In this episode, Ari Joury (PhD, particle physics; Founder & CEO of Wangari Global) turns his attention to the 2026 World Cup — and to why the machine learning models built to predict it are confidently wrong in specific, predictable ways. Drawing on his upcoming O’Reilly book Soccer Analytics with Machine Learning, he walks through four failure modes: the distribution shift between club and international football, the small-sample limits of Expected Goals (xG), the form-transfer illusion, and the incentive structures that push analysts to publish flashy numbers over honest ones. He then flips the argument: what actually does predict tournament outcomes, and what does that tell us about where ML earns its keep versus where it just looks like it does? The bigger lesson here is not about soccer — it is about knowing which questions your model can actually answer with the data you have.
Topics covered: Soccer analytics, World Cup prediction, Expected Goals (xG), distribution shift, small-sample statistics, feature engineering, predictive modeling, O’Reilly Media, enterprise AI context.
Wangari is the newsletter and podcast for practitioners and leaders navigating the real work of enterprise AI. New episodes every Friday.
https://wangari.global/contact
Ari’s book (O’Reilly Media, early release out now and officially out around June 25): Soccer Analytics with Machine Learning
