Core Insights - The AI investment competition concluded with Qwen 3 Max emerging as the winner, followed by DeepSeek, both from China, while GPT-5 ranked last with significant losses [1][4]. Group 1: Competition Overview - The competition involved six top AI models, including Qwen 3 Max, DeepSeek V3.1, and GPT-5, trading cryptocurrency perpetual contracts on the Hyperliquid platform [3][4]. - Each AI model was given $10,000 to trade various cryptocurrencies, with performance measured by risk-adjusted returns [4][5]. Group 2: Performance Metrics - Qwen 3 Max achieved an account value of $12,232 with a return of +22.32%, completing 43 trades with a win rate of 30.2% and a Sharpe ratio of 0.273 [5][6]. - DeepSeek V3.1 followed with an account value of $10,489 and a return of +4.89%, having the highest Sharpe ratio of 0.359, indicating strong risk management [6][7]. - Other models, including Claude Sonnet 4.5, Grok 4, Gemini 2.5 Pro, and GPT-5, experienced significant losses, with GPT-5 suffering a -62.66% return [5][7]. Group 3: Trading Strategies - Qwen 3 Max employed a straightforward strategy, focusing solely on BTC with high leverage, while DeepSeek utilized a more analytical approach with independent logic for each asset [8][10]. - GPT-5 maintained all positions despite substantial losses, reflecting a less adaptive strategy compared to others [10][11]. - Gemini 2.5 Pro adopted a strict short-selling approach, disregarding short-term market fluctuations [10][12]. Group 4: Future Implications - The competition highlighted the effectiveness of simple, focused strategies over complex ones, suggesting that less can be more in trading [14][15]. - The organizers aim to use financial markets as a training ground for AI, believing that the dynamic nature of markets presents a greater challenge than fixed-rule environments [12][14].
AI 交易大赛结果出炉：只有中国 AI 赚钱，跟 GPT-5 反着买，别墅靠大海
3 6 Ke·2025-11-04 03:26