Poker variance: how big can a downswing get?
On this page
Every poker player who has kept records knows the feeling: you are sure you play well, and you have lost for a month. Is that bad luck or bad play? The honest answer is that over short and even medium stretches, you often can't tell. This guide puts numbers on that uncertainty.
We simulated 20,000 poker "careers" for each of several win rates, and checked the results against standard formulas. The figures show how often a winning player is behind, how deep downswings get, and how many hands it takes before a win rate means much. They are a guide to expectations, not a strategy, and not a promise that anyone will win.
Assumptions, stated up front. Results are measured in big blinds per 100 hands (bb/100), after rake. A buy-in is 100 big blinds. We assume each block of 100 hands is independent and roughly normally distributed, with a standard deviation of 80 or 100 bb/100. We chose those values to illustrate the range for six-handed no-limit hold'em; your own figure will differ by game, format and style, and tracking software will report it. Win rates of 2, 5 and 10 bb/100 are illustrative, not typical or achievable for most players.
Win rate and standard deviation
Two numbers describe a cash-game player's results:
- Win rate: average profit, in big blinds per 100 hands, after rake. Rake is the fee the poker room takes from pots, and it comes straight out of your win rate.
- Standard deviation: how much results swing around that average over 100 hands. In no-limit hold'em it is many times larger than any realistic win rate.
That gap is the whole story of variance. With a win rate of 5 bb/100 and a standard deviation of 100 bb/100, a single 100-hand session ranges routinely from a loss of more than a buy-in to a win of more than one. Over 100 hands your win rate is barely visible underneath that.
How often a winning player is behind
The chance of being behind after a given number of hands follows directly from the two numbers. It is the probability that a normally distributed result, with mean equal to win rate × (hands ÷ 100) and standard deviation equal to SD × √(hands ÷ 100), is below zero.
| Win rate | SD | After 10,000 hands | After 50,000 | After 100,000 | After 500,000 |
|---|---|---|---|---|---|
| 2 bb/100 | 80 | 40.1% | 28.8% | 21.5% | 3.9% |
| 2 bb/100 | 100 | 42.1% | 32.7% | 26.4% | 7.9% |
| 5 bb/100 | 80 | 26.6% | 8.1% | 2.4% | <0.1% |
| 5 bb/100 | 100 | 30.9% | 13.2% | 5.7% | <0.1% |
| 10 bb/100 | 80 | 10.6% | 0.3% | <0.1% | <0.1% |
| 10 bb/100 | 100 | 15.9% | 1.3% | 0.1% | <0.1% |
Calculation: P(behind) = Φ(−win rate × √(hands/100) ÷ SD), where Φ is the standard normal distribution. Our simulation (below) agreed with these figures to within about one percentage point.
What this means. A genuine 2 bb/100 winner, a solid result in many games, has roughly a two-in-five chance of being behind after 10,000 hands. Even after 100,000 hands, about one in four or five of those careers is still losing.
How wide the range of results is
Another way to see it: the range in which about 95% of results fall after 100,000 hands, which is 1,000 blocks of 100.
| Win rate | SD | Expected result | 95% range |
|---|---|---|---|
| 2 bb/100 | 80 | +20 buy-ins | −30 to +70 buy-ins |
| 2 bb/100 | 100 | +20 buy-ins | −42 to +82 buy-ins |
| 5 bb/100 | 80 | +50 buy-ins | 0 to +100 buy-ins |
| 5 bb/100 | 100 | +50 buy-ins | −12 to +112 buy-ins |
| 10 bb/100 | 80 | +100 buy-ins | +50 to +150 buy-ins |
| 10 bb/100 | 100 | +100 buy-ins | +38 to +162 buy-ins |
Calculation: expected result ± 1.96 × SD × √1,000, converted to buy-ins (÷ 100).
Two players with exactly the same skill can finish 100,000 hands one hundred buy-ins apart, purely from luck.
How big downswings get
Being behind overall is one thing. A downswing, the largest drop from a high point to a later low point, is what tests nerves and bankrolls. There's no simple formula for the whole distribution, so we simulated it: 20,000 careers for each combination, seeded so the results can be reproduced.
| Win rate | SD | Hands | Median worst downswing | Chance of a 10 buy-in downswing | 20 buy-ins | 30 buy-ins |
|---|---|---|---|---|---|---|
| 2 bb/100 | 80 | 10,000 | 7.5 | 26.5% | 1.0% | <0.1% |
| 2 bb/100 | 80 | 100,000 | 21.8 | 99.2% | 59.0% | 22.2% |
| 2 bb/100 | 100 | 100,000 | 28.5 | 100.0% | 82.8% | 44.8% |
| 5 bb/100 | 80 | 10,000 | 6.6 | 17.8% | 0.3% | <0.1% |
| 5 bb/100 | 80 | 100,000 | 16.2 | 95.0% | 27.8% | 4.9% |
| 5 bb/100 | 100 | 100,000 | 22.1 | 99.7% | 62.0% | 21.0% |
| 10 bb/100 | 80 | 100,000 | 11.6 | 70.2% | 4.1% | 0.2% |
| 10 bb/100 | 100 | 100,000 | 16.3 | 96.8% | 25.9% | 3.5% |
Simulation: 20,000 paths per row; results in 100-hand steps; worst downswing measured from the highest point reached (starting at zero). Downswings are in buy-ins of 100 big blinds.
What this means. Over 100,000 hands, a 5 bb/100 winner with a 100 bb/100 standard deviation has about a 62% chance of a downswing of 20 buy-ins or more, and about a one-in-five chance of 30 buy-ins or more. For a 2 bb/100 winner, a 20 buy-in downswing is more likely than not. These are winning players, losing in a way that is entirely normal for them.
