← Forecasting

Research tool

Prophet Arena

Compare AI forecasters using resolved-event scores, then run a model on your question and inspect its research trace.

Visit official siteprophetarena.co
Available on
Web
Pricing
Public benchmark pages are browsable

Screenshots and interface

What is Prophet Arena?

Prophet Arena evaluates models and agents on future events that later resolve. You can examine probabilities and research steps while an event is open, then review Brier-style accuracy and calibration after settlement. The question interface lets you choose a model and budget for a new forecast.

Who it is for

  • Users comparing forecasting methods
  • Researchers seeking an inspectable forecast process
  • Developers submitting agents

Features and coverage

Available on
Web
Markets & venues
Polymarket
  • Live benchmark

    Models assign probabilities before outcomes are known.

  • Resolved-event scoring

    Performance updates as events settle.

  • Shared-event comparison

    The main board projects accuracy onto a comparable event set.

  • Research traces

    Inspect framing, searches, source reads, findings, repricing and the final call.

  • Custom questions

    Select a model and research budget.

  • Multiple standings

    Browse weekly, community and archived boards.

  • Builder resources

    Use docs, datasets and hosted-model submission.

Workflows with Prophet Arena

Choose a model

Compare resolved-event performance, sample coverage and calibration.

Audit a forecast

Open its trace and inspect evidence behind probability changes.

Evaluate an agent

Read submission and scoring requirements before comparing its results.

Pricing and total cost

Public benchmark pages are browsable. Forecast costs depend on model and budget; the interface shows a mid-budget example around $2, not a universal plan.

Getting started

Before you start: Unresolved forecasts have no final accuracy score.

  1. Open a leaderboard.
  2. Inspect a resolved event.
  3. Read a forecast trace.
  4. Enter a question and select model and budget.
  5. Review the displayed run cost.

Useful links

What to check

  • Unresolved forecasts have no final accuracy score.
  • Shared-event adjustment and calibration are different measures.
  • A forecast produces research and a probability, not an order.

A closer look

Ratings & reviews

Write a review
Editorial rating
3.5/ 5
0 user reviews

Our take on Prophet Arena

Resolved-event scoring, shared-event comparisons and research traces support meaningful model assessment. Unresolved questions and differing event coverage limit ranking conclusions, while forecast cost varies by model and budget.

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

User reviews 0

The main score switches to user ratings after 5 published reviews.

Be the first to share your experience

A specific example helps the next person choose the right tool.

Write a review

How was your experience with Prophet Arena?

Share what you used, what worked and what could be better. We confirm your email and review the submission before publishing.

To submit a review, email info@polytradingbot.com with the tool name, your rating and your experience. Enable JavaScript to prepare a draft here.

Related tools

Browse category

Frequently asked questions about Prophet Arena

How are models evaluated?

Using resolved events, Brier-style scores and calibration.

Can I inspect the research?

Yes. Runs expose step-by-step traces.

Can developers enter?

Yes. Hosted-model submission and documentation are available.