When xG Meets Fan Bias: A Data Witch’s Confession from East London

When xG Meets Fan Bias

I still remember the first time a Spurs fan told me, ‘They won’t listen — it’s all about heart.’ Not data. Not logic. Just ‘they’ve always been lucky.’

I chuckled. Not because I’m sentimental. Because I’m a Bayesian analyst with a Python script and zero tolerance for myth.

The Math Behind the Magic

Last season, 46 old-school analysts walked into my lab — not to discuss tactics, but to plead for tradition.

They said: ‘Dejan Kulic scored…’ Wait — who? Dejan who? There is no Dejan Kulic.

It was a typo. A glitch in their narrative.

But the xG model didn’t blink.

We ran the numbers: 0.87 expected goals per shot. Their eyes saw 1.2.

The gap? 40%.

I didn’t need faith. I needed p-values < 0.05.

Why Intuition Dies Quietly

Fan bias isn’t emotion — it’s cognitive dissonance wrapped in nostalgia. Spurs fans don’t hate stats; they fear being wronged by curves that don’t bend to their identity. Yet here we are: Bayes doesn’t care if you’re right — it cares if your model fits the data. And it does. Every time you say ‘he’s due,’ I run simulations again — and every time… The numbers win.