Variance in betting: why losing streaks are normal
A losing streak is what independent bets produce by arithmetic. If each bet loses with probability q, n losses in a row occur with probability q^n. Ten straight losses on a 50% bet happen about once in 1,024 attempts, and a longer series of bets makes that run likely to appear somewhere.
What variance is
Variance is the spread of results around the average. Two bets can have the same expected value and still behave very differently: one wins often in small amounts, the other wins rarely in large amounts. The average cost of each is covered in the guide on the expected value of a bet; variance describes how bumpy the road is on the way to that average.
A losing streak is simply a visible piece of that spread. It does not need a cause such as a cold run, a jinx or a rigged market. It needs only independent bets and a probability of losing that is not tiny.
The chance of a streak: q to the power n
Assume each bet is independent of the others, which means the result of one does not change the probability of the next. Let q be the probability that a single bet loses. The probability that the first bet loses is q. The probability that the first two both lose is q × q. Carrying on, the probability of n losses in a row is:
The table gives exact results for three kinds of bet: an even chance (p = 0.50), a one-in-four chance (p = 0.25, q = 0.75) and a one-in-ten chance (p = 0.10, q = 0.90). Values are rounded to the digits shown.
| Bet type | q | 3 losses in a row | 5 losses in a row | 10 losses in a row | 20 losses in a row |
|---|---|---|---|---|---|
| Even chance (p = 0.50) | 0.50 | 12.5% | 3.125% | 0.0977% | 0.0001% |
| One in four (p = 0.25) | 0.75 | 42.19% | 23.73% | 5.63% | 0.317% |
| One in ten (p = 0.10) | 0.90 | 72.9% | 59.05% | 34.87% | 12.16% |
Read the last row carefully. For a one-in-ten bet, five losses in a row is the normal state of affairs, and about a third of the time ten consecutive bets all lose.
Streaks across a whole series
The table above is the chance that one specific block of bets all lose. A series is made of many overlapping blocks, so streaks show up far more often than the single-block figure suggests. In a series of N bets there are N − n + 1 places where a block of n bets can start, and each block is all losses with probability qn. The expected number of all-loss blocks is therefore (N − n + 1) × qn. Overlapping blocks are counted separately, so a run of six losses counts as two blocks of five.
| Series | Block length n | Blocks (N − n + 1) | qn | Expected all-loss blocks |
|---|---|---|---|---|
| 100 bets, p = 0.50 | 5 | 96 | 0.03125 | 3.00 |
| 100 bets, p = 0.25 | 5 | 96 | 0.2373 | 22.78 |
| 500 bets, p = 0.50 | 5 | 496 | 0.03125 | 15.50 |
| 500 bets, p = 0.50 | 10 | 491 | 0.000977 | 0.48 |
So in 100 even-chance bets, five-loss blocks are expected about three times, and in 500 bets the expected count of ten-loss blocks is close to one half. None of that signals a fault. It is the arithmetic of repetition.
Why the average does not rescue you
People often hear that results "even out" and conclude that a streak must be followed by a compensating run of wins. Independence says otherwise. After five losses the chance of winning the sixth bet is still p, not higher. This belief is called the gambler's fallacy, and it is the reason that raising stakes after a streak feels reasonable and is not.
What does happen over time is that the average result per bet becomes a better description of the total, and for bets priced by a bookmaker that average is below zero, as the guide on calculating the bookmaker's margin explains. A streak of losses is therefore not a debt that the future repays. It is simply part of the total.
Measuring the spread
For one bet of stake S at decimal odds D with win probability p, the win pays S × (D − 1) and the loss costs S. The spread of one bet is its standard deviation, which for a win-or-lose bet works out to S × D × √(p × (1 − p)). For N independent bets of the same kind, the average result is N times the one-bet average and the spread is the one-bet spread times √N.
Illustrative example
Two bets each cost KES 5 on average per KES 100 staked, and both are bought 100 times at KES 100.
Bet A: p = 0.50, D = 1.90. EV per bet = 100 × (0.50 × 1.90 − 1) = −KES 5. Spread per bet = 100 × 1.90 × √0.25 = KES 95.
Bet B: p = 0.10, D = 9.50. EV per bet = 100 × (0.10 × 9.50 − 1) = −KES 5. Spread per bet = 100 × 9.50 × √0.09 = KES 285.
Over 100 bets both average −KES 500. The spread is 95 × √100 = KES 950 for A and 285 × √100 = KES 2,850 for B, so B's ordinary range is three times as wide.
Same average cost, very different experience. Bet B will often go many bets without a win and then pay a large amount once, which feels like skill or a change of luck when it is neither.
What this means for a budget
Because streaks are normal, a stake size should be chosen so that a long streak does not break the plan. A KES 100 stake on an even chance has a 1 in 1,024 probability (0.0977%) of losing ten in a row, which would cost KES 1,000. On a one-in-ten bet the same ten-bet run has a 34.87% probability. The same stake can be comfortable in one case and a problem in the other.
- Decide the total you can afford to lose in a session, treating it as already spent.
- Divide it by the stake to get the number of bets it covers.
- Compare that number with the streak table for the kind of bet you place. If a normal streak would exhaust the budget, the stake is too large for that bet type.
The session budget planner splits a monthly ceiling into sessions and stakes, and the guide on setting a betting budget and limits covers the order of steps. Regulation 87 of the 2026 operations regulations requires a licensee to offer deposit limits (daily, weekly and monthly), loss, session and expenditure limits and reality checks, so the limits can be set in the account itself.
Common misreadings
- "I am due a win." Independent bets have no memory. The next bet has the same probability as the first.
- "A hot streak shows a method works." Wins in a row occur by the same arithmetic as losses in a row, and a short series cannot separate skill from luck.
- "Bigger stakes will recover the losses faster." Each added bet carries its own negative expected value, and a bigger stake multiplies both the cost and the spread.
- "Variance is what makes betting profitable." Variance changes the path, not the average. A negative average stays negative.
If a streak leads to urgency, anger or a wish to recover money, that is a sign to stop. The responsible gambling page lists the national helpline and the self-exclusion process.
Questions and answers
Does a losing streak mean the next bet is more likely to win?
No. If bets are independent, the chance of the next one winning is the same as before, whatever happened earlier. The only thing that changes after a streak is the total you have lost.
Is a losing streak a sign that something is wrong?
Not by itself. Streaks of several losses appear regularly in any long series, as the tables above show. Judge a series by the expected cost of the bets, not by how the losses are arranged.
Does variance mean I will win back what I lose if I keep going?
No. Variance describes the spread of results around the average. The average itself is negative for bets priced with a margin, so a longer series moves toward that average cost, not toward zero.
Why does a long-shot bet produce longer losing streaks?
Because its losing probability q is higher, and q^n shrinks slowly when q is close to 1. At q = 0.9, twenty losses in a row have a probability of about 12%; at q = 0.5 the same run is about one in a million.
What can I do about variance?
You cannot remove it, but you can decide in advance what you can afford to lose and set deposit, loss and session limits. Licensees must provide such tools under regulation 87 of the 2026 operations regulations.
Sources
- Gambling Control (Conduct of Gambling Operations) Regulations, 2026, reg. 87, accessed 2026-10-05