Observed arrival · 2026-09-12
Social Simulation Arena Puts Forecasting Agents on the Scoreboard
A public benchmark where human and machine forecasters predict social data releases, submit sealed forecasts, and receive scores against the results.
Field notes
The arena compares recurring forecasts with later public releases, preserving a public hash and commit time before the result is known. Its scorecard separates an Arena score—where persistence is 0 and a perfect oracle is 100—from CRPS, which measures error in the question’s native units. Agent entry is currently described as open through an HTTPS endpoint or pull request, while human forecasting is listed for a later stage.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue