Poker Variance Explained: Is This Downswing Normal?

Every poker player has the same conversation with themselves eventually. The graph has been going down for weeks, the hands all look like coolers, and the question underneath is not really about the cards. It is: am I actually any good at this?

The honest answer is that your results are far weaker evidence than they feel. Here is what variance actually does to a poker graph, and how to tell a normal downswing from a real problem.

What variance is

Variance is the spread of outcomes around your expectation. Two players with identical skill, playing identical hands, will end a month thousands of dollars apart purely because the cards fell differently.

The number that quantifies it is your standard deviation, usually expressed in big blinds per 100 hands alongside your win rate. Typical figures for online no-limit hold'em:

• 6-max cash — 80 to 110 bb/100
• Full ring — 70 to 90 bb/100
• Heads-up — 120 to 150 bb/100
• Pot-limit Omaha — 120 to 180 bb/100

Now look at those next to a good win rate. A strong 6-max regular wins perhaps 5 bb/100 with a standard deviation of 100 bb/100. The noise is twenty times the size of the signal.

That ratio is the whole story. Everything below is a consequence of it.

Why your graph tells you so little

The uncertainty in a measured win rate shrinks with the square root of the number of hands, which is much slower than intuition suggests. Quadrupling your sample only halves your error bar.

For a player whose true win rate is 5 bb/100 with a standard deviation of 100:

• After 10,000 hands — the 95% confidence interval runs from about -15 to +25 bb/100. Essentially no information.
• After 100,000 hands — roughly -1 to +11. Still includes losing.
• After 500,000 hands — roughly +2 to +8. Now you know something.

A genuine 5 bb/100 winner is still behind after 100,000 hands about 6% of the time. Not behind their expectation — actually losing money, over a sample most players never reach.

Turn that around and the implication is uncomfortable: a break-even player can be up several buy-ins after 50,000 hands with nothing unusual happening at all. Results confirm skill far too slowly to be used as feedback.

How big a downswing is normal

The question everyone actually wants answered.

For a winning 6-max player at 5 bb/100 with a 100 bb/100 standard deviation, over a few hundred thousand hands:

• Downswings of 10 buy-ins are routine — several per year.
• Downswings of 20 buy-ins happen to essentially every player eventually.
• Downswings of 30 buy-ins or more are uncommon but entirely within normal experience.

And for a break-even player, or a small winner at 2 bb/100, the figures are considerably worse, because there is less edge grinding against the noise.

The deeper point is that downswings are not a separate phenomenon that happens to you. They are what a random walk with a small upward drift looks like. A graph that only went up would be evidence of something wrong with the sample, not evidence of skill.

Running your own numbers

General figures are useful; your figures are better. A poker variance calculator takes your win rate, your standard deviation and a sample size and simulates thousands of possible careers, showing the range of outcomes you could reasonably expect.

Three things are worth reading off it:

The confidence interval. How wide the plausible range of true win rates is, given what you have observed. Usually wider than expected, and usually the most useful number on the page.

The worst downswing across the simulations. This is what you are actually preparing for. Seeing that a 25 buy-in downswing appears in a meaningful fraction of simulated careers reframes the one you are currently in.

The probability of being stuck. How often a player with your win rate is below break-even after this many hands. For most realistic win rates over realistic samples, it is not a small number.

Risk of ruin and bankroll

Variance has one consequence that is not psychological: with a finite bankroll, a bad enough run ends your ability to play.

Risk of ruin depends on three things — win rate, standard deviation and bankroll — and it falls off very quickly as the bankroll grows. For a 5 bb/100 winner with a standard deviation of 100:

• 20 buy-ins — roughly a 14% chance of eventually going broke
• 30 buy-ins — roughly 5%
• 50 buy-ins — under 1%

Halve the win rate to 2.5 bb/100 and every one of those numbers rises sharply: 30 buy-ins becomes over 20%. Bankroll requirements depend on your edge, not just on the stake.

Two notes that matter in practice. Risk of ruin is exactly 100% for a break-even or losing player given enough time, no matter how large the bankroll — which is why bankroll management cannot rescue a strategy problem. And tournament players need far larger bankrolls, often 200 buy-ins or more, because tournament variance dwarfs cash game variance: the top few finishes carry most of the prize pool, so you can play well and cash nothing for months.

Telling variance from a leak

The genuinely useful skill is distinguishing "I am running badly" from "I am playing badly", because they feel identical from the inside.

Signs it is variance:

• Your all-in equity graph runs well above your actual winnings
• Your statistics are stable and your decisions look right on review
• The losses are concentrated in coolers — set over set, flush over flush
• Different stakes and formats show the same pattern at the same time

Signs it is a leak:

• Your all-in equity graph tracks your results closely, or sits below them
• Your win rate has stepped down and stayed there since a specific change
• The losses cluster in one identifiable spot — big blind defence, river calls, one position
• Your session length has grown, or you have been playing after losing

The single most useful diagnostic is the all-in equity comparison. If you are consistently getting money in ahead and losing, that is variance, and it will correct. If you are getting money in behind, no amount of running good will save you.

One caveat worth stating, because it is where the diagnostic gets misused: all-in equity only covers hands that got all the money in. It says nothing about the far larger number of hands where you folded too much, called too light, or missed a value bet on the river. A player can run perfectly at showdown and still be losing steadily from decisions that never reached one.

There is also a failure mode where variance and a leak reinforce each other. A downswing prompts small compensatory changes — calling a little wider to "make something happen", playing longer sessions to get back to even — and those changes are real leaks that arrived because of the downswing rather than causing it. When you review, check whether your statistics changed during the losing stretch, not just before it.

Reviewing individual hands with a poker equity calculator settles the question directly: enter the spot, see whether you were ahead when the money went in. A run of losses where you were a 70% favourite each time is a story about the deck. A run where you were a 40% underdog is a story about you.

The mental side

Variance does most of its damage indirectly. The downswing itself costs you money once. The tilt, the stake-jumping and the abandoned strategy that follow cost you repeatedly.

Four things help more than anything else:

1. Judge decisions, not results. A correct call that loses is still a correct call. This is easy to say and genuinely hard to practise, and it is the entire discipline.
2. Set stop-losses before you sit down. Not because losing three buy-ins predicts losing a fourth, but because your play measurably deteriorates after a bad run.
3. Track sample size alongside results. "Down four buy-ins" means nothing. "Down four buy-ins over 8,000 hands" is well inside normal noise, and knowing that changes how it feels.
4. Keep enough bankroll that no session matters. Playing scared costs more than the variance you are afraid of.

What actually moves the needle

The uncomfortable conclusion of all this is that you cannot measure your own progress from your graph on any useful timescale. By the time your results are statistically meaningful, you are a different player.

So the feedback loop has to come from somewhere else: reviewing hands against a solver baseline, checking whether the money went in ahead, and confirming that your decisions were right regardless of how they turned out. That is the only signal available that is not buried under twenty times its own size in noise.

Two concrete places to start. Make sure the routine decisions are right — most of the ten most common beginner mistakes cost more per hour than any downswing. And get the arithmetic of calling right, because pot odds and why they matter is the one place where a small, permanent edge compounds across every hand you will ever play.

Variance decides what happens this month. It decides nothing at all about what happens over a career — provided you last long enough for the edge to show up.