How to build a winning football betting strategy
A seven-step method with the maths worked out: value, staking, sample size and losing runs. No strategy removes the risk.
You have a theory. It feels sound. You back three teams on a Saturday, and two win. You feel sharp, like you've found an edge. Then comes the next month. The theory stops working, or you get unlucky with one leg of an accumulator. You give back all the profit, and you are left wondering: was the theory ever any good, or did I just get lucky?
This guide gives you a seven-step method for building a football betting strategy and finding out whether it works. It uses simple arithmetic to strip away the noise and show whether your approach has an edge. It won't promise you profit, because no strategy removes the risk. What it can do is limit the cost of finding out, using clear rules you can test. This guide is for adults, 18+. If you are not sure about your relationship with betting, read our responsible gambling page first.
What "winning" actually means
Most people confuse hit rate with profit. A high percentage of winning bets does not guarantee a profit, and a low percentage does not guarantee a loss. Profit depends on the odds you take.
Consider a simple example: 100 bets of $10 each, at average odds of 2.10. You stake $1,000 in total.
- If 50 bets win, you get $1,050 back. Your profit is $50 and your return on investment (ROI) is 5%.
- If 48 bets win, you get $1,008 back. Your profit is $8 and your ROI is 0.8%.
The difference between those two results is just two winning bets out of 100.
To break even at odds of 2.10, you need a hit rate of 47.6%. Win 40% of your bets at odds of 3.00 and you are in profit. Win 70% at odds of 1.20 and you are losing money. The formula is simple:
ROI = profit / total staked
So a strategy that wins 40% of its bets at high odds can beat one that wins 70% at short odds. Look at the price, not just the outcome.
The bookmaker's margin: what every strategy has to beat
Every price you take includes the bookmaker's margin. It is the hidden cost your strategy must overcome.
Take a hypothetical match priced at 2.10 for the home win, 3.40 for the draw and 3.60 for the away win. To find the implied probabilities, divide 1 by the odds. Add them together and you get 104.81%. The extra 4.81 points are the overround, the bookmaker's margin. Back all three outcomes and you would have to stake $104.81 to get $100 back.
Remove the margin and you get the fair chances and fair odds:
| Outcome | Bookmaker odds | Implied probability | Fair probability | Fair odds |
|---|---|---|---|---|
| Home | 2.10 | 47.62% | 45.4% | 2.20 |
| Draw | 3.40 | 29.41% | 28.1% | 3.56 |
| Away | 3.60 | 27.78% | 26.5% | 3.77 |
Betting at random on such a book returns about $95.41 for each $100 staked over the long run. Being right often is not enough: you must be right by more than the margin. If you think the home win's fair price is 2.20 and the bookmaker offers 2.10, there is no value in it, even if the home side is the likeliest winner. If the bookmaker offers 2.30, there might be.
Value betting and expected value (EV betting)
The core idea of any football betting strategy with a chance of winning is value betting. A value bet is one where the odds pay more than the true chance of the outcome justifies.
The formula for expected value (EV) is:
EV = (probability × odds) − 1
If you put a team's chance at 50% and the odds are 2.20:
EV = (0.50 × 2.20) − 1 = +0.10
That means you expect to make 10 cents for every dollar staked in the long run. At 2.20, the break-even hit rate is 45.45%.
Now look at what happens if your estimate is wrong.
- If the true chance is 45%, EV is −1%. You lose money over time.
- If the true chance is 42%, EV is −7.6%.
Value depends completely on your probability being right, and a small error turns a good bet into a bad one. Value bets also lose often. A value bet is a judgment about the price, not a prediction that the team will win. You can stake $10 on a value bet, lose it, and have made the right decision. Any profit comes from the total of many such bets, never from one.
Where your probabilities come from
You need a source for your estimates. There are three.
Your own judgment. Cheap, but unmeasured. It is prone to bias: you may favour big clubs because they are famous, or lean too hard on recent results. You can't tell how good your judgment is without a strict record.
A model. A statistical model turns data into probabilities and has no favourite team. FootyMath is an open example: a Poisson / Dixon-Coles style goal model that builds team strength from goals and shots on target, adds an Elo rating, then calibrates the output. You can read how the model works.
The market itself. The odds reflect what bettors and bookmakers collectively believe, and prices in big leagues are hard to beat. Taking the margin out of them, as above, gives you a benchmark to compare your own numbers against.
Here is what a model can and can't tell you. In a backtest over 21,272 matches from July 2023 to October 2026 across 18 leagues, FootyMath's model named the right result 50.9% of the time (51.6% in the top European leagues), against 44.3% for always picking the home side. Picks rated 60% or more hit 69.5% of the time. Picks rated 70% or more hit 78.0%. You can follow the live public record of predictions stored before kick-off.
Now the honest catch: these are accuracy figures, not profits. On average, a group of picks that hits 69.5% only makes money at odds above 1.44, and a group that hits 78.0% only above 1.28. Favourites tend to be priced short. At odds of 1.35, a 69.5% hit rate gives an EV of −6.2%. At 1.20, a 78.0% hit rate gives −6.4%.
