Bankroll management in sports betting means betting only from a fixed sum kept apart from the money you live on, and sizing every bet as a small, fixed share of that sum. The share has to be small because long losing runs are certain to come: at the break-even rate for -110 bets, a run of at least 7 straight losses turns up in 94.8% of 1,000-bet stretches (exact). A sports betting bankroll staked at 1% of its size absorbs that run. Staked flat at 5%, it was ruined in 38.95% of simulated sequences even at a 53% win rate.
Every figure on this page is for a -110 side with 1 unit staked, where a win pays 0.9091 units and a loss costs 1. Exact figures are calculated. Simulated ones come from 20,000 simulated sequences of 1,000 bets at each of three win rates: 52.38%, 53% and 55%.
Your sports betting bankroll: separate money, counted in units
The bankroll is money you can lose without it touching rent, bills or anything else you owe. Set it once, and do not top it up in the middle of a run, because a top-up hides the result you are trying to measure. That result has one bar: 52.38%, the share of -110 bets you must win to break even. The arithmetic behind that figure is on how betting odds work.
One unit is one flat stake. What a unit is in betting covers how to size it. Here the only question is how many units the bankroll holds. A 1% unit means the bankroll holds 100 stakes, a 2% unit holds 50, and a 5% unit holds 20. That count decides almost everything in the tables that follow.
What variance does at 52.38% to 55%
Variance is the spread of actual results around what a win rate should produce. At -110 that spread is wide compared with the edge, even for a bettor who wins more often than the break-even rate.
| Win rate | ROI per bet | Expected units per 100 bets | 95% range per 100 bets | Behind after 100 bets | Behind after 1,000 bets |
|---|---|---|---|---|---|
| 52.38% | 0.0% | 0.0 | -18.7 to +18.7 | 50.9% | 49.2% |
| 53% | 1.18% | +1.18 | -17.5 to +19.9 | 45.9% | 34.0% |
| 55% | 5.0% | +5.0 | -13.6 to +23.6 | 30.7% | 4.6% |
All figures exact.
At 55%, the expected gain is 5 units per 100 bets, against a standard deviation of 9.5 units. The noise is larger than the edge. A 55% bettor is still behind after 100 bets 30.7% of the time (exact).
The thinner edge is worse. A 53% bettor is behind after 100 bets 45.9% of the time, and after 1,000 bets still 34.0% of the time (exact). A thousand bets of real profit can look like a losing record, and the bankroll has to survive long enough for the edge to show.
Losing runs are normal
At the break-even rate, the chance of at least one run of 7 straight losses inside 100 bets is 24.6% (exact). Stretch that to 1,000 bets and it is 94.8%. A better win rate does not make them rare: at 55%, a run of 8 or more straight losses inside 1,000 bets happens 60.4% of the time (exact).
The simulation gives the typical worst case. Across 20,000 simulated sequences of 1,000 bets, the median longest losing run was 8 at every one of the three win rates. The 95th percentile of the longest run was 12 at 52.38% and 53%, and 11 at 55%. Plan for 8 straight losses as ordinary, and size stakes so that 12 does not end the bankroll.
Flat staking against percentage staking
Flat staking means the same stake every bet, set as a percentage of the starting bankroll. Percentage staking means a percentage of the bankroll as it stands now. Stakes shrink after losses and grow after wins. For flat staking, “ruined” means the bankroll can no longer cover one full stake. “Fell to half” means it dropped to half the starting sum at some point in the 1,000 bets.
| Stake size | Stakes in starting bankroll | Flat: ruined | Flat: fell to half | Percentage: fell to half |
|---|---|---|---|---|
| 1% | 100 | 0.03% | 4.5% | 1.2% |
| 2% | 50 | 4.95% | 27.7% | 21.9% |
| 5% | 20 | 38.95% | 62.0% | 73.6% |
53% win rate, 20,000 simulated sequences of 1,000 bets.
Stake size does most of the work. At 53%, flat stakes of 5% ruined 38.95% of sequences, even though every bet carried a positive expectation of 0.0118 units. Flat stakes of 2% ruined 4.95%, and flat stakes of 1% ruined 0.03%. A better win rate does not rescue a large stake: at 55%, flat stakes of 5% still ruined 11.09% of sequences.
Percentage staking cuts the bet after losses, so at 1% and 2% it fell to half the bankroll less often than flat staking did. At 53%, it did so in 1.2% and 21.9% of sequences, against 4.5% and 27.7% flat. At 5% the order reverses. Percentage staking fell to half in 73.6% of sequences, and the median sequence finished down 42.2% despite the positive expectation on each bet. Large percentages lose money to swings, because a win at a shrunken stake does not repair a loss taken at a larger one.
Percentage staking also does not protect the bankroll by itself. At 5% of the current bankroll and a 53% win rate, the median sequence’s lowest point was 27.1% of the starting sum. The stake size matters more than the method.
Why chasing losses and doubling up fail
Chasing means sizing the next bet to win back every loss so far, plus the 0.91 units the first bet would have won. At -110, that rule multiplies the stake by 2.1 after every loss.
| Bet | Stake (units) | Total lost if this bet loses (units) |
|---|---|---|
| 1 | 1.0 | 1.0 |
| 2 | 2.1 | 3.1 |
| 3 | 4.41 | 7.51 |
| 4 | 9.26 | 16.77 |
| 5 | 19.45 | 36.22 |
| 6 | 40.84 | 77.06 |
| 7 | 85.77 | 162.83 |
| 8 | 180.11 | 342.94 |
After 7 straight losses, 162.83 units are gone, all to win back 0.91. Bet 7 alone asks for 85.77 units, most of a bankroll that holds 100 stakes. This is the martingale strategy under another name. The run that breaks it is not rare: at the break-even rate, a run of 7 straight losses inside 1,000 bets is 94.8% likely (exact).
Keep records
Log every bet with its date, market, price, stake in units and result. The running total in units is the only thing that tells you whether you are beating 52.38% at -110, the price explained on standard betting lines. Record the price you actually took, because a worse price raises the win rate you need to break even.
A record only means something over hundreds of bets. At a 55% win rate, the 95% range for 100 bets runs from -13.6 to +23.6 units (exact). A losing 100-bet record fits a 55% bettor, and so does a winning one well above the expected +5.0 units. Whether a bet had value is a question for expected value, and a short run of results cannot settle it either way.
Set your deposit, loss and time limits before the first bet, as covered in responsible gambling.
Sources: our exact calculations and a simulation (Python, seed 20261113, 20,000 sequences of 1,000 bets per win rate), run on 13 November 2026, all at -110.