NeuPortal runs a public forecasting experiment: every forecast is sealed before the event, timestamped into Bitcoin with OpenTimestamps, and scored against the Polymarket price at the moment of the call.

Resolves YES if, over all non-weather forecasts that resolve between 9 Nov 2026 00:00 UTC and 31 Dec 2026 23:59 UTC, our mean Brier score is lower than the market's. Otherwise NO. A tie resolves NO.

Included: crypto, sports, markets, politics and tech tracks. Weather is excluded because it is most of the record and would decide the question on its own.

Where we stand now (lower is better):

- Non-weather, all time: ours 0.1298, market 0.1233 (920 forecasts)

- September 0.1559 vs 0.1517, October 0.1111 vs 0.1077, November so far 0.1387 vs 0.1210

- All tracks: ours 0.1941, market 0.1606. The market is ahead.

Anyone can check it: each forecast file holds our probability, the market price at the call and the Polymarket market ID.

Data: https://github.com/neuportalai-ui/experiment-proofs

Archived copy (CERN Zenodo): https://doi.org/10.5281/zenodo.24051884

At resolution the list of included forecasts and both scores will be published in the repo. If the data stops being published, the market resolves N/A.

The creator will not trade in this market.