RSO-F-ST placed joint 35th among 463 entries on ForecastBench’s tournament leaderboard on 19 October 2026. Counting each named team once by its best entry, RysoAI placed joint tenth.
The model scored 67.0 on the Overall Brier Index across 567 resolved questions. Its score placed it ahead of the listed Claude Opus 5 and Claude Fable 5 configurations, and Google Cloud’s NanoForecaster entries.
| Measure | Brier Index | 95% interval | Questions |
|---|---|---|---|
| Dataset | 62.9 | [61.6, 64.3] | 485 |
| Market | 71.7 | [70.5, 72.9] | 82 |
| Overall | 67.0 | [66.1, 68.0] | 567 |
Joint entry rank: 35 of 463. Joint named-team rank: 10, counting each named team once by its best overall score, including ForecastBench. Source: ForecastBench, CC BY-SA 4.0. CC BY-SA 4.0
From preliminary to tournament results
Our September article covered the early dataset results from our first submission. The October result adds the tournament comparison we said we would publish, bringing dataset and prediction-market performance into the same evaluation.
ForecastBench admits submissions to the tournament leaderboard after 50 days. That waiting period gives more questions time to resolve before an entry joins the comparison. The 19 October result is a dated snapshot; the board continues to change as outcomes arrive.
The scores behind the ranking
RSO-F-ST scored 62.9 on the Dataset Brier Index across 485 questions and 71.7 on the Market Brier Index across 82 questions. Its Overall Brier Index was 67.0, with a 95% confidence interval of 66.1 to 68.0.
ForecastBench gives the two categories equal weight in the overall calculation, after adjusting Brier scores for question difficulty. The resulting indices are not percentages of correct answers, and the overall index is not a simple average of the two displayed category indices.
From the archive. The date refers to the period described.