At its core, variance is the natural scatter of outcomes around a true long-run edge. Because each hand is decided partly by chance, a player's results bounce above and below their real win rate from session to session. Over a small number of hands those bounces dominate the picture entirely; the underlying skill is there the whole time, but it is buried under noise. This is the single most misunderstood feature of the game.
01Why short-term results deviate from skill
The edge a strong player holds is small compared with the size of the swings around it. A respectable cash-game win rate amounts to a few big blinds per hundred hands, yet the outcomes of those hundred hands can range across hundreds of big blinds in either direction. When the signal is that small and the noise that large, short samples are governed by luck, not ability.
This is why a handful of sessions reveal almost nothing. The same final result can describe a skilled player on a cold run or a weak player on a heater, and only the size of the sample separates the two. The edge that variance is busy hiding is the poker win rate, and it only surfaces once enough hands have been played for the swings to start cancelling out.
| Sample size | What the results can still look like |
|---|---|
| 5,000 hands | Mostly noise - heavy swing potential, almost no signal. |
| 20,000 hands | An early hint of an edge, but still very volatile. |
| 50,000 hands | More useful, yet still variance-heavy and easily misleading. |
| 100,000 hands | A stronger picture of the true win rate starts to form. |
| 250,000+ hands | A much more reliable long-run read of the real edge. |
02Standard deviation: measuring the scatter
The scatter itself can be measured, and the usual tool is standard deviation - a figure that captures how widely results spread around the average. In cash games it is quoted in big blinds per hundred hands, just like the win rate, which allows a direct comparison between the size of a player's edge and the size of their swings.
That comparison is sobering. The standard deviation of a typical cash-game style is many times larger than the win rate it surrounds, which is the mathematical reason short-term results are so unreliable. The bigger the standard deviation relative to the edge, the more hands it takes for the true rate to emerge - and tournaments, with their top-heavy payouts, carry far larger swings still.
A way to hold both numbers in mind: the win rate is the slope of the long-run graph, and the standard deviation is how far the line wobbles around that slope. A small slope under a large wobble can point clearly downhill for a long time while still trending up overall.
Every point on this line is the same skilled, winning player. A real edge can still dive deep below break-even and stay there for a long, painful stretch - and only volume eventually pulls it back up where it belongs.
03How big a normal downswing really is
Because the swings are so large relative to the edge, normal downswings are far deeper and longer than intuition suggests. A clearly winning cash-game player can lose across tens of thousands of hands without anything being wrong with their game, and a tournament player can run dozens of buy-ins below where they should be while still being a long-term winner. These are not disasters - they are the expected texture of a high-variance pursuit.
Mistaking a normal downswing for a broken edge is one of the most expensive errors in poker, because it pushes players to abandon good strategy or move stakes at exactly the wrong moment. The healthier read is to expect the swings, size the bankroll for them, and judge results only over a sample large enough to mean something.
| Situation | What it feels like | What it may actually be |
|---|---|---|
| Losing 10 buy-ins fast | "I'm playing terribly" | Could still be normal variance |
| A bad month after a good run | "My edge is gone" | The sample may simply be too small |
| Big all-in losses in clusters | "I always lose flips" | Random clustering, not proof of anything |
| Flat results over medium volume | "I'm break-even now" | A real edge can still be hidden by noise |
04What variance demands of the player
Variance has two practical consequences. The first is financial: the bankroll must be deep enough to outlast the swings, which is precisely why the buy-in counts in poker bankroll management look so conservative. A thin roll exposed to large variance can go broke even when the edge is real. The second is emotional: a player has to act on the long-run value of their decisions rather than on the short-run results those decisions happen to produce.
- A bankroll built for swings. Deep enough buy-in counts so a normal downswing never threatens going broke.
- Patience over meaningful volume. Trusting the edge across hundreds of thousands of hands, not judging it by a week.
- Emotional stability in downswings. Not tilting, quitting or jumping stakes the moment the graph turns ugly.
- Decisions judged by EV, not by one session. Measuring the line by its long-run value, never by the last card.
- Honest review instead of panic adjustments. Checking whether the play was sound before blaming - or rebuilding - the strategy.
That second point is where variance connects to the idea of expected value. A correct decision can lose money on the night and remain correct, because its merit lives in the average outcome over many repetitions, not in this one. Reading results through that lens - value over time, not outcome on the day - is what keeps variance from steering a winning player into losing habits.