Can You Profit From Sports Betting? Break-Even and Margins
You win 52 of 100 bets at -110 and still finish behind.
For the typical bettor sports betting is not profitable, and telling a real edge from luck takes far more bets than most people expect.
Every price carries a margin. At -110 (decimal 1.909) you must win 52.38 percent of your bets just to break even, and a bettor who wins exactly half loses 4.55 percent of every unit staked. An edge is possible in the maths. Yet a bettor whose true win rate is 55 percent is still behind after 1,000 bets about 4.6 times in 100, and proving a small edge can take thousands of bets. No public, reliable statistic says what share of bettors, counted as people over a defined period, come out ahead.
What win rate does sports betting need to break even?
At -110 you stake 110 to win 100. Win 52 of 100 bets and the winners pay 5,200 while the losers cost 5,280, so you end 80 down. Win 53 and it flips (5,300 paid against 5,170 lost) to a profit of 130.
The break-even win rate is 1 divided by the decimal odds. At -110, which is 1.909 in decimal, it is 52.38 percent: a bettor who wins that share of bets at that price only breaks even, on average.
To see what happens to bettors who do beat the price, read why bookmakers limit winning accounts. This page starts one step earlier, with how much beating it takes.
| Decimal odds | American odds | Break-even win rate |
|---|---|---|
| 1.80 | -125 | 55.56% |
| 1.909 | -110 | 52.38% |
| 1.95 | no exact match | 51.28% |
| 2.00 | +100 | 50.00% |
| 2.10 | +110 | 47.62% |
The 1.909 row is American -110.
Price both sides of a two-way market at -110 and the implied probabilities, 52.38 percent each, add up to 104.76 percent. The extra 4.76 percent is the overround, the margin built into the prices. A bettor who wins exactly half of their bets at that price loses 4.55 percent of the money staked, which over 1,000 bets of one unit is an average loss of about 45.5 units. Do not mix up the two: 4.76 is the margin in the prices and 4.55 is the share of the stake a 50 percent bettor loses.
Move both sides to 1.95 and the overround falls to 2.56 percent and the break-even rate to 51.28 percent, 1.10 points lower. The same 50 percent bettor then loses 25 units per 1,000 bets instead of 45.5. A bettor who wins 51.5 percent of the time clears the hurdle at 1.95 and falls short at -110.
To turn any set of prices into an overround, use the bookmaker margin calculator.
How many bets does it take to prove an edge?
A record is one draw from a win rate you never get to see. The figures from here on are illustrations, and none describes a real bettor. They assume a fixed stake of one unit at the same price, the same true win rate each time, independent bets, no voided bets, no commission and no selection bias.
A bettor whose true win rate is 55 percent at -110 expects +5.0 units after 100 bets. Results typically scatter by about 9.5 units around that figure (one standard deviation), nearly twice the expected gain. After 1,000 bets the expected gain is +50.0 units and the scatter is 30.0: ten times the bets bring ten times the gain and about three times the noise.
For that 55 percent bettor, the chance of being ahead, meaning a net result above zero, is 69.3 percent after 100 bets, 78.3 percent after 200, 88.7 percent after 500 and 95.4 percent after 1,000. A bettor with a true 50 percent win rate expects to be 45.5 units down after 1,000 bets. These are exact binomial probabilities.
