Post by Matillion

One of our team, built a football prediction model in 2021. It took months. Last week, he rebuilt it for the 2026 World Cup with Maia. It took an afternoon. The 2021 version was hand-built in legacy ETL: dozens of transformation jobs, no Python, every piece of logic written by hand. The 2026 version was a conversation. Maia recommended the approach, wrote the Python, built the orchestration, ran 10,000 Monte Carlo simulations on Snowflake, and wrote up the results. The best moment wasn't the speed. Partway through, he noticed every predicted score was coming back 1-0 or 1-1. He asked Maia why. It diagnosed its own scoring method as too conservative, proposed a better statistical model, and rebuilt it on the spot. That's the shift. The work that used to eat a data team's months becomes the work of an afternoon – and the engineer spends their time asking better questions instead of hand-coding logic. (For the record, the model picks Spain to beat England 1-0 in the final.) Read the full blog -> https://bit.ly/4ebsTFW