Bet Sizing and Losing Streak Calculator
Bet 1% to 2% of your bankroll a play, or a quarter of what the Kelly criterion says, and plan for a losing streak longer than feels possible: a bettor who wins 54% of 500 bets has an even chance of losing 7 in a row somewhere in the season. Put in your own numbers to see your stake, your streak and how the season could run.
Bet size, losing streaks and a season's swings
Your bankroll, the price you usually bet at and how often you really win. See the stake, the losing streak a season will throw at you, and how the bankroll could run.
How you size bets
Kelly re-sizes every bet to the bankroll as it stands, so stakes shrink in a slump and grow in a run.
The losing streak to plan for
| Losses in a row | Chance this season | What it costs |
|---|
1,000 seasons like yours
The same seasons, other ways to stake
| Staking | First bet | Median end | Finish down | Halved at some point |
|---|
The simulation bets every play at the same price and win rate, with no pushes, one bet at a time. A flat bettor whose bankroll falls under a unit bets what is left; a bankroll under $1 counts as broke. A real season is lumpier: your true win rate is unknown and changes, prices vary, and bets on the same game move together.
How much of your bankroll should you bet on a prop?
Two rules cover most bettors. Flat units: decide a unit, usually 1% to 2% of the bankroll you start the season with, and bet one unit a play, more only for a play you rate far above the rest. A fraction of Kelly: size each bet to your edge and the bankroll as it stands, so stakes fall in a slump and rise in a good run.
Whichever you use, the bankroll is money set aside for betting and nothing else, and the unit is fixed before the season, not after a bad night. Bankroll management, step by step goes through setting one up.
What is the Kelly criterion?
John Kelly showed in 1956 that a bettor with an edge grows a bankroll fastest by staking a fixed share of it, the edge divided by the net odds: (b × p − q) ÷ b, where b is the profit a $1 bet makes, p the chance of winning and q the chance of losing. At -110, b is 0.909. A bettor who wins 54% of the time has an edge of 3.1% a bet, and full Kelly is (0.909 × 0.54 − 0.46) ÷ 0.909 = 3.4% of the bankroll.
The catch is p. Kelly is the best stake only if your win rate is exactly right, and nobody knows theirs that well: stake full Kelly on a win rate you have overestimated and you are betting more than Kelly, where the swings grow and the growth falls. Betting twice Kelly or more shrinks the bankroll over time even with a real edge. That is why most bettors who use Kelly use a quarter or a half of it, which gives up a little growth for far smaller swings.
How long a losing streak should you expect?
Longer than you think. Over a season of 500 bets, the chance of a losing run of each length somewhere in it, worked out exactly from the win rate:
| Losses in a row | Win 54% of bets |
|---|---|
| 6 or more | 92.8% |
| 7 or more | 69.4% |
| 8 or more | 41.7% |
| 9 or more | 21.9% |
| 10 or more | 10.7% |
| 12 or more | 2.4% |
The longest losing streak averages 7.5 bets at a 54% win rate and 8.3 at 50%, so a winning bettor and a coin flipper see streaks of much the same length; a run of seven straight losses says almost nothing about whether you have an edge. At 1% units, a 10-bet slide costs 10% of the bankroll. At 5% units it costs half. The table in the calculator puts your own stake against each streak. For getting through one without chasing, see how to survive a losing streak and tilt triggers and how to control them.
Why simulate a season?
An edge shows itself only over hundreds of bets, and even then not always. In the calculator's 1,000 seasons of 500 bets at 54% and -110, a bettor staking a flat 1% still finished down in about 1 season in 5. Raising the unit to 5% made the median season better and the bad ones much worse: about 16% of those seasons went broke. Full Kelly fell to half the bankroll at some point in about 1 season in 5, where a quarter of Kelly almost never did. The same luck is used for every rule in the comparison table, so the rows differ only in the staking.
The simulation is simpler than a real season: every bet is at the same price and win rate, one at a time, with no pushes and no props on the same game moving together. Correlated bets, like the same player's points and assists, swing harder than the chart shows. Price your plays with the player prop odds calculator and keep the record honest with the prop bet tracker, then come back with the win rate it shows.
Sources: the stake rule from J. L. Kelly Jr., "A New Interpretation of Information Rate", Bell System Technical Journal 35 (1956); the streak chances and the seasons are arithmetic on the figures typed in (an exact recursion over the length of the current losing run, and 1,000 seeded random seasons). Nothing here is financial advice, and an edge is never certain. Sports betting is for adults where it is legal in your state; bet only what you can afford to lose. If gambling stops being fun, call or text 1-800-MY-RESET.