In recent weeks, I’ve been working on a framework to better articulate the impact of what analysts often label as “variance.” It feels overly simplistic to call it “luck,” because match outcomes are shaped by multiple contributing elements. Still, football’s low-scoring structure arguably makes it the most unpredictable major team sport.
One lapse in concentration or one unstoppable strike can decide a game, and even the weaker side can defend resolutely or seize a single defining moment.
But just how powerful is standard variance in football? In my opinion, it’s not examined nearly enough. Among the many analytical voices I follow, there’s surprisingly little sustained focus on the scale of its influence.
Supporters prefer to believe victories come purely from superiority — sharper coaching, stronger players, smarter tactics. Acknowledging how frequently fortune intervenes is uncomfortable. No one wants to credit randomness for success.
Think about what can realistically be controlled. Can a team dictate how clinical the opposition will be? Can they stop an opposing goalkeeper from producing an extraordinary display? Refereeing decisions remain subjective. Even individual form fluctuates — a striker who scores six in six can suddenly go ten games without a goal. Tactical systems can improve the probability of chances, but when the decisive moment arrives — when foot meets ball — control is limited.
Consider Ange Postecoglou’s second season at Celtic F.C. in 2022/23, when they were dominant. That campaign saw them benefit from positive expected-goals variance of 1.06 xG per match. In practical terms, swings in finishing and goalkeeping at both ends effectively granted them more than one additional expected goal per game. Across 38 matches, that totals just over 40 goals. Celtic finished with a goal difference of +80 — meaning roughly half of that margin can be attributed to favourable variance. Opposition goalkeeping alone contributed 0.79 xG per game in their favour.
Now contrast that with Dundee United F.C., who were relegated that same season, finishing three points behind Ross County F.C. and seven goals worse off in goal difference. Dundee United’s expected-goal variance stood at -0.48 per match — a negative swing of more than 18 expected goals across the season. Had their variance simply balanced out to neutral, survival may well have been within reach.
To understand where measurable process meets unpredictable outcome, we turn to expected goals — the backbone of modern football analytics. xG difference is widely regarded as a better predictor of future results than actual goals scored. That philosophy underpinned the betting models built by Tony Bloom and Matthew Benham before they moved into club ownership.
Bloom now owns Brighton & Hove Albion F.C. and Royale Union Saint-Gilloise, and holds a stake in Heart of Midlothian F.C.. Benham controls Brentford F.C. and previously invested heavily in FC Midtjylland.
Their success reinforces a key idea: while short-term results can swing wildly due to variance, underlying performance metrics — particularly xG differential — provide a far clearer indication of a team’s true level and likely trajectory.
Leave a Reply