Name: Metaculus: editorial review
Rating: 4
Author: PredictionToolkit

What is Metaculus?

Read a question’s resolution criteria before entering a percentage or distribution. Binary questions accept a slider or numeric field; multiple-choice questions allocate probability across mutually exclusive outcomes. Numerical and date forecasts describe uncertainty across a range, with mixtures of up to 4 weighted components. Community aggregates and individual reasoning provide additional perspectives.

Scoring needs its own interpretation. Baseline measures performance against a fixed reference, while Peer compares forecasts on the same question. Standard time-averaged scoring accounts for how long each forecast stood, so a late revision does not erase earlier estimates. These scores are forecasting measures, not invested-money returns or executable trading prices.

Who it is for

Users practicing calibrated probability forecasting.
Researchers comparing human and AI forecasts.
Teams organizing structured forecasting questions.

Features and coverage

Binary and multiple-choice probabilities
Enter a binary percentage or allocate probabilities across mutually exclusive answers. Multiple-choice totals must equal 100%.
Numerical and date distributions
Represent a range of possible values or dates and combine up to 4 weighted components when using mixture controls.
Community comparison
Read the aggregate forecast alongside individual reasoning and disagreement.
Distinct scoring measures
Separate the fixed-reference Baseline score from the Peer score comparing other forecasts on the question. Account for time-averaging when reviewing results.
Tournaments and AI programs
Explore question sets with their own scoring and rules, plus a separate automated-forecaster tournament track.
Web and API access
The platform documents web and API access for its forecasting ecosystem. Professional service paths have their own terms.

Workflows with Metaculus

Form an initial probability

Read the resolution criteria, record your estimate and identify the evidence that would change it.

Revise a distribution

Update the uncertainty range as evidence changes rather than retaining only a point estimate.

Review a forecasting record

Compare many resolved questions using the chosen score definition and the time each forecast remained in force.

Pricing and total cost

Use the community platform for public forecasting and research. A private instance, a commissioned tournament or professional forecasting requires its own service terms or quote; public access does not establish what those services cost.

Getting started

Before you start: Scores do not represent investment returns.

Choose a question and read its resolution criteria.
Create an account to submit a forecast.
Enter probabilities or a numerical/date distribution.
For multiple choice, check that the total is 100%.
Update the estimate when evidence changes.
Review the resolved score and reasoning.

A forecasting score is not a return on invested money.

Read the forecasting tutorial

Useful links

What to check

Scores do not represent investment returns.
Baseline and Peer scores make different comparisons and should not be mixed.
Time-averaging includes earlier forecasts; a last-minute revision does not erase them.
Tournament prizes depend on the rules of the specific tournament.

A closer look

Ratings & reviews

Editorial rating

4.0/ 5

0 user reviews

PredictionToolkit editorial

Our take on Metaculus

Editorial rating: 4/5. Explicit probabilities, distributions, resolution criteria and scoring support rigorous forecasting practice. Community scores and professional services should remain separate from investment returns and executable market prices.

Based on documented features, setup and access terms. How we rate