Course · Starts September 1, 2026
Deep Learning & AI in Sport
For data scientists, researchers and analysts in football, basketball and ice hockey who want to move from off-the-shelf models to decision-grade AI systems used inside clubs, research labs and sports-tech companies.
Overview
Across eight weeks you build a complete modern toolkit — Bayesian inference, causal modelling, graph neural networks, reinforcement learning and LLM-driven communication — on real tracking and event data.
What makes this course different: every learner works alongside a team of purpose-built AI agents trained on the course material, academic literature and football context. The agents help you think through problems, find the gap in existing work, sanity-check your modelling, and translate your output into the formats clubs actually use — research papers, scouting tools, coaching reports and live dashboards.
By Twelve Football and Skillcorner. Led by Dr. Pegah Rahimian and Professor David Sumpter, with invited experts from leading clubs and analytics providers. Limited cohort to ensure agent-supported feedback for every learner.
What you'll leave with
- A portfolio-ready project — paper, dashboard, scouting tool or coaching report
- A workflow of AI agents you keep using after the course
- Fluency in the modern football AI stack: Bayesian, causal, graph-based, RL and LLM
- Network access to Twelve Football, Skillcorner, Uppsala University and invited club and tech-company experts
Curriculum
Module 1
Week 1 — Tracking & Event Data Foundations
From raw frames to model-ready features. Team shapes, defensive lines and formations — the geometry under every modern analysis.
Module 2
Week 2 — Bayesian Inference for Football
Think under uncertainty. Separate player skill from team effects, judge prospects on small samples and update beliefs as the evidence comes in.
Module 3
Week 3 — Causal Inference for Football Decisions
Move from prediction to what-would-have-happened. Counterfactuals and synthetic controls applied to manager bounces, signings and tactical changes.
Module 4
Week 4 — Representation Learning & Playing Styles
Embeddings that recruitment and tactical analysis can actually use. PCA, autoencoders, t-SNE and UMAP — the math behind modern player profiling.
Module 5
Week 5 — Graph Neural Networks
Pass risk, pass reward, pitch control. The architecture quietly powering today's top club analytics — built from scratch on tracking data.
Module 6
Week 6 — Action Valuation & Reinforcement Learning
xT, EPV, VAEP, plus-minus and RL. Value every action, then critique your own model the way a sporting director would.
Module 7
Week 7 — Interpretability & LLM Communication
SHAP, attention and wordalisations. Turn black-box output into language a coach, scout or sporting director will actually act on.
Module 8
Week 8 — Project Studio & Demo Day
Ship one: research paper, club-ready dashboard, scouting tool or coaching report. Presented live to Pegah, David and invited industry guests.
Course leaders
Dr. Pegah Rahimian
Football Data Scientist, Twelve Football
Dr. Pegah Rahimian is a football data scientist at Twelve Football and a post-doctoral researcher at Uppsala University. She holds a PhD in Football Data Science and has applied AI to derive in-depth analyses from football data, influencing various aspects of the sport.
Professor David Sumpter
Co-founder, Twelve Football
Professor David Sumpter is a professor of applied mathematics at Uppsala University and the author of Soccermatics, a book that explores the application of mathematical modelling to football. He co-founded Twelve Football and has worked with major clubs and national federations worldwide.
Who it's for
- Researchers moving from one-off papers to full modelling pipelines
- Data scientists at clubs and federations going from descriptive to decision-grade
- Engineers at sports-tech providers building tracking-data and LLM products
- Senior analysts transitioning into modern spatiotemporal AI
Testimonials
"Even after years leading data initiatives in professional football, I found the course genuinely valuable. It offered a clear and practical view of how modern AI techniques can unlock additional value from tracking, performance, and scouting data, while remaining grounded in the realities of elite sport. More importantly, it helped me think differently about the next generation of tools and workflows that are beginning to shape decision-making across clubs. I would recommend it to anyone serious about applying AI in a high-performance sporting environment."
Iñigo López Pérez
Head of Data & Analytics, Atlético de Madrid · Chief Data & AI Officer, VidasPrime
"This course was very impressive because it showed how research, algorithms, and football knowledge can be combined and translated into real, proactive insights to plan and prepare games. For me, it helped a lot by giving me a new source of knowledge and curiosity, and it changed the way I approach problems compared with what I knew before taking the course. Regardless of your current role, if you love football and have a curious mind, this course is definitely for you, because it will give you a new way of approaching problems in football."
Tiago Monteiro
First Team Data Scientist, FC Porto
"The course played an important role in shaping the way I approach AI and modeling in football today. Since taking it, I've had the opportunity to improve the Data Science projects at Racing Santander, applying many of the ideas explored in the course to areas such as scouting, player evaluation, tactical analysis, tracking data, and predictive modelling. What I value most is its focus on practical applications rather than theory alone. I would recommend it to any sports professional who wants to understand how modern AI can be translated into real tools and decision-making processes inside a football club."
Manuel Pablo Duran
Senior Data Scientist, Real Racing Club
"This course provided a hands-on approach to soccer-specific analytic concepts — such as soccer maps and EPV — that I had read about but hadn't known how to implement in practice. It also gave me a foundational understanding of graph neural networks and the range of features that can be used to improve model performance. Since joining Houston Dynamo, this has given me a clear roadmap for where I want to take the data analytics department — leveraging tracking data to drive better decision-making across performance, recruitment, and sport science."
Miguel Vidal
Head of Analysis, Houston Dynamo Football Club
Get the latest on football data and AI
Get sharp insight on how the game is changing, practical ideas you can use in your work, and early access to new features and updates from Twelve.