I had a weird idea during the World Cup: what if I asked every major AI — both Chinese and American — to predict the same 14 matches? Then bought real tickets based on the results?
So I ran Kimi, Doubao, and Qianwen (China) against ChatGPT, Gemini, and Claude (US) on 14 real World Cup fixtures.
The result that surprised me
They agreed on 12 out of 14 matches.
Six AI models from two completely different ecosystems, trained on different data in different languages — and they converged on the same prediction 86% of the time.
The one match where they split
Norway vs Senegal.
- 🇨🇳 All 3 Chinese AI: Norway wins
- 🇺🇸 All 3 American AI: Draw
Not a single crossover. Every Chinese model picked one result, every American model picked another — like a clean cultural fault line running through the data.
I have no idea if it means anything. But I bought two real lottery tickets (¥2 each) — one following the Chinese consensus, one following the American consensus.
Results drop June 24. I'll post the follow-up here.
Why I did this
I'm an ordinary guy in China running a public experiment: can AI actually help a regular person build side income? I'm documenting everything — wins, failures, and the weird stuff in between.
This is one of the weird ones.
Full breakdown with all 14 predictions: ordinarymantrying.com (https://ordinarymantrying.com/6-ai-models-predict-world-cup-china-vs-us/)
Part 2 (results + analysis) coming June 24.
Top comments (1)
Update coming June 24 when results drop — will edit this post with who got it right.