OFFBEAT
Spoiler: The Register team fares poorly in the simulation.
The developers behind the AI Octopus Euro 2024 predictor have upgraded their simulator for the 2026 FIFA World Cup, enabling users to submit natural‑language scenarios and view projected tournament outcomes.
“Sensible queries — such as a red card, a key injury, a heat wave, or a change of base camp — yield useful insights, but the model also handles whimsical ones, for example ‘What if the tournament were played under rugby rules?’” noted Haroen Vermylen, CTO and co‑founder of Luzmo.
Using the tool is straightforward: type a scenario into the prompt box and the predictor returns projected results. The underlying data encompasses player‑based squad strength, heat and altitude considerations, injury statistics, and similar factors. A Monte Carlo simulation runs 5,000 match iterations to produce win‑loss‑draw probabilities, from which a predicted score line is derived.
The original Euro 2024 AI Octopus engine was built in TypeScript; for this edition the team switched to Rust. “We adopted Rust to gain speed, especially now that a real‑time component is required,” Vermylen explained to The Register.
“Previously the simulation could take up to five minutes. Our goal is to deliver predictions within two to three seconds of actual runtime.”
OpenAI models interpret the user’s prompt and produce summaries, while an orchestrating agent transforms the scenario, invokes the calculation engine, and returns answers. No data‑science expertise is required to pose questions or comprehend the output.
The system is notably swift, recalculating outcomes for even unconventional prompts — such as pondering the impact of politically charged emissions from a prominent leader. Not every scenario is viable; Vermylen noted that filters block profanity and prevent inputs that could be harmful to specific groups.
A lingering challenge remains the parser’s occasional failure to grasp the intent behind a prompt. Clear phrasing is essential; while natural‑language input offers a friendly alternative to sliders and toggles, it can also lead to misinterpretation.
As the competition unfolds, the dataset will be continuously updated. At present the baseline projection shows Spain defeating England in the final, with an 18 % chance of winning the trophy and a 26.8 % chance of reaching the final match. These probabilities shift when users inject their own scenarios.
For instance, querying “What if the Spanish team eats a bad paella?” reduces Spain’s title odds to 1.5 %, projecting France as the likely champion.
We also tested a scenario swapping the England squad for Register writers; the outcome, predictably, was unfavorable.
When asked about future plans, Vermylen replied, “The Olympics would be appealing… or perhaps Eurovision. We’d love to see the United Kingdom secure a victory.” ®