Cumulative result after n bets at -110, for a 50 percent and a 55 percent bettor
Published formula, calculatedView the data as a table
| Number of bets | 55 percent bettor, median | 50 percent bettor, median | 55 percent bettor, 95th percentile | 50 percent bettor, 95th percentile | 55 percent bettor, 5th percentile | 50 percent bettor, 5th percentile |
|---|---|---|---|---|---|---|
| 0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
| 100 | 5.0 | -4.5 | 20.3 | 10.7 | -10.3 | -19.8 |
| 200 | 10.0 | -9.1 | 32.9 | 13.8 | -12.9 | -32.0 |
| 300 | 15.0 | -13.6 | 41.7 | 13.1 | -11.7 | -40.4 |
| 400 | 20.0 | -18.2 | 50.5 | 12.4 | -10.5 | -48.7 |
| 500 | 25.0 | -22.7 | 59.4 | 11.6 | -9.4 | -57.1 |
| 600 | 30.0 | -27.3 | 68.2 | 10.9 | -8.2 | -65.5 |
| 700 | 35.0 | -31.8 | 77.0 | 10.2 | -7.0 | -73.8 |
| 800 | 40.0 | -36.4 | 83.9 | 7.5 | -3.9 | -80.3 |
| 900 | 45.0 | -40.9 | 92.7 | 6.8 | -2.7 | -88.6 |
| 1000 | 50.0 | -45.5 | 99.6 | 4.2 | 0.4 | -95.1 |
After 1,000 bets, the 55 percent bettor's result falls between about 0 and about +100 units in more than 9 cases out of 10, and the 50 percent bettor's between about -95 and about +4. The lower edge of the 55 percent range only rises above zero at 1,000 bets, at +0.4 units. Until then, the unluckiest 1 in 20 of the bettors with a real edge are behind.
How many bets before the record itself gives a verdict? At -110, a standard test (one-sided, 5 percent level, 80 percent power) gives the figures below.
| True win rate | Points above break-even | Expected gain per one-unit bet | About this many bets |
|---|---|---|---|
| 53% | 0.62 | +0.0118 | 40,200 |
| 54% | 1.62 | +0.0309 | 5,900 |
| 55% | 2.62 | +0.0500 | 2,250 |
| 57% | 4.62 | +0.0882 | 720 |
Bets needed: n = [z1 × s0 + z2 × s1]² / m². Here s0 is the standard deviation of one bet's result under no edge, 0.9535 units at -110; s1 is the same at the true win rate; m is the expected gain per bet at that rate. Constants: z1 = 1.6449 for a one-sided 5 percent test and z2 = 0.8416 for 80 percent power. The values come from a normal approximation and are rounded.
Because a 55 percent bettor gains only 0.05 units per bet on average while a single bet swings by nearly one unit either way, the gain shows through the noise only after enough bets have piled up, since noise grows with the square root of the count and gain grows in step with it.
At one bet a day, 2,250 bets take more than six years and 40,200 more than a century. A real record can need more or fewer bets, so read the table as an illustration.
If nobody capped your stakes, the wait would not get shorter: a tenfold stake makes the swings ten times larger and leaves the number of bets needed where it was.
What share of sports bettors is profitable, and where do the percentages come from?
Search this question and one claim keeps coming back: only a small share of bettors, 3 percent or 5 percent, make money, and the rest lose. Those figures circulate online without a named study. What is missing is a source you can check, with a definition of who counts as a bettor, how long they were watched and whether the count is of people or accounts.
A bettor whose true win rate is 50 percent at -110 is ahead after 10 bets 37.7 percent of the time, after 100 bets 30.9 percent of the time and after 1,000 bets 6.9 percent of the time. A share of winners therefore means little until it says how many bets each bettor had placed.
The number that would settle it is a share of people over a defined period, from a large group followed for years. It is not public.
These three claims recur on betting pages, each without a named study.
- “Only 3 percent of bettors are profitable”, sometimes 5 percent.
- “95 percent of bettors lose.”
- A typical yield or win rate for professional bettors.
No public, reliable statistic gives the share of bettors, meaning people, who make money over a defined period. The nearest regulator figure counts accounts, and it is covered below with its limits.
What do the solid numbers say, and what do they leave out?
What the industry's own numbers show
The American Gaming Association reported, as relayed by ESPN, that legal sports betting in the United States generated 10.92 billion dollars of revenue on 119.84 billion dollars wagered in 2023, a sportsbook win percentage of 9.1 percent. The figure covers the states that reported and leaves out some months in Arizona and Kentucky, and the AGA report itself could not be read at the source. The 9.1 percent averages over every bet from every bettor, which makes it a different quantity from the margin on one bet at -110.
