Sports Journalists Asked Microsoft’s Copilot to Predict World Cup Matches, and the Results May Surprise You
TL;DR
USA Today asked Microsoft Copilot to predict four World Cup matches: Spain-Cabo Verde 3-0, Belgium-Egypt 2-1, Uruguay-Saudi Arabia 2-1, and Iran-New Zealand 1-0. All four matches ended in draws instead. Copilot appears not to have treated a draw as a serious possibility. The clearest miss was Spain-Cabo Verde: Copilot expected Spanish dominance, but goalkeeper Josimar ‘Vozinha’ Dias helped Cabo Verde hold Spain to 0-0.
Nauti's Take
This is a useful reality check for anyone trying to turn every chatbot into a crystal ball. Copilot did not just miss the scorelines; it missed the shape of the outcomes: four draws, zero hits.
As an editorial stunt, that is entertaining. For betting, scouting, or business decisions, it is weak ground.
LLMs can write convincing sports narratives, but that does not mean they model uncertainty better than a human with the table, context, and judgment.
Briefingshow
This is not just about a few bad picks. It highlights a structural gap: language models can sound analytical without being strong forecasting systems. If their output mainly recombines media hype, team reputation, and plausible narratives, confidence starts to masquerade as accuracy.
Sources