World Cup AI predictor now lets users ask daft what-ifs
Spoiler: It doesn't end well for Team Register
By The Register
An AI-powered World Cup predictor now lets football fans type their own “what if” scenarios and see how the tournament could change.
Football fans can now test unusual World Cup scenarios using an AI-powered predictor that turns plain English questions into tournament simulations.
The tool, built by Luzmo, has been updated for the 2026 FIFA World Cup after the company previously created an AI Octopus predictor for Euro 2024.
Instead of only showing a fixed prediction, the new version allows users to type a scenario into a prompt box and see how the model thinks the tournament might change.
That could include sensible football questions, such as what happens if a key player is injured, a team receives a red card, or a heatwave affects a match.
It can also handle more playful questions, including unlikely rule changes or deliberately unrealistic tournament conditions.
According to The Register, the model uses data including squad quality, player information, heat and altitude factors, injury data and other inputs. It then runs simulations to estimate win, draw and loss probabilities.
The system uses a Monte Carlo simulation approach, with scorelines derived from thousands of match runs.
For readers, the interesting part is not just who the model thinks will win. It is how quickly AI tools are moving from static predictions to interactive scenario testing.
That means a user does not need to understand statistics, coding or data modelling to ask a question and get a simulated outcome.
Luzmo chief technology officer and co-founder Haroen Vermylen told The Register that the system can respond to both realistic and daft scenarios. He said the new version was rebuilt in Rust so predictions could run more quickly, with results expected in seconds rather than minutes.
OpenAI models are used to parse user requests and generate summaries, while an agent helps create or adjust scenarios, call the calculation engine and explain the results.
The tool also includes filtering to block harmful or abusive prompts, while still allowing lighter “what if” experiments.
At the time of The Register’s test, the baseline prediction suggested Spain would beat England in the final. Spain was given an 18% chance of lifting the trophy and a 26.8% chance of reaching the final.
Those numbers can change if users feed different scenarios into the model.
The idea highlights one of the more practical uses of consumer-facing AI: making complex modelling easier to explore through normal language.
The same concept could be applied well beyond football, from business planning and weather disruption to logistics, finance and public services.
But there is also a warning. Natural language makes tools easier to use, but it can also create misunderstandings if the prompt is vague or the model interprets the question differently from what the user intended.
For football fans, that means the predictor should be treated as an entertaining simulation rather than a guaranteed forecast.
For businesses, it is a useful reminder that AI tools are becoming more accessible — but still need clear questions, good data and human judgement.