A National Bureau of Economic Research working paper by Scott Baker and four co-authors quotes almost the same figures. Its 2023 totals, over 120 billion dollars wagered and 11 billion of revenue, come from the ESPN article, and it computes no margin of its own. The authors reason that the industry's profitability suggests the typical bettor faces negative expected returns, an inference from operators' results, and their data, household bank and card records, do not measure who wins. The text read here is the working-paper version.
What the British regulator counted
The closest figure to a share of winners comes from the UK Gambling Commission, and its unit is the account. For large British online operators in calendar year 2024, 25.42 percent of active accounts were in net profit, measured over the life of the account on settled bets only. The regulator's post does not say how many bets each account had placed.
Winners are over-represented among restricted accounts, yet 51.29 percent of them were net losers, so being restricted does not show that a bettor is ahead. You can find the full regulator breakdown on the limits guide, and restricted bettors deciding what to look at next can turn to the guide to venues that treat winning accounts differently.
One football strategy tested with real money
In a 2017 preprint on arXiv, Lisandro Kaunitz, Shenjun Zhong and Javier Kreiner describe a football betting strategy that they report was profitable in historical simulations and in five months of real-money bets. They also report that some bookmakers began to limit their accounts, which ended the experiment.
The real-money part was 265 bets. The authors' own significance test on their 672 bets, paper trading and real money together, gives a p-value of 0.089, above the usual 0.05 threshold. The strategy won fewer than half of its bets, 44 to 48 percent in the paper's table, while still reporting a profit, which means it bet at long odds. Its win rate cannot be compared with the break-even rates on this page, and the preprint gives no journal reference.
Tax, with one country as the example
In the United States, the IRS topic page says its guidance applies to casual gamblers who are not in the trade or business of gambling. It says winnings, sports betting included, are fully taxable and must be reported, and that losses can be deducted only if you itemise and keep records, up to the gambling income reported.
The federal tax code goes further. In the text compiled by Cornell's Legal Information Institute under 26 USC 165, subsection d, the deduction for tax years beginning after December 31, 2025 is limited to 90 percent of wagering losses, and only up to wagering gains. The IRS page as read does not mention the 90 percent limit, and this page draws no conclusion about why. Nothing here covers professional gamblers, whom that page leaves out. This is not tax advice: check the rules where you live.
How many bets would 30,000 units a year take?
Betting income is the amount staked multiplied by the yield, which is profit as a share of the stakes. To make 30,000 units in a year at a yield of 2 percent, you would have to stake 1,500,000 units. At 3 percent it takes 1,000,000 units and at 5 percent 600,000. At an average stake of 100, that is 15,000, 10,000 and 6,000 bets a year (about 41, 27 and 16 a day).
| Yield | Units staked in a year | Bets a year at an average stake of 100 | Bets per day, about | Win rate needed at -110, flat stakes |
|---|---|---|---|---|
| 2% | 1,500,000 | 15,000 | 41 | 53.43% |
| 3% | 1,000,000 | 10,000 | 27 | 53.95% |
| 5% | 600,000 | 6,000 | 16 | 55.00% |
| Yield professionals actually reach | Not public, so the row stays empty: use your own assumption. | |||
Yield = win rate × decimal odds - 1, so at -110 with flat stakes a 2 percent yield means winning 53.43 percent of bets.
No reliable public data shows what yields professional bettors actually achieve. The yields are inputs, and the table promises nothing. What running a betting operation involves beyond the maths is a separate subject: becoming a professional bettor.
What should you do with these numbers?
If you keep no record of your bets, with the stake, the price you took and the price at the close, you cannot tell which line of the chart you are on. Even with a record, a few hundred bets would settle little.
To see how a fair price is estimated, start with value betting. You need to be 18 or over, and betting law and tax on winnings differ by country. If a losing run has you staking more to recoup it, the tools for staying in control are a better next step than another bet.