FutureLedger

How scoring works

FutureLedger records what public figures say will happen, waits for the facts, and scores the result. This page explains the rules, so you can judge the numbers for yourself.

1. What counts as a prediction

A prediction is a claim, by a named person or organization, about something that has not happened yet, and that can be checked later. We look for:

We leave out:

2. How predictions get in

Predictions are found in articles, interviews and official publications. Software helps find them, but a person reviews every prediction before it appears: checking the wording, who said it, and the date. Every prediction links to its source, so you can read the original. The date shown is when the claim was made, not when we added it.

3. How outcomes are decided

When the deadline has passed, the facts are checked and the prediction gets an outcome: correct, incorrect or partial (or unknown when the evidence doesn't settle it). Each outcome links to its evidence.

4. How a prediction is scored

Each resolved prediction gets a score from 0 to 100, depending on its type.

TypeExampleScore
Yes / no“The merger will close this year.”Correct 100 · Partial 50 · Incorrect 0
Price or level threshold“The index will end the year above 6,000.”Correct 100 · Incorrect 0 — how far past the line doesn't matter
Number (a point estimate)“Inflation will be 3.1% in March.”100 minus the percentage error, never below 0. A miss of 10% scores 90.
Direction“Oil will rise.”Correct 100 · Partial 50 · Incorrect 0
Timing“The launch will happen in 2028.”Right and on time 100 · right but outside the window 50 · Incorrect 0

An unknown outcome gets no score, so it neither helps nor hurts. A prediction with no outcome yet is unresolved and is not counted at all.

5. A person's accuracy

A person's accuracy is the average score of their resolved predictions. It is also shown per topic, using only that topic's predictions. Correct, incorrect and partial counts are shown next to it.

Small samples are hidden. One lucky call is not a track record, so a score and rank are shown only once someone has at least 5 resolved predictions. Until then you see their predictions and results but no percentage. The leaderboards apply the same minimum (and let you raise it).

6. Rankings

Leaderboards sort people by their average score, among those who meet the minimum. They can be filtered by topic and by date range, using only predictions resolved in that range. A high score on a handful of predictions is not the same as a strong long-term record, which is why the minimum exists.

7. What the scores do not mean

8. Corrections

If you think a quote, date or outcome is wrong, write to privacy@futureledger.org with the link to the prediction. We review corrections and fix mistakes. See also our Terms and Privacy Policy.

See the leaderboards →