TL;DR
- DeepSeek Chat V3.1 achieved a 19.96% gain in 72 hours in an AI bot competition.
- Changpeng Zhao (CZ) warns that shared AI strategies cannot outperform the market if everyone uses them.
- Experts point out that public LLMs lack real data and do not explain asymmetric risks.
The debate over the effectiveness of AI trading has intensified after Binance co-founder Changpeng Zhao (CZ) publicly questioned how shared algorithms can outperform the market. His comments come in response to a 72-hour trading tournament by Alpha Arena (Nof1), where the AI bot DeepSeek Chat V3.1 showed dominant performance.
CZ was skeptical on X (formerly Twitter) about strategies that many traders use simultaneously. “Saw this a lot in my feed. DeepSeek out performing the rest in AI trading. How does this work? … You are just buying and selling at the same time as others,” he reacted. CZ argued that “AI strategies work if traders have their own plan that is better than others, and no one else has it.”
The tournament that caught CZ’s attention showed DeepSeek leading the leaderboard with a 19.96% return (a total account value of $11,995.57). The bot used leveraged long positions in ETH (15x), SOL, and BNB. In second place, Claude Sonnet 4.5 obtained a 5.84% gain, while at the lower end, GPT-5 registered a 36.82% loss after making short bets on XRP and DOGE.
Experts warn: “Algorithm-assisted gambling”?
Despite the enthusiasm, several experts are tempering expectations about the effectiveness of AI trading for retail investors. Markus Levin, co-founder of XYO, argued that public LLMs, like ChatGPT, often rely on a “small, self-reinforcing pool of sources” (press releases, Reddit threads) and lack real market data, unlike the proprietary AI systems used by large firms.
The risk, according to Levin, is that users treat the responses of these LLMs as investment advice.
Eric Croak, president of Croak Capital, was more direct, calling the retail use of generative AI in crypto markets “algorithm-assisted gambling.” Croak warned that the danger lies in the “[AI’s] inability to explain asymmetric risk in real terms,” omitting key factors such as tax consequences or liquidity concerns. This skepticism highlights the gap between theoretical performance in tournaments and the practical and safe application in the effectiveness of AI trading in the real world.
