterly titles and the Editors’ Set

Members choose their own questions, so for titles we compare everyone on the same ones. Each quarter our editors choose about ten events across the markets, the Editors’ Set. Enter at least 70% of the set’s events that are not cancelled (7 of 10, rounded up) to be a contestant for Best Track Record and Rising Predictor. Those titles are scored on the whole set: an event you did not enter counts as a 50% prediction, so everyone is compared on the same events. If you skip the set, you stay on every scoreboard and can still win Prediction of the Quarter. Join early in the quarter to enter the full set. Each contestant’s record shows events entered, events counted as 50%, and the set score, and counted values never appear in your prediction history or answered count.

A plain accuracy score

Your accuracy score runs from 0 to 100. Always predicting 50% earns 50. Higher is better. The highest possible score is 99, because predictions run from 1% to 99%. It is not a percent correct: an 80 does not mean 80% right. It shows how close your numbers were to what happened. Under the hood we use a standard accuracy rule called the Brier score, where lower is better, and show it on this 0 to 100 scale. The order of the rankings is the same either way. The formula is 100 × (1 − the square root of your average Brier score).

Enough answered questions

You are ranked once you have at least 10 answered questions overall, and on a market scoreboard once you have at least 5 answered in that market. Until then your card is an early card. A few lucky predictions can look brilliant, which is why each rank shows how many answered questions sit behind it. Early results show how a group is doing, not proof of skill.

Cancelled means cancelled

If an event is cancelled, ruled invalid, or still undecided by its resolution date, the question is cancelled and left out of scores, with a public note.

When the words are unclear

If a question’s written rule and its plain meaning point different ways and we cannot settle it, we cancel the question instead of stretching the words.

Real people behind models

A team or an AI model is ranked with the named people behind it shown.

One standard for every question

Every question is chosen on the same standards, whoever suggests it. Sponsors support the program and a market scoreboard, BanxCorp is the title sponsor, and sponsored questions are labeled “Sponsored by”.

A question page (public)

The three best predictions, each with the member’s display name, the number given and the score. Also the answer, where we checked it, and how many other predictions the question drew.

A member’s page (public)

Display name, market, titles and best predictions. Once a member passes the minimum, an overall accuracy score is shown with the number of answered questions behind it. A few wins are not proof of overall judgment.

The scoreboard (public)

Members listed by display name and ranked by accuracy at the top of each market, with how many answered questions each rank rests on. Scores count every answered question, including predictions not shown publicly. Other records stay private unless shared.

Your private card (only you)

Every prediction you made, your numbers, your scores and your position. Before the minimum, it is an early card. Prefer to stay anonymous? Use a private ID instead of your name.

Timing

Before the deadline nobody’s number is public. Between the deadline and the answer, no player’s number is public either, unless a player shares their own. After the answer, the best predictions are announced by name.

One method for every title

Every award winner comes from the same record and the same scoring.

No one judges their own question

Our editors do not resolve a question in which they or a sponsor has a stake. A second person confirms every result.

Review on request

If you ask us in writing to review a record, we pause its use while we check.

Public corrections

If we get an answer wrong, we fix it, log the correction by ID and show it beside the record.

House forecasts

Profety also publishes AI forecasts, always labeled AI and shown apart from members. They are entered before the deadline like everyone else’s and never use other members’ numbers. They are a benchmark: they do not count as members and cannot win a title. On the Editors’ Set questions, we show how members and AI forecasts compare, with the number of questions behind it.