RSO-F Design correctly called all 14 chess-ranking outcomes in the seven-day portion of ForecastBench’s 27 September 2026 question set. The questions resolved on 4 October, covering nine players who held or improved their ranking and five whose ranking fell.
The five declines were particularly clear in the model’s forecasts. RSO-F Design assigned roughly 79% to 85% probability to those outcomes. The corresponding median probabilities across the 108 eligible forecast entries ranged from 39% to 50%.
| Player | Resolved outcome | RSO-F Design: probability of that outcome | Median: probability of that outcome |
|---|---|---|---|
| Carissa Yip | Held or improved | 99.50% | 78.00% |
| Vincent Keymer | Declined | 85.05% | 50.00% |
| Amin Tabatabaei | Declined | 84.22% | 46.00% |
| Tan Zhongyi | Held or improved | 83.10% | 85.62% |
| Fabiano Caruana | Held or improved | 87.62% | 84.47% |
| Wei Yi | Held or improved | 99.50% | 84.00% |
| Aleksandra Goryachkina | Held or improved | 95.48% | 94.00% |
| Nodirbek Abdusattorov | Held or improved | 99.50% | 88.05% |
| Koneru Humpy | Held or improved | 80.58% | 83.00% |
| Bibisara Assaubayeva | Declined | 81.68% | 47.52% |
| Vaishali Rameshbabu | Declined | 85.14% | 50.00% |
| Anna Muzychuk | Held or improved | 86.13% | 85.00% |
| R Praggnanandhaa | Held or improved | 97.55% | 84.50% |
| Ju Wenjun | Declined | 79.46% | 39.00% |
Probabilities are expressed for the outcome that occurred. A decline probability is 1 minus the submitted probability of holding or improving the ranking. Median uses all 108 eligible entries. Source: ForecastBench forecasts and resolutions; calculations by RysoAI, CC BY-SA 4.0. CC BY-SA 4.0
Five ranking declines
The players were Vincent Keymer, Amin Tabatabaei, Bibisara Assaubayeva, Vaishali Rameshbabu and Ju Wenjun. Each finished with a lower ranking in the benchmark’s resolved data.
Checking all fourteen
The other nine questions concerned Carissa Yip, Tan Zhongyi, Fabiano Caruana, Wei Yi, Aleksandra Goryachkina, Nodirbek Abdusattorov, Koneru Humpy, Anna Muzychuk and R Praggnanandhaa. RSO-F Design favoured a held or improved ranking for each, matching the recorded outcomes.
Across this complete fourteen-question subset, its mean Brier score was 0.0178. Using the median submitted probability for each question gave 0.1179. Lower is better. These are our calculations on the public forecasts and resolutions, not separate official leaderboard scores.
From the archive. The date refers to the period described.