(Yicai) July 14 -- Artificial intelligence models developed by 12 Chinese tech firms have shown greater accuracy than human fans at predicting matches during a 2026 FIFA World Cup competition, posting an overall accuracy of 66 percent so far.

The World Cup prediction contest, co-hosted by Chinese personal computer and tech giant Lenovo Group and streaming service Migu Video, set the 12 AI models against around 35 million human fans, who gave correct outcomes with 59 percent accuracy after the first 100 matches, according to a report released by the organizers yesterday. The human group led after seven days but has remained second since.

The AI lineup consists of DeepSeek, Qwen, China Mobile's Jiutian, Baidu's Ernie Bot, Tencent Holdings' Hunyuan, Kimi, Zhipu AI, MiniMax, JieYue Star, iFlytek Spark, SenseTime Xiaohuan, and Lenovo's Tianxi AI. They got 788 of their 1,200 win-draw-lose predictions correct after the end of the 100th fixture between Argentina and Switzerland in the quarter-finals.

Hu Yanping, distinguished professor at Shanghai University of Finance and Economics, told Yicai that the AI collective accuracy aligns with his prior projection that it would land between 60 percent and 80 percent. "The World Cup prediction contest serves as a practical testbed for evaluating the reasoning capacity and inherent limitations of AI models.

"Exposing both their strengths and shortcomings provides tangible insights to guide further optimization," Hu noted.

As the World Cup progressed, richer data on team form, squad rotations, group standings, and tactical tendencies gradually emerged, amplifying AI's strength in aggregating and processing massive volumes of information, the report said. By contrast, human judgments are easily skewed by team popularity, personal fandom, and emotional bias.

Unexpected draws and other "upsets" emerged as a consistent blind spot for the AI participants during the first 100 matches, with all 12 models giving a wrong prediction for 11 fixtures that ended in a draw and in four that were won by a lower-ranked team and hence dubbed upsets.

AI's general deficiency in predicting draws stems from its solid capability to gauge relative team strength, paired with its inability to reliably judge whether this strength advantage can translate into goals within the duration of a match, a dynamic worthy of deeper research, Hu pointed out. However, the models' failure to anticipate unforeseen in-game incidents falls within expected performance boundaries, he added.

Historical dominance of traditional powerhouse nations leads AI to produce highly uniform pre-match projections, resulting in a collective error when elite teams suffer unexpected defeats. All 36 forecasts submitted by the AI models for the elimination matches of Germany, the Netherlands, and Brazil turned out wrong.

Predicting correct scores was also a major weakness of the AI models. Out of 1,200 score predictions, only 145 were on the mark for a 12 percent accuracy. Even the top-performing model only accurately predicted the exact score in 17 percent of the time.

In addition, all 12 AI models favored Brazil to claim the title, but the nation was eliminated in the Round of 16, while none picked England, a nation that has made it into the final four.

Editors: Tang Shihua, Martin Kadiev