Foto: Ascannio / Shutterstock.com
How effective are AI trading bots, and which one is the best? An experiment gave six different artificial intelligences €10,000 each to trade crypto, yielding surprising results. According to expert Jan Scheele, Asia is leading the way, but it will be some time before AIs dominate the market. Read on to find out why.
Six Autonomous AI Trading Bots #
In an experiment by Alpha Arena, six trading bots based on ChatGPT, Gemini, DeepSeek, Grok, Claude, and Qwen were each given €10,000 to trade cryptocurrencies on the decentralized exchange HyperLiquid for one month. The goal was to see which AI would achieve the best results, without any human intervention.
During the experiment, the AIs could only trade bitcoin, ethereum, solana, BNB, dogecoin, and XRP. They had access to a wide range of information sources.
“They combine order book microstructure, liquidity flows, funding rates, on-chain network analyses, and fractal patterns from historical volatility. That provides a kind of real-time ‘thermal scan’ of the market,” deep-tech expert Jan Scheele told Crypto Insiders. “Some systems process more than 50 million price and volume changes per day.”
This gives trading bots access to more information and the ability to trade faster, Scheele explains: “In practice, the magic lies in execution engines that operate in microseconds.” Many financial professionals therefore believe it’s only a matter of time before AI completely takes over the sector, he says:
Notable Results #
The experiment therefore produced interesting and surprising results, states Scheele:
“Notably, the Asian models performed better than the Western ones. DeepSeek and Qwen managed to make a profit, while the rest incurred losses.”
The big winner turned out to be the AI model Qwen, which finished in first place with a final amount of $12,287. Deepseek finished in second place with a small profit, at $10,476.
The other AI models lost significantly during the experiment. “Even a simple ‘buy and hold bitcoin’ strategy performs better,” Scheele states. Claude was left with just $6,740 of the initial amount, Grok with only $5,226, and Gemini with $4,485. ChatGPT performed the worst, finishing with a meager $3,734.
Asia and Open Source Take the Lead #
“Some see this as a sign that open-source models, which learn and adapt faster, handle volatile markets better than closed systems like ChatGPT,” says Scheele.
Asian models also have an advantage over Western AI models, he told Crypto Insiders, because they are trained with more privacy-sensitive data:
“China is partly ahead because its ecosystem is much more centrally organized. Large models get access to datasets on a scale that is simply not possible in Europe due to privacy rules. This yields more accurate predictions for patterns that are difficult to train in the West.”
He thinks, however, that it will be a while before the market is dominated by autonomous AI trading bots. The experiment shows, according to him, ‘also a lot about how complex markets are: unpredictable, emotional, and hardly capturable in code’. Real intelligence isn’t just about data, but also about insight, Scheele concludes:
“The technology is making impressive progress, but the market remains a human game of behavior, trust, and timing.”
He is particularly struck by the speed with which China is innovating. However, that lead is not absolute, he tells Crypto Insiders: “The West should see this as an invitation to organize innovation more broadly, for example through open-source infrastructure and shared research platforms.”