How Football Stats Can Beat the Bookies
When most people think about predicting football matches, they focus on goals. After all, goals decide games. But what if goals are actually one of the least useful pieces of information when you’re trying to work out whether a match will go over or under 2.5 goals?
That’s exactly what a study from the London School of Economics found – and the results are surprising, especially for anyone who bets on football.
In this post, we’ll break the research down into simple terms and explain how certain match stats can give you a genuine edge in the over/under goals market.
You can read the actual paper on this link – But I’ve summarised it all for you in this post.
The Problem With Using Goals to Predict Goals
Goals look straightforward, but they’re incredibly noisy. A team can dominate a match, create ten good chances, but still score none. Another team can have one shot all match and sneak a winner.
Because so much luck is involved in actually scoring, goals alone don’t tell you very much about a team’s true attacking strength.
The study found that relying on goals to predict future goals is so unreliable that using goals as the main input produced a consistent loss across every league examined.
A Better Way: Look at What Creates Goals
Instead of looking at goals, the researcher built a rating system called Generalised Attacking Performance (GAP).
This measures how strong a team is going forward and defensively based on the things that lead to goals, such as:
- Shots
- Shots on target
- Corners
These are much more stable from match to match. They reflect how much pressure a team creates. Over thousands of matches, they turn out to be far better indicators of how likely a match is to go over or under 2.5 goals.
How the Model Worked
The model tracked:
- Each team’s attacking and defensive strength at home
- Each team’s attacking and defensive strength away
After every match, the ratings were adjusted depending on whether a team performed better or worse than expected in the attacking stats.
The model then used these ratings to estimate the probability that a match would finish with three or more goals.
If the probability was higher than the bookies’ implied probability, this counted as a value bet.
The Big Question: Can This Actually Make Money?
The researcher tested two simple strategies:
- Level stakes – bet 1 unit whenever the odds offer value.
- Kelly staking – increase the stake when the value is bigger.
He repeated these tests across 12 years, 10 European leagues, and over 68,000 matches.
The Results Were Striking
When using shots, shots and corners, or shots on target and corners:
- The strategy made a consistent long-term profit
- The profit averaged around 0.8% per bet
- This held across tens of thousands of bets
Meanwhile, using goals alone produced a loss in every league.
The conclusion is simple:
Bookmakers price goals well, but they do not price underlying performance as accurately.
Why Only the Best Odds Matter
There is one important catch.
The strategy only made money when using the best odds available across all bookmakers. When using average odds, the profit disappeared and became a loss.
In other words:
- The edge exists
- But only if you consistently take the best price on the market
If you always settle for average odds, the overround (bookies’ margin) wipes the advantage out.
Could the Results Be Just Luck?
To test this, the researcher simulated up to one million random betting sequences.
The model’s performance almost never occurred through chance alone.
This confirms the strategy wasn’t simply getting lucky – it was genuinely identifying value.
Why This Matters for Bettors
This research shows that:
- Goals are too random to rely on
- Stats like shots and corners tell you far more about a team’s attacking quality
- Models built on these stats can outperform bookmaker pricing
- But only if you shop around and take the best available odds
It also hints at a wider truth: the betting markets are not perfectly efficient, especially in the football totals market.
Has the Edge Reduced Over Time?
Interestingly, the profit curve flattens a little in more recent years.
A plausible explanation is that bookmakers now incorporate more advanced performance data into their models.
The edge may still be there – but perhaps smaller, or requiring more refined modelling.
Final Thoughts
This study shows that if you want to predict goals, don’t look at goals.
Look at what causes goals.
Shots and corners offer a more reliable picture of a team’s attacking threat, and when used correctly, they can give you a measurable advantage in the over/under 2.5 goals market.
The key lessons are:
- Use performance metrics, not just results
- Bet only when the odds give you value
- Take the best price available
- And remember: small edges add up over thousands of bets