Hey everyone,
Wanted to share something we’ve been building in the Lab. MatchGraph is a football analytics platform that turns regular 2D match footage into structured, queryable data.
A Bit of Background
At 44, I’m still playing full competitive 11-a-side football. I came back to the game after 16 years of retirement because with 3 boys, football has become life (albeit my eldest has retired in favour of downhill mountain biking). Some weeks I spend every day at Wimborne Town Football Club. One of my sons in particular lives for his football. He plays for WTFC U13s, who are JPL National Champions, and he’s navigating his way from grassroots through the development pathway.
That journey has brought me in touch with many people in the system, none more so than Leon Best, former Premier League striker and Republic of Ireland international, who himself has 3 sons in the AFC Bournemouth professional academy. Together, we’ve been exploring how bleeding-edge AI and computer vision technologies, brought into real product experiences, can benefit our children’s development and perhaps the system as a whole.
That exploration is what led to MatchGraph.
The Problem
The tools for recording grassroots and academy football have never been better. Veo, Hudl, phones on tripods. But recording a match and actually understanding it are two very different things. Most clubs outside the professional tier capture video that sits on a hard drive. There’s no scalable way to extract insight from it.
What MatchGraph Does
We built a processing pipeline that takes 2D football video and runs it through deep learning models for:
- Player detection and tracking to identify and follow individuals across frames
- Ball tracking for continuous position data through the match
- Event recognition including passes, shots, tackles, and set pieces
- Positional data generation per player, per team, over time
The output is what we call the MatchGraph, a time-series graph that captures movement, momentum, key moments, and the micro-decisions that shape a game. Think of it as a structured, queryable representation of an entire match.
Natural Language on Top
On top of the MatchGraph, we’ve layered a natural language interface. Instead of needing to understand data science or build custom queries, you can just ask questions like:
- “Show me all successful through balls by midfielders in the second half.”
- “Compare pressing intensity between the first and second halves.”
- “How did our left back’s positioning change after the substitution?”
The goal is to make advanced analytics accessible to people who understand football but aren’t analysts.
Why This Matters
In grassroots and academy football, talent is often judged by visibility rather than potential. Players outside top-tier academies, late developers, or those in regions with limited scouting coverage get overlooked. MatchGraph is designed to change that by giving every level of the game access to the kind of insight that was previously reserved for professional clubs.
For players, meaningful feedback to improve and showcase their skills.
For parents, visibility into their child’s development.
For coaches, advanced analytics without needing a dedicated analyst.
Where We’re Headed
We’re scaling across the UK, USA, and Gulf regions, building what we see as a new infrastructure layer for football development. The vision is broader than analytics. It’s about democratising opportunity through data, ensuring talent doesn’t go unseen.
More detail on the project page: beach.io/labs/matchgraph
Would love to hear your thoughts. Particularly interested in:
- Anyone working in grassroots or academy football who’d want to test this
- Ideas on what questions you’d want to ask of match data
- Thoughts on how natural language interfaces change who can use analytics
Let’s discuss!