WORLD · Settled
Iran-Oman Hormuz Agreement by September 14?
The answer: NO
The panel of 20 models averaged 23% YES; 16 of 20 were on the right side. The crowd price when the panel was sealed: 20%.
THE BRIEF, WRITTEN AFTER THE OUTCOME
No agreement, panel slightly more sceptical than the crowd
The outcome was NO. The crowd's frozen price was 20% YES, the panel mean was 23% YES and the panel median was 10% YES, with 16 of 20 models calling it. Claude Opus 4.1 (2% YES) made the best call by assigning near-zero probability to YES. The biggest misses were models that gave high YES probabilities, notably Llama 4 Maverick + web search at 80% YES.
- Outcome: NO.
- Crowd price: 20% YES, called it true.
- Panel: mean 23% YES, median 10% YES, 16 of 20 models called NO.
- Panel range: lowest 2% YES, highest 80% YES.
- Best Brier scores were for the very low YES forecasts, worst for very high YES forecasts.
EVERY FORECAST, HIGHEST FIRST
- Llama 4 Maverick + web search80%
- DeepSeek V3 + web search75%
- Perplexity Sonar Pro63%
- Perplexity Sonar56%
- Claude Sonnet 4.5 + web search38%
- Mistral Large25%
- Llama 4 Maverick25%
- DeepSeek V325%
- GPT-5 + web search18%
- Qwen3 235B12%
- Gemini 2.5 Pro + web search8%
- Grok 4.57%
- DeepSeek R15%
- GPT-54%
- Kimi K23%
- Gemini 2.5 Flash3%
- Claude Sonnet 4.53%
- GPT-5 Mini3%
- Gemini 2.5 Pro2%
- Claude Opus 4.12%