Almost nobody does. Every price carries a fee of around 6–8%, and across 113,026 matches over ten seasons no simple rule made a profit. The few who win consistently tend to have their accounts restricted by the bookmaker.
Read in
The fee you cannot see
A bookmaker does not bet against you: it charges a fee hidden inside every price. Add up the probabilities implied by a match’s odds and you get more than 100%. That excess, around 6–8% on the 1X2 market, is what the bookmaker keeps. You can work it out for any match with the fee calculator.
To make a living from betting, beating the average punter is not enough. You have to beat the bookmaker by enough to pay that fee and have something left. And bookmakers set their prices with a lot of information.
What ten seasons of matches say
We tested eight ways of betting on 113,026 matches from 2016 to today, with real closing odds. None made a profit.
Return of eight betting strategies
Strategy
Bets
Wins
Result per €100
The favourite, at the best price of ~17 bookmakers
113,026
50%
-€0.29 [-€0.93, €0.30]
Always the favourite
113,026
50%
-€4.52 [-€5.07, -€3.91]
Always the home team
113,026
44%
-€5.96 [-€6.64, -€5.27]
Always the draw
113,026
26%
-€6.56 [-€7.57, -€5.68]
At random
113,026
33%
-€6.91 [-€7.83, -€6.02]
“Value bets” from an Elo model
36,671
30%
-€8.84 [-€10.67, -€7.08]
Always the away team
113,026
30%
-€9.70 [-€10.65, -€8.88]
Always the underdog
113,026
22%
-€10.42 [-€11.39, -€9.39]
One unit per bet, at the reference bookmaker’s closing price unless stated. In technical mode, 95% bootstrap interval. Data: Football-Data.co.uk and ClubElo, via A. Gábor (2026).
The only one close to zero is backing the favourite while always taking the best price among about 17 bookmakers. Even that does not win. And in real life nobody holds accounts with 17 bookmakers and gets the best price on every match. In the full analysis we also tried machine learning, with the same result.
What if I really am good?
How often you must win not to lose, and how many bets it would take to tell skill from luck.
You must win · just to break even
50.0%
True chance of the result · without the fee
46.9%
Bets to show it is not luck
4,265 · ≈ 8.2 years at 10 bets a week
Even if you really were better than the bookmaker, it would take about 4,265 bets before your results proved it. Until then, a good run and real skill look the same. That is why so many people believe they have a method: chance produces runs in plenty.With true ROI r and odds o, p = (1 + r)/o and the standard deviation of the return per bet is o·√(p(1−p)) = 1.000. For the 95% interval of the mean ROI to exclude zero, n ≥ (1.96·σ/r)². Break-even: 1/o = 50.0%.
Why so many people think they can
Because chance manufactures winners. When thousands of people bet, some will string together good months with no skill at all, and they are the ones who tell their story. Those who lose post nothing. In the betting simulator you can watch those runs appear among bettors who pick at random.
Winning now and then does not prove a method. And if you are betting to win back what you lost, it is worth stopping to look at it calmly.
TechnicalWhere these figures come from
Margin (overround) = Σ 1/odds − 1, median 6.7% at the reference bookmaker and 1.01% taking the best price in the market. Each strategy’s return is the mean profit per unit staked, with a percentile bootstrap interval. The number of bets needed for the 95% interval of ROI to exclude zero is n ≈ (1.96 · σ / r)², where σ is the standard deviation of one bet’s return. The code is in analysis/.
Related questions
Do accumulators pay more?
They pay higher odds, but the fee multiplies with every match. With a typical fee of 6.7%, a 3-leg accumulator loses 18% of the stake on average, and a 5-leg one 28%. That is why bookmakers push them so hard.
Do tipsters work?
Telling whether someone beats the odds takes thousands of bets recorded before kick-off, not screenshots of winners. Among thousands of tipsters, some string together spectacular runs by pure chance, and those are the ones you see advertised.
What about sure bets or arbitrage?
They exist: sometimes the prices of several bookmakers combined guarantee a small profit. They are increasingly rare, vanish within minutes and require accounts with many bookmakers. Anyone who exploits them systematically tends to end up with restricted accounts.
Why do bookmakers restrict winners?
Because their business is the fee in every price, not risk. An account that wins consistently costs them money, and bookmakers can cut the maximum stake they accept from it. Losing accounts, on the other hand, have no limit.
Can AI or machine learning win?
We tried, with 235,796 matches and machine learning models validated season by season. The models get close to the accuracy of the odds but do not beat it, and none made money reliably.