Aster pitted 70 human traders against 30 AI models in a trading competition, with each participant starting with $10,000 and a prize pool of $200,000. The competition ran from December 9th to December 23rd, and was streamed live on Aster's event page. Currently, the top performer is renowned trader Aoying Capital, who tripled his position in a single day. Statistics show that human traders consistently ranked in the top 20, but also occupied the bottom 20 spots. AI traders, on the other hand, were concentrated in the early to mid-tier and later stages of the rankings.
In the AI trader category, Alibaba's Qwen (extremely conservative version) currently leads with a profit of $618, while Ernie 4.5 (aggressive version), Gemini 3 (aggressive version), and various versions of Claude Sonnet 4.5 are the worst-performing models. In terms of trading volume and number of open positions, AI employs a strategy of small, diversified long positions, while humans tend to close positions or concentrate them in a single position. Human PNL standard deviation: 4875.0. AI PNL standard deviation: 224.8. This shows that human performance fluctuations are more than 20 times greater than AI's.
Data shows that human transactions have a higher upper limit and a lower lower limit.
Currently, the top spot on the leaderboard belongs to renowned trader Ao Ying. In fact, on the first day of the event, he was at the bottom of the list, indicating he must have caught a major trend. He has now tripled his initial investment of $10,000, with a floating profit of $23,220.
( From 3,000 to 40 million: The bittersweet story of cryptocurrency trader Ao Ying Capital )
Second place goes to renowned Texas Hold'em player Wesley, who has already profited $20,000. Third place is represented by the English-speaking region @nextfckingthing, currently profiting $17,870. Notably, his trading volume of $6.36 million is the second highest on the list. Familiar names from the Chinese-speaking region, such as Feng Wuxiang and veteran crypto traders, are also in the top ten. However, those ranked 81st and below are all human traders, with even trader Alert's account already at zero. Overall, human traders have a higher ceiling and a lower floor, while AI is concentrated in the early to mid-late stages of trading.
In the AI trader category, Alibaba's Qwen (extremely conservative version) currently leads with a profit of $618, followed by Ernie 4.5's conservative and balanced versions in second and third place, respectively. The remaining models, ranked in order, are DeepSeek 3.1 balanced version, ChatGPT 5 balanced version, DeepSeek 3.1 aggressive version, ChatGPT 4o conservative version, and ChatGPT 4o basic version. Ernie 4.5 aggressive version, Gemini 3 aggressive version, and various versions of Claude Sonnet 4.5 are the worst-performing models.
Human trading performance fluctuates more than 20 times more than that of AI.
Statistics show that human traders average approximately $623,000 in trading volume and hold an average of 0.83 positions. AI traders, on the other hand, average approximately $99,000 in trading volume and hold an average of 5.30 positions. While AI's trading volume is significantly lower than that of humans, its number of positions is considerably larger, indicating that AI employs a strategy of diversifying with small long positions, whereas humans tend to close out positions or concentrate their holdings in a single position.
If we further define the profitable group as those with profits exceeding $1000 and the losing group as those with losses exceeding $1000, the average number of trades in the losing group (187 trades) is significantly higher than that in the profitable group (132 trades). This may suggest the risk of long mistakes. Both traders who made large profits and those who suffered large losses had significantly higher than average number of trades. This again confirms the previous observation: high activity is positively correlated with high volatility. When we further calculate the standard deviation:
- Human PNL standard deviation: 4875.0 (extremely volatile)
- AI PNL Standard Deviation: 224.8 (Very Minimal Fluctuation)
This shows that human performance fluctuates more than 20 times more than that of AI. This perfectly aligns with the human tendency to take high risks in exchange for high returns (or suffer high losses), while AI strictly adheres to risk control and pursues stable, low-volatility strategies. As of press time, the human team's trading profits exceeded $60,000, far surpassing the AI team's $741.
This article, "Can AI Really Replace Human Traders? Real-Time Trading Competition Data: Humans Temporarily Have the Upper Hand, with a 20-Fold Performance Fluctuation Difference," first appeared on ABMedia, a ABMedia .
