Elon Musk once again drew attention to the capabilities of his Grok neural network from xAI, this time in the context of sports predictions. The reason was a post by an X user claiming that back in April, long before the World Cup started, Grok analyzed bookmaker odds, prediction market data, and supercomputer models, concluding that the Spanish national team had the highest probability of winning the trophy. Several months later, the Spanish team indeed lifted the cup.
How Grok "Predicted" Spain's Triumph
The post author presented this case as an example of how the neural network can combine several independent signals and, based on them, determine the most likely outcome long before the event. According to him, this case proves once again that Grok is particularly strong in forecasting and predicting results. Musk, commenting on the post, agreed with this assessment, calling the AI's work "great."
However, not all users share the billionaire's optimism. Some commenters were skeptical of the "prediction." They note that the bot did not so much predict the outcome of the final as it collected an already established consensus: Spain was already considered the favorite with a probability of about 58-60% in prediction markets before the start. From this perspective, it is not about prophecy, but about data search and aggregation, whereas the real skill of forecasting is calibration across a hundred different predictions, not one lucky hit.
There were also those who spoke more harshly, calling the coincidence simple luck and claiming that the bot cannot accurately predict anything. An argument was also made that with proper prompting, the neural network can be steered toward any desired answer, so such "predictions" should not be taken seriously. One commenter stated that they asked the same question before the match and received, in their opinion, a random result.
At the same time, there were those who agreed with Musk. Some users called the bot's answer accurate and supported the idea that combining independent signals yields a more reliable result than guesses from a single source.
Honesty as a Competitive Advantage
Interestingly, Grok's reputation as a "predictor" contrasts with its recent reputation as the "most honest model." In a simulation by the Kradle project with four AI agents that needed to choose one of the rooms (three with food, one with death), an informed Grok 4.20 lied in only 5% of runs and directly warned the others about the deadly room in 92% of cases. This openness also yielded practical results: Grok's honesty ensured the survival rate of other agents at 59% — the best indicator among all participants.
| AI Model | Opponent Survival Rate |
| Grok 4.20 | 59% |
| Gemini 3.1 Pro | 33% |
| Claude Sonnet 4.6 | 31% |
| GPT-5.5 | 24% |
| Claude Fable 5 | 10% |
The worst situation was with the informed Claude Fable 5, whose manipulations left opponents with only a 10% chance of survival.
Expert opinion: The World Cup prediction case is more a demonstration of the power of data aggregation than true foresight. For the crypto industry, where analyzing many disparate signals is a daily necessity, such Grok skills could be much more useful than guessing sports results. However, investors should remember: even the most honest model is not immune to errors, and relying on a single prediction, even an accurate one, is a risky strategy.