Why your results don’t reflect your skill
In poker, your result over any period is your edge plus the luck of the cards. Over a few thousand hands, luck dominates. Over hundreds of thousands, your edge eventually shows. In between, a good player can lose for months and a losing player can win, without either of them changing level.
The calculator puts a number on that margin: for a given true win rate, it shows the range your results fall into 68% and 95% of the time, and the chance of being down over the period.
The calculation
Over N hands, the result roughly follows a normal distribution:
- expected result = win rate × N ÷ 100
- standard deviation of the result = standard deviation × √(N ÷ 100)
The 68% range is one standard deviation either side of the expected result, the 95% range is 1.96 standard deviations either side. The expected result grows with the number of hands, the standard deviation only with its square root: that’s why your edge eventually dominates, but slowly.
Read the other way round, the same calculation gives the accuracy of your observed win rate: plus or minus 1.96 × standard deviation ÷ √(N ÷ 100) bb/100.
Example: 50,000 hands at 5 bb/100
A 6-max NL Hold’em player has a true win rate of 5 bb/100 and a standard deviation of 90 bb/100. They play 50,000 hands.
- Expected result: 5 × 500 = 2,500 big blinds, or 25 buy-ins
- Standard deviation of the result: 90 × √500 ≈ 2,012 big blinds
- 95% range: −1,444 to +6,444 big blinds
- Chance of being down: about 10.7%
So one winning player in ten finishes these 50,000 hands in the red. And their observed win rate can land anywhere between −2.9 and 12.9 bb/100.
The chance of being down shrinks with volume: 28.9% over 10,000 hands, 10.7% over 50,000, 4% over 100,000.
How many hands to know your win rate
With a standard deviation of 90 bb/100, here is how accurate your observed win rate is (95% interval):
| Hands played | Win rate accuracy |
|---|---|
| 10,000 | ±17.6 bb/100 |
| 50,000 | ±7.9 bb/100 |
| 100,000 | ±5.6 bb/100 |
| 500,000 | ±2.5 bb/100 |
Even after 100,000 hands, a displayed win rate of 5 bb/100 is still consistent with a true win rate close to zero. That’s why you should enter a conservative value in the risk of ruin and bankroll calculators.
Limits of the model
- Normal distribution. The approximation holds well in cash games over a few thousand hands or more. It doesn’t suit tournaments, where results are heavily skewed.
- Constant win rate. The calculation assumes a fixed true win rate. In practice it varies with the tables, the stakes and your form.
- Final result only. The range applies to the result at the end of the period. Along the way, downswings can go deeper than the bottom of the range.
- Rake. Use your win rate after rake, which is what your tracker shows.