A few thousand hands proves nothing. Plan on roughly 50,000 hands as a bare minimum, 100,000+ for a result you can lean on, and 250,000–500,000 before a win rate is genuinely nailed down.
Every poker win rate is an average pulled from a sample, and every average carries a margin of error. The fewer hands behind it, the wider that error — to the point where a few thousand hands can show a confident-looking win rate that’s really just noise. This is the single most misunderstood number in poker, and the one most apps are happiest to gloss over.
Why small samples lie
No-Limit Hold’em has a standard deviation of roughly 100 bb/100 for online 6-max play — a measure of how violently results swing around their true average. Run that through the math of a confidence interval and the picture is stark. Over 10,000 hands, the 95% range around your observed win rate spans nearly ±20 bb/100. In plain terms: a player who’s genuinely break-even can easily show +10 bb/100 over 10k hands, and a real winner can show a loss.
It gets worse for our egos. Even a true 5 bb/100 winner — a strong rate — has roughly an 11% chance of being down after 100,000 hands. Variance doesn’t politely cancel out just because you’ve put in volume.
What each sample size actually buys you
- 10,000 hands: a fun number to look at and nothing more. The error bars swallow it whole.
- 50,000 hands: the rough floor where trends start to mean something. Still a wide range — expect ±9 bb/100 or so at 95%.
- 100,000 hands: a result you can reasonably lean on. The 95% range tightens to about ±6 bb/100.
- 250,000–500,000 hands: high confidence. This is where serious players consider a win rate genuinely established.
Plug your hands and win rate into the win-rate confidence calculator and watch the range shrink as the sample grows. It’s the fastest way to feel how little a small sample really says.
Live players: it’s even slower
Live poker runs around 25–30 hands an hour. Hitting even 50,000 hands means well over a thousand hours at the table — years for most recreational players. That’s not a reason to despair; it’s a reason to stop treating a few dozen sessions as proof of anything, and to judge your game on process rather than a noisy headline figure.
What to do with this
Keep logging, and read your win rate as a range, not a verdict. Until your sample is large, the honest statement isn’t “I win at 6 bb/100” — it’s “my results are consistent with anything from a small loss to a big win, and I need more hands to narrow it.” That’s precisely why StackWise withholds a headline win rate until the data can support it, and shows a confidence range and sample-size label in the meantime.
A tracker that won’t lie about your sample.
StackWise shows a 95% confidence range and a sample-size label on every rate, and only reveals the headline number once your data earns it. Free to start, no account.
