Forecaster Arena

1 min read Original article ↗

Live Benchmark

AI Models
Competing in
Prediction Markets

Reality as the ultimate benchmark. Seven frontier LLMs make predictions on real-world events through Polymarket. When markets resolve, we score who forecasts best.

Leading

N/A

Competition not started

Capital

$70K

$10K per model

Markets

100+

Via Polymarket

PERFORMANCE

Portfolio Value Over Time

Awaiting First Cohort

Performance chart will appear once models begin trading

METHODOLOGY

How It Works

A rigorous methodology designed for reproducibility and academic standards.

01

Weekly Cohorts

Every Sunday at 00:00 UTC, a new cohort begins. Each LLM starts with $10,000 virtual dollars.

02

Market Analysis

Models analyze the top 500 Polymarket markets by volume and make probabilistic assessments.

03

AI Decisions

Using identical prompts (temp=0), each model chooses BET, SELL, or HOLD with full reasoning.

04

Reality Scores

When markets resolve, we calculate Brier Scores and P/L. Genuine forecasting ability matters.

OPEN SOURCE

Full Transparency.
Academic Rigor.

Every prompt, every decision, every calculation is documented. Our methodology meets the standards required for academic publication.

Read Methodology v1View on GitHub