Zero hands
Expected result and spread both remain zero regardless of the entered rate assumptions.
Poker Variance Calculator
Explore how a win rate and standard deviation scale across a chosen hand sample. The poker variance calculator reports theoretical spread and runs a seeded local simulation in a Web Worker, keeping uncertainty distinct from expected value.
Seeded cash-game variance model
Percentiles use a normal approximation in bb/100, not an exact forecast.
How the result is built
The Same Win Rate Can Produce Very Different Short-Term Results
Seeded sample paths can finish far apart even though they share one expected line. The bands describe model percentiles, not guaranteed limits.
- Expected
- P50
- P25–P75
- P5–P95
- Seeded paths
| Endpoint | P5 | P25 | P50 | P75 | P95 |
|---|---|---|---|---|---|
| 20,000 hands | -1,093.6 bb | 141.5 bb | 1,000 bb | 1,858.5 bb | 3,093.6 bb |
Mean grows linearly; spread grows by square root
For H hands, expected big blinds scale with H/100, while total standard deviation scales with the square root of H/100. More volume narrows relative noise but not absolute swings.
Percentiles are model outputs
The fifth percentile is not a worst case. A normal approximation still assigns probability beyond the displayed range and may understate real heavy tails.
Seeded paths aid comparison
Using the same seed makes interface and test comparisons repeatable. It does not make the simulated path more likely than another legal path.
Validate assumptions before simulating
Estimate the average result and its spread from games with comparable stakes, formats, and fees. Mixing unrelated samples can make a simulation look precise while its inputs are not representative. Review several plausible input pairs and compare their negative tails and distribution shapes.
Worked examples
Double the sample
Expected result doubles, while total standard deviation grows by roughly the square root of two.
Frequently asked questions
Is the fifth percentile a guaranteed floor?
No. Outcomes can fall below it under the model.
Why use bb/100 units?
They keep win rate and standard deviation on the same 100-hand scale.
Does the normal model exactly match poker?
No. It is a transparent approximation.
What does the seed change?
It changes the reproducible pseudo-random sample, not the theoretical mean.