There are profitable gamblers, but their win rate alone says little about what they earn. A roulette system can win on 99% of attempts and still lose money on average. A skilled bettor can find favourable bets and still earn very little for the hours spent finding them.
The system that wins 99% of the time
Consider this roulette scheme: bet £1 on red, double after each loss, and stop when you win. Each successful sequence recovers the earlier losses and leaves £1 profit. With £127 available, the sequence can run to seven bets: £1, £2, £4, £8, £16, £32 and £64.
On a standard single-zero wheel, the chance of winning before those seven bets run out is 99.06%. A long record of successful attempts would be entirely plausible.
But the remaining 0.94% of attempts lose the entire £127. One such loss wipes out 127 successful sequences. Allow for both outcomes and the average result is a loss of about 21p each time the sequence starts.
The arithmetic is fixed by the game's rules: red wins on 18 of 37 numbers, but pays as though it wins half the time. On average, ordinary red/black bets lose 2.7p per £1 staked. Doubling changes the pattern of wins and losses, not the disadvantage on each pound.
Sports odds, poker and bonuses
Sports betting is different because the bookmaker has to put a price on an uncertain event. A favourite can be a bad bet if the odds are too short; an outsider can be a good bet if the price more than compensates for the chance of losing.
At decimal odds of 1.90, a winning £10 bet returns £19, including the stake. Win half your bets and each £20 staked brings back £19. You need to win more than about 52.6% at those odds to make money. Knowing a sport well helps only if it leads to better estimates than the prices available.
In poker, players compete against each other and pay the operator a commission. A study of 456 million online poker player-hand observations found that stronger players tended to remain stronger in a later period. Yet the operator's cut was substantial: 37.5% of players finished ahead before commission, compared with 32% afterwards. Those were games at one operator in 2009–10, excluding unobserved bonuses and rebates; they are evidence of persistent skill, not a current success rate for new players.
Bonuses offer another possible source of profit because the operator contributes money. A successful promotion does not establish that the bets would be profitable without the subsidy. Required bets, withdrawal conditions and the supply of further offers determine whether it can be repeated.
£450 profit: skill or luck?
Suppose a bettor wins 550 of 1,000 bets, each for £10 at odds of 1.90. The record shows a £450 profit. That is a return of 4.5% on £10,000 staked.
That record alone does not establish a lasting advantage. Assuming independent bets with the same underlying chance of winning, the 95% confidence interval for that chance runs from 51.9% to 58.1%. It includes the 52.6% break-even point. A bettor with exactly break-even chances would produce a record at least this good about 7% of the time.
The £450 is profit already made; the uncertainty concerns whether the success will continue. Choosing a system because it did well in last year's results makes the evidence weaker still: it has already had the benefit of hindsight.
A useful record includes every stake, return and fee, across all accounts, with the selections recorded before the results arrive. Several bets on the same match also provide less independent evidence than the same number spread across unrelated events.
Profit, losses and hours worked
Suppose the bettor really can sustain a 55% chance of winning at odds of 1.90. Each £10 bet then earns 45p on average. The calculation below assumes independent bets and enough money to complete them all.
| Number of bets | Total staked | Average profit | Chance of finishing with a loss |
|---|---|---|---|
| 100 | £1,000 | £45 | 30.7% |
| 1,000 | £10,000 | £450 | 6.8% |
Even with this advantage, a hundred bets produce a loss nearly a third of the time. The table describes the result at the end; losses along the way can be larger. Someone who runs out of money cannot carry on to the planned thousand bets.
Finding and placing a thousand bets would take about 167 hours at ten minutes each. An average £450 profit would pay £2.70 an hour, before subscriptions or other costs. These are illustrative timings, not a survey of professional bettors.
Increasing each stake from £10 to £100 would multiply the profit by ten, but would also multiply every possible loss by ten. It would not reduce the chance of finishing down. To average £20,000 profit at a 4.5% return requires about £444,000 of stakes. That is turnover over time, not necessarily £444,000 deposited at once.
Bookmakers can limit stakes or close accounts. The Gambling Commission examined almost 15 million accounts active during 2024. Commercial restrictions affected 4.3%, including stake limits and closures. About 47% of restricted accounts were in lifetime profit, compared with 25% overall. These are accounts, not unique people or annual incomes, but the restrictions are a practical limit on repeating successful bets.
Sources and calculations
Roulette. The published game rules specify 37 pockets, 18 red numbers, even-money payouts on red and a loss on zero. No zero-refund rule is assumed. The seven-bet scheme has P(all losses)=(19/37)7 ≈ 0.94%. Its expected result is £1×(1−P)−£127×P ≈ −£0.21 per sequence. It is analysed as a losing scheme, not recommended as a strategy.
Sports bets. At decimal odds d and win probability p, expected profit per £1 is pd−1. The 550/1,000 record has a Wilson 95% confidence interval of 51.9%–58.1%. Under p=1/1.90, P(W≥550) ≈ 7.1% for W ~ Binomial(1000,p). This is not the probability that the bettor has no skill.
For the conditional 55% example, profit after n bets is £10(1.90W−n), W ~ Binomial(n,0.55). P(loss)=P(W<n/1.90), giving about 30.7% for n=100 and 6.8% for n=1000. Actual sports bets have different odds and chances, may share risks, and cannot be assumed to retain a 55% chance at these prices. The worked example isolates what a genuine advantage would and would not imply; it does not estimate a reader's prospects. £20,000÷0.045 ≈ £444,000 of turnover. £450÷(1,000×10/60 hours)=£2.70 per hour.
The Commission sample contained 14,923,840 accounts placing at least one bet in 2024, including subsequently closed accounts. Profit was lifetime profit on settled bets. Missing data and break-even accounts explain why its profit and loss shares do not add to 100%. Restrictions overlap and their severity varies. The poker study separates analysis of persistent skill from the cost of commission; bonuses and rebates were not observed.
Checked 14 October 2026.