FootyMath's backtest has no historical odds in it, and a forward test against real prices only began on 22 October 2026. So profitability against bookmakers is not proven. The model also does not see team news, injuries or suspensions, while the market may. Use it as one input, not as a verdict.
Seven steps to build your strategy
Step 1: Write the rule down
You need a rule so clear that someone else could read it and place the same bets.
- Good rule: "Singles on League X, match result only, when my estimate beats the price by at least 5%, odds between 1.80 and 3.50."
- Bad rule: "Back teams in form."
Vague rules cannot be tested. When you win, form was good. When you lose, it wasn't really form. A vague rule always gets rewritten after the fact, and a specific one can't.
Step 2: Narrow the field
Start with one market and a few leagues. Don't try to bet on every match in the world.
There is a trade-off here. Big leagues like the Premier League or La Liga have sharper prices, because more money and public information go into them. Smaller leagues may have softer prices, but the data is less reliable and bookmakers may limit your stakes. Don't assume either is more profitable. Pick one and measure it.
Step 3: Set the entry threshold
Require a margin of safety between your probability and the price, because your estimate has an error of its own. FootyMath's simulator, for example, flags "value" when the model's chance is at least 5% better than the odds pay for.
Choose an odds range too. Longer odds have more variance, so you need many more bets to judge a strategy at 4.00 than at 1.50.
Step 4: Fix your bankroll and staking
Decide how much money you are prepared to lose and how much goes on each bet. Do this before the first bet, not during a bad run. The next section covers the options.
Step 5: Keep a record of every bet
Record the date, league, market, odds taken, your estimate, stake, result and closing odds. Add a brief reason for the bet. A record cuts through selective memory: you will forget the bets you lost, but the spreadsheet won't.
Step 6: Test without money first
Use our free betting simulator to run your rule. It uses real fixtures and, where available, the median bookmaker price, with $100 of virtual money a week. Its Strategy tab splits your ROI by bet type, odds, outcome, league, and with or against the model. At $5 a bet you can place 20 bets a week, so 100 bets take about five weeks.
Step 7: Review on a schedule, not after a bad weekend
Decide in advance after how many bets you will judge the strategy. Change one thing at a time. When you change the rule, start a new count. A single weekend of losses is noise, and so is a single weekend of wins.
Bankroll management and staking plans
Your bankroll is money set aside that you can afford to lose completely, separate from living costs. If it goes, your life should not change.
There are three common staking plans.
Flat staking. You stake the same amount on every bet. On a $1,000 bankroll, 2% means $20 per bet. It is simple, and it makes your results the easiest to read. Ten straight losses cost 20% of the bankroll.
Percentage staking. You stake a fixed percentage of your current bankroll, so stakes shrink as you lose. At 5%, ten straight losses cost 40.1% of the bankroll, against 50% if you staked a flat 5% of the original amount.
Kelly staking. The Kelly formula works out a stake from your edge:
Stake fraction = (b × p − q) / b
Here b is the odds minus 1, p is your estimated probability, and q is 1 − p. At 50% and odds of 2.20, Kelly says to stake 8.33%. Half Kelly is 4.17% and quarter Kelly is 2.08%.
If the true chance is 45%, the formula goes negative at −0.83%, which means no bet at all. Kelly assumes you know p exactly, and nobody does, so full Kelly is too aggressive for an estimate.
| Staking plan | Stake | Ten straight losses cost |
|---|---|---|
| Flat | 2% of the starting bankroll | 20.0% |
| Percentage | 2% of the current bankroll | 18.3% |
| Quarter Kelly | 2.08% of the current bankroll (example above) | 19.0% |
| Half Kelly | 4.17% of the current bankroll (example above) | 34.7% |
| Full Kelly | 8.33% of the current bankroll (example above) | 58.1% |
For most people starting out, flat staking is the sensible default, because it keeps your results easy to read.
Why Martingale and other chase systems fail
The Martingale system is one of the best-known football betting systems: you double your stake after each loss at odds of 2.00, starting at, say, $10.
| Bet | Stake | Total staked |
|---|---|---|
| 1 | $10 | $10 |
| 2 | $20 | $30 |
| 3 | $40 | $70 |
| 4 | $80 | $150 |
| 5 | $160 | $310 |
| 6 | $320 | $630 |
| 7 | $640 | $1,270 |
A win at any point recovers your losses and leaves a profit of $10. All that risk, to win $10.
Six losses in a row at a 50% chance has only a 1.56% probability from any given starting point. But over 100 even-money bets, the chance that such a run appears at least once is 54.6%. A $1,000 bankroll can't pay for bet 7, and a bookmaker's maximum stake may stop you sooner.
Staking cannot turn a negative-EV bet into a positive one. It only rearranges when you lose. The same logic applies to any "zero risk" system.
Arbitrage is different, since it rests on a real price difference between bookmakers. But it isn't risk-free in practice: it needs many accounts and fast execution, and bookmakers can restrict the accounts doing it.
