Core Sports AI · Co-Founder & CEO
An AI performance intelligence platform for squash
Core Sports AI turns match footage into performance intelligence. Computer vision reads the match — players, ball, shots, positions — and the analytics layer turns that into the kind of read a good coach gives you, at a scale no one can watch by hand.
Why Core exists
Almost every squash match ever played has been analysed by memory. Ours is a sport with enormous tactical depth and almost no analytical infrastructure.
At the top of the game, the analysis that does happen is manual: a coach watching tape, pausing, rewinding, counting by hand. It works, and it does not scale. Below the professional tier it mostly doesn't happen at all — not because players don't want it, but because nobody has the hours.
Core exists because the intelligence that coaches and elite athletes develop over years is, in large part, pattern recognition over a very large number of observations. That is a thing software is good at — provided it is built by people who understand what the patterns actually mean on a court.
I lead technology and product end-to-end — AI architecture, computer vision, analytics, product systems and technical direction.
How it works
Video in, structure out, intelligence on top
Four layers. Each one is only as good as the one beneath it, which is why most of the engineering effort goes into the parts nobody sees.
- 01Footage
A match, filmed however you film it
The input is ordinary match video. No sensors on the player, no instrumented court, no special camera rig — the constraint that makes the problem hard is also what makes the product usable by anyone with a phone on a tripod.
- 02Vision
Players, ball, court, shots
Computer vision recovers the structure of the match: where each player is on court over time, where the ball goes, and what kind of shot was played. Court geometry turns pixel positions into real positions you can reason about in metres rather than pixels.
- 03Structure
From detections to rallies
Individual detections are only useful once they become sequences: rallies, exchanges, recoveries, the shot that preceded the error. This is the temporal layer, and it is where most of the difficulty lives.
- 04Intelligence
The read a good coach gives you
On top of that structure sit the things a player actually wants: court positioning, tactical patterns, performance metrics, and AI-generated insight about what a match revealed — including the patterns that recur across matches rather than within one.
Platform
What the system does
Vision
- Automated match analysis
- Computer vision
- Player tracking
- Ball tracking
- Shot recognition
Analysis
- Tactical pattern analysis
- Court positioning
- Performance metrics
- AI-generated match insights
Preparation
- Opponent scouting
- AI game plans
- Match simulations
Development
- Digital player twins
- Player development tracking
- Shareable performance profiles
Core Sports AI · Digital Twin
A model of a player that keeps learning them
A Digital Twin is an evolving representation of an athlete, built from their match history rather than from a questionnaire.
Every match a player uploads adds to it: where they move, how they recover, which shots they choose from which positions, what they do when a rally gets long, what changes when the score gets tight. The twin is the accumulation of that behaviour — tendencies, strengths, weaknesses, positioning and tactical patterns — held in a form the system can reason over.
That representation makes two things possible. The first is deeper analysis of your own game: not what happened in one match, but what keeps happening across all of them.
The second is scouting. A player can study an opponent they have never faced, and use simulations between the two twins to explore tactical approaches before they walk on court. It is a way of generating and testing hypotheses about a match — a preparation tool, not a prediction of the result.
Scouting
Preparing for someone you've never played
Draws are published days before a tournament. You get a name, maybe a ranking, and if you're lucky a friend who has played them. Everything else is guesswork.
With Core, an athlete can study an opponent through that opponent's own match history: where they hold position, what they do under pressure, which shots they favour from which parts of the court, how their patterns change as a match goes on.
Because both players can be represented as Digital Twins, the system can simulate the interaction between them and surface tactical approaches worth considering — the length to target, the pattern to avoid, the position that tends to put this particular opponent under strain.
Simulations are a way to generate and test tactical hypotheses before a match. They are not predictions of results, and Core does not claim to forecast outcomes.
Why this team
“I spent a decade asking these questions as a player, and I now spend my time building the systems that can ask them at scale.”
Ziad Sakr — Co-Founder & CEO, Core Sports AI
Further reading
How the system actually works
Digital Twins
What Is a Digital Twin for an Athlete?
A Digital Twin is not a 3D model of a player. It is an evolving representation of how someone actually competes, built from their match history — and it is useful precisely because it keeps changing.
5 min readComputer Vision
Teaching AI to Understand a Squash Match
Detecting a player is the easy part. Understanding a rally means recovering structure over time — and that is where most of the real engineering in sports computer vision goes.
6 min readSports Analytics
AI Opponent Scouting in Sports
Scouting used to mean a coach with a notebook and whatever footage they could get hold of. AI changes the economics of it — and changes what a scouting report can honestly claim.
4 min readSports Analytics
How AI Identifies Tactical Patterns in Sports
Finding patterns in sports data is easy. Finding the ones that are real, repeatable and worth changing your game for is the actual problem — and it is mostly a discipline problem, not a modelling one.
4 min readComputer Vision
Player Tracking vs Ball Tracking in Sports Computer Vision
They sound like the same problem and they are not. One is an identity problem, the other is a signal problem — and treating them alike is the most common mistake in sports vision.
5 min readSports Analytics
Beyond the Scoreboard: What AI Can Learn From Sports Video
The score tells you who won. Everything explaining why is in the video — and most of it has never been extracted, because until recently nobody could afford to watch it all.
5 min read
Contact
Building something at the intersection of AI and the real world?
I'm glad to hear from engineers, founders, researchers, coaches and athletes — and from anyone working on AI systems that have to be right rather than merely impressive.