Binance co-founder Changpeng Zhao (CZ) recently expressed skepticism about the efficacy of shared AI trading strategies after observing DeepSeek's dominance in live cryptocurrency trading competitions. On October 20, 2025, CZ questioned on X how AI models like DeepSeek could outperform the market when multiple traders use similar algorithms simultaneously, stating, "I thought trading strategies work best if you have your own unique strategy that is better than others, AND no one else has it. Otherwise, you are just buying and selling at the same time as others."
The discussion was triggered by a trading leaderboard from Alpha Arena by Nof1, where various large language model (LLM)-powered bots competed over 72 hours. DeepSeek Chat V3.1 led the rankings with a total account value of $11,995.57, achieving a 19.96% gain. The bot executed leveraged long positions, including an ETH 15x long worth $19,375 with an unrealized profit of $1,648.53, along with SOL and BNB positions showing profits of $696.63 and $327.30, respectively.
In second place, Claude Sonnet 4.5 posted a total account value of $10,584.58, up 5.84%, with an XRP 8x long valued at $12,335 generating $441.78 in profit and a BTC 20x long worth $67,357 yielding $380.99. At the lower end, GPT-5 saw a 36.82% loss on a total account value of $6,318.04, with short bets in XRP and DOGE posting losses of $378.52 and $169.89, and a small gain from a BTC 10x long.
An advisor from Amber Group shared the chart, noting, "Right now, DeepSeek is way ahead! This feels so much more impactful than the AI memes from last year." The event spurred broader industry debate, with experts like Markus Levin of XYO warning that public LLMs often rely on biased sources, while Eric Croak of Croak Capital likened retail AI use to "algorithm-assisted gambling" due to omitted risk factors.
MIM
LTC
ZK