Singles or accumulators
Accumulators (parlays) compound the bookmaker's margin. Picks get less likely to all come in, too. The table assumes each leg pays back 95% of its stake in the long run, and that the picks are independent, with a 69.5% chance each.
| Legs | Long-run payback | Chance all picks win at 69.5% each |
|---|---|---|
| 2 | 90.2% | 48.3% |
| 3 | 85.7% | 33.6% |
| 4 | 81.5% | 23.3% |
| 5 | 77.4% | 16.2% |
| 8 | 66.3% | 5.4% |
Accumulators aren't wrong as entertainment, but they are the hard way to run a strategy. Wins are rare and each one is large, so a few results can swing your numbers, and you need a far larger sample to judge an edge.
How many bets it takes to know
This is where most betting strategies fool their owners. Assume bets at odds of 2.00 with flat stakes. The table shows the typical swing in ROI, and how often the true quality of a strategy stays hidden.
| Bets | Typical swing in ROI (one standard deviation) | Chance a real +3% edge is showing a loss | Chance a real −5% strategy is showing a profit |
|---|---|---|---|
| 50 | ±14.1 points | 36.2% | 31.0% |
| 100 | ±10.0 points | 34.4% | 27.4% |
| 250 | ±6.3 points | 29.5% | 19.6% |
| 500 | ±4.5 points | 23.7% | 12.2% |
| 1,000 | ±3.2 points | 16.3% | 5.3% |
| 2,000 | ±2.2 points | 8.6% | 1.2% |
| 5,000 | ±1.4 points | 1.6% | under 0.1% |
These figures are calculated from the binomial distribution. After 100 bets, a strategy that loses 5% in the long run still shows a profit more than a quarter of the time. So 100 bets tell you very little, and a strategy with no edge can look brilliant for months. Results at longer odds take longer still. There is no magic number: 1,000 bets is a far better checkpoint than 100, and even that can mislead.
Losing runs to expect in advance
You will have losing runs. Expect them, so they don't panic you into changing a sound plan. Here is the chance of at least one run of that length in 500 bets:
| Hit rate | 6 losses in a row | 8 in a row | 10 in a row | 12 in a row |
|---|---|---|---|---|
| 50% | 98.3% | 62.5% | 21.5% | 5.8% |
| 45% | 99.9% | 85.4% | 43.3% | 15.6% |
| 40% | 100% (rounded) | 97.1% | 70.5% | 35.1% |
At a 45% hit rate, a run of ten losses is close to a coin toss over 500 bets, so it isn't by itself a sign the strategy has broken. Plan your stake so such a run can be survived, and decide in advance the drawdown at which you stop and review.
Closing line value and comparing prices
Comparing prices across bookmakers is a free improvement. Taking 2.20 instead of 2.10 on a bet you would place anyway adds 4.8% to every winning return, and it costs nothing in extra risk.
Closing line value (CLV) compares the odds you took with the final price before kick-off. Say you took 2.20 and the match closed at 2.00. You got a price 10% better than the close. Remove a 4.81% overround from that 2.00 and the fair close is about 2.10, so if the closing price was right, the bet's edge was about +5%.
CLV gives a much faster signal than profit, as it doesn't depend on results. But it is an indicator, not proof. Record closing odds in your log.
Common mistakes
- Chasing losses. Raising stakes after a loss to win it back adds risk and no edge.
- Raising stakes after a win. Betting bigger because you feel hot is emotion, not a plan.
- Changing the rule after every bad week. You never get a sample to judge.
- Backing short favourites because they feel safe. A likely win at a poor price is still a bad bet.
- Counting only what you remember. Wins stick and losses fade. Keep the spreadsheet.
- Judging by the last ten bets. Ten bets is noise, and as the table shows, so is a hundred.
- Betting on a match because it's on television. A televised match is no more likely to offer value.
When to stop
A strategy with no edge should be dropped, and finding that out is a good result of testing, especially when it cost nothing in the simulator. Set a limit before you begin. If betting stops being affordable or fun, stop, and see our responsible gambling page.
Football betting strategy FAQ
Is there a betting strategy that always wins?
No. Every result carries uncertainty and every price carries a margin, so no method can guarantee a win. The realistic goal is to find out whether your bets have positive expected value over a large sample, and to drop them if they don't.
What makes a good football betting strategy?
A written rule, a clear price threshold, fixed staking, a full record of bets and enough of them to judge. It should be specific enough that someone else could follow it and place the same bets.
How much should I stake per bet?
As an illustration, 1 to 2% flat of money you can afford to lose completely. This isn't financial advice. Small flat stakes make results easier to read and losing runs easier to survive.
How many bets do I need to test a strategy?
More than most people expect. Under about 100 bets, luck moves the result more than skill does. A checkpoint at 1,000 bets is far more reliable, though the table above shows it can still mislead.
Can I test a betting strategy without real money?
Yes. Our simulator is free: you sign in with Google and get $100 of virtual money a week. There is no deposit and no real money involved.
Does FootyMath's model beat the bookmakers?
That is not proven. The model's accuracy is measured and public, but its results against bookmaker prices have only been in a forward test since 22 October 2026.
FootyMath doesn't take bets and isn't a bookmaker. Your next step needs no money at all: write your rule in one sentence, then run it in the simulator.