NewsMore in News→
AI Soccer Betting Test Shows Frontier Models Still Struggle With Real-World Forecasting
Key Takeaways
- The KellyBench report tested eight AI models on a simulated Premier League betting season.
- All models lost money on average, with several suffering total bankroll ruin.
- The benchmark highlights the gap between language fluency and robust real-world decision-making.
DE
DT Editorial Team··via arstechnica.com