Every dataset has a heartbeat, though most of us never hear it.
We open a file, run a few commands, and chase the output.
But if you pause before the code — before the graphs and models — you can feel a kind of pulse: patterns drifting and reforming, like players warming up on an empty pitch.
That sense of movement is what first drew me to soccer analytics.
In 2016, when Leicester City won the Premier League at 5000-to-1 odds, the data community called it a miracle.
Yet once we examined the numbers — expected goals, pass recoveries, pressing intensity — the miracle began to look like mathematics.
Leicester hadn’t defied probability; they’d changed the game.
They played for moments, not for possession, and our models simply hadn’t learned to see that yet.
That realization shaped my book, Soccer Analytics with Machine Learning.
It’s not really a book about sport; it’s a book about perception.
How to listen to data before forcing it to speak.
How to find rhythm where others see noise.
In analytics, just like in soccer, the field matters.
A team can’t play beautifully without boundaries; an analyst can’t explore freely without structure.
Clean environments — notebooks, libraries, reproducible workflows — are our version of a well-kept pitch.
Once the surface is smooth, creativity takes over.
Exploration comes next.
Before a coach sets tactics, they study film; before we build models, we explore.
EDA is our slow walk around the field — noticing what’s missing, what’s surprising, what’s quietly consistent.
The process teaches humility: before prediction, there must be attention.
Then comes design — what in soccer would be tactics, and in analytics becomes feature engineering.
We translate intuition into numbers: rolling averages for momentum, venue effects for context, expected goals for quality.
Each feature is a belief made testable.
Finance has its own versions: moving volatilities, regime effects, expected returns.
Different games, same logic.
And finally, the model learns to play.
A neural network moves information the way a team moves the ball — through layers of cooperation, weighted chemistry, and adaptation.
Sometimes it finds signals we never expected, like the one that once told me that lost possession predicts defeat more than shots or goals.
That’s when I realized: deep learning isn’t about making machines human.
It’s about helping humans see differently.
We analyze because we care, and we model because we wonder what else the data might reveal if we listen a little longer.
When we stop demanding answers and start hearing patterns, the data begins to move again.
And for a moment — brief, clear, and beautiful — it almost sounds like thought.
A rough and un-edited early release of our book is already out now! Check it out here.