Editorial view. If a 20 buy-in downswing would end your poker, or worse, affect money you need for anything else, you are playing stakes that are too high for your circumstances, whatever your win rate.
Risk of ruin
Risk of ruin is the probability of losing an entire bankroll, starting from a given number of buy-ins, if you keep playing the same stakes. A standard approximation for a player with a known positive win rate is:
risk of ruin ≈ e^(−2 × win rate × bankroll ÷ SD²) (with win rate, bankroll and SD in the same units)
| Win rate | SD | 20 buy-ins | 30 buy-ins | 50 buy-ins | 100 buy-ins |
|---|---|---|---|---|---|
| 2 bb/100 | 80 | 28.7% | 15.3% | 4.4% | 0.2% |
| 2 bb/100 | 100 | 44.9% | 30.1% | 13.5% | 1.8% |
| 5 bb/100 | 80 | 4.4% | 0.9% | <0.1% | <0.1% |
| 5 bb/100 | 100 | 13.5% | 5.0% | 0.7% | <0.1% |
| 10 bb/100 | 80 | 0.2% | <0.1% | <0.1% | <0.1% |
| 10 bb/100 | 100 | 1.8% | 0.3% | <0.1% | <0.1% |
Calculation: the formula above. It assumes the win rate is known exactly and never changes, and that the player never moves down in stakes.
Read this table with care, because its assumptions flatter the player:
- Nobody knows their true win rate (see the next section). If your real win rate is lower than you think, your real risk is higher, and for a losing player the risk of ruin, over enough hands, is effectively 100%.
- Win rates change. Games get tougher, and players' form changes.
- The formula assumes you keep playing. In practice, moving down in stakes after losses (a stop-loss rule) is the main way players manage this risk.
How many hands before you know your win rate?
The uncertainty in an observed win rate shrinks only with the square root of the number of hands. To be 95% confident that your true win rate is within a given margin of your observed one, you need roughly:
| Standard deviation | Within ±2 bb/100 | Within ±1 bb/100 |
|---|---|---|
| 80 bb/100 | about 615,000 hands | about 2.46 million hands |
| 100 bb/100 | about 960,000 hands | about 3.84 million hands |
Calculation: hands ≈ (1.96 × SD ÷ margin)² × 100.
So after 100,000 hands, a respectable sample for most recreational players, an observed win rate of 5 bb/100 with a standard deviation of 100 is consistent with a true win rate anywhere from about −1 to +11 bb/100. Short-term results can't show whether you are a winning player.
Why rake matters: a worked example
Rake is easy to ignore because it's taken a little at a time, from each pot. But because it comes straight off your win rate, small differences compound. Suppose a change of stakes, format or room reduced the rake you pay by 2 bb/100, turning a 3 bb/100 winner into a 5 bb/100 winner, with a standard deviation of 100 bb/100 throughout.
| 3 bb/100 | 5 bb/100 | |
|---|---|---|
| Expected result after 100,000 hands | +30 buy-ins | +50 buy-ins |
| Chance of being behind after 100,000 hands | 17.1% | 5.7% |
| Risk of ruin from a 30 buy-in bankroll | 16.5% | 5.0% |
Calculation: the formulas earlier in this guide. The 2 bb/100 difference is an illustrative assumption, not a measured difference between any poker rooms.
The same player, with the same skill, carries about three times the risk of ruin at the higher rake. Before comparing poker rooms on promotions, compare what each charges, using the rake information it publishes.
What this means for how you play
- Expect long losing runs, even if you play well. The tables above are for winning players.
- Judge decisions, not results. Over thousands of hands, results mostly reflect luck. Hand reviews and study tell you more about your play than your graph does.
- Rake matters more than it looks. It comes straight off your win rate, and a small win rate is the difference between the top and bottom rows of the risk-of-ruin table.
- Keep poker money separate, and never play with money you need for bills or savings.
- Watch for tilt. Downswings are when players chase losses, move up to "win it back", or play longer than planned. If that's happening, stop for the day.
Tax and regulation in the UK
The Gambling Commission states that for customers in the UK, "winnings from gambling for customers are tax-free, regardless of the amount won." If your situation is unusual, for example poker is part of a business you run, take professional tax advice.
Licensed online operators must let you set a deposit limit. You can also self-exclude from all UK-licensed online gambling, including poker, through GAMSTOP. If poker has stopped being fun, the National Gambling Helpline, run by GamCare, is free and confidential: 0808 8020 133.
Questions and answers
What is a normal downswing in poker?
It depends on your win rate and game, but in our model a winning six-handed no-limit player playing 100,000 hands has a median worst downswing of roughly 12 to 29 buy-ins, depending on win rate and standard deviation. A downswing of 20 buy-ins or more is common.
How do I find my own standard deviation?
Poker tracking software usually reports it in big blinds per 100 hands. It is more stable than win rate, so a few tens of thousands of hands give a reasonable estimate.
Are tournaments different?
Yes. Tournament results are far more volatile than cash-game results, because most of the return comes from rare deep finishes. The normal approximation used here doesn't describe tournament results well.
Related on DannyBet
Sources and method
- Simulation and calculations:
editorial/calculations/poker_variance.pyin the DannyBet repository (Python with NumPy; 20,000 paths per row; seed 20261022). - Gambling Commission, Winnings from gambling for customers (FOI response, request dated 12 May 2026). Checked 22 October 2026.
- Help and self-exclusion: GamCare, GAMSTOP.