Poker variance explained: downswings and standard deviation

Updated 12 September 2026

Why a winning player can lose for months, what standard deviation actually measures, and how big a downswing has to be before it stops being normal. Cash games, with every number simulated rather than asserted.

The short answer. Variance is the spread of your results around your expected results. It is measured as standard deviation in big blinds per 100 hands, and for online 6-max no-limit Hold’em it is about 100 bb/100 — twenty times larger than a strong 5 bb/100 win rate. That gap is why a good player’s graph looks like a bad player’s for months at a time: over 100,000 hands the typical worst drop from a peak is around 22 buy-ins, and one run in ten falls 36 buy-ins or more.

What variance actually is

Your poker graph has two ingredients. The first is your win rate: the money you expect to make per hundred hands, written in big blinds as bb/100. The second is variance: the randomness around that expectation, which comes from the cards and from your opponents’ choices rather than from anything you did.

Variance is usually quoted as standard deviation — the typical distance between a result and the average result, in the same unit as your win rate. If your standard deviation is 100 bb/100, then over any single 100-hand block your result will typically land within about 100 big blinds of your expectation, and occasionally much further. The important thing is not the definition but the scale: for most cash-game players the standard deviation is fifteen to thirty times the win rate.

There is one saving grace. Expected profit grows in a straight line with the number of hands, but the swing around it grows with the square root. Play four times as many hands and you expect four times the profit with only twice the uncertainty. That is the whole reason volume matters, and the whole reason nobody can judge their own play from a weekend.

Standard deviation by format

These are the ranges a tracker typically reports. Your own figure moves with your style: more three-betting, more bluffing and more big pots push it up; a tight, straightforward game pulls it down.

FormatTypical σ (bb/100)Use this if unsureHands to know your win rate within ±2 bb/100
Online 6-max NLH75120100960,400
Online full ring NLH608070470,600
Live NLH7010085693,900
PLO 6-max1201601401,882,400

Two things jump out. Full ring is the calmest no-limit game, because you are in fewer pots and fewer of them get big. Pot-limit Omaha is a different sport: at σ 140 the same win rate needs roughly twice the bankroll and takes twice as long to prove. And the last column is the one nobody wants to read — even the calmest format needs hundreds of thousands of hands before your win rate is known to within a couple of big blinds.

How big a downswing is normal?

This is the question people actually come here for. The table below simulates 100,000 hands at σ 100, 1,200 times over, at three win rates: a small winner (2 bb/100), a strong online winner (5 bb/100) and a live-style crusher (10 bb/100). Everything is measured in buy-ins of 100 big blinds.

Win rateTypical worst drop from a peak1 run in 10 falls1 run in 100 fallsChance of a 20 BI dropChance of a 30 BI dropChance of losing over 100k hands
2 bb/10028 BI49 BI71 BI83%44%26%
5 bb/10022 BI36 BI53 BI61%21%5.7%
10 bb/10016 BI24 BI35 BI26%3%<0.1%

Drop from a peak means peak-to-trough: the fall from your best point to your worst point afterwards. It is always larger than the amount you end up down, which is why downswings feel worse than the final damage suggests.

Read the middle row carefully, because it is the one most online players are in. A 5 bb/100 winner — a genuinely good player — has a 61% chance of a 20 buy-in downswing inside 100k hands and a 21% chance of a 30 buy-in one. Those are not disasters. They are the normal texture of a winning career, and they are the reason bankroll management exists.

Notice what happens as the win rate rises. The swings do not get smaller — the standard deviation is identical in all three rows — but the drift pulls you out of them faster, so the same randomness produces shallower holes. Edge does not remove variance; it shortens it.

What it looks like

Ten simulated careers for the same 5 bb/100 player over the same 100k hands. One line ends far above expectation, one ends well below, and several spend long stretches looking like a losing player.

-0.050100050k100kBI
  • Expected value
  • Middle 70%
  • Middle 95%
  • 10 sample runs
Buy-ins up the side, hands along the bottom. The green line is the expected result; the shaded bands hold the middle 70% and 95% of outcomes. Every grey line is the same player.

Why your graph cannot tell you if you are good

The chance of being ahead depends as much on how long you have played as on how well. Here is the probability of showing a profit at σ 100, by win rate and sample size.

Hands2 bb/1005 bb/10010 bb/100
10k58%69%84%
25k62%79%94%
50k67%87%99%
100k74%94%>99.9%
250k84%99%>99.9%
500k92%>99.9%>99.9%

A 2 bb/100 winner — a real, money-making edge at most low stakes — is still behind after 50,000 hands a third of the time. Run the same logic backwards and you get the number that ends most arguments: at σ 100 it takes about 3,841,500 hands to pin your win rate down to within ±1 bb/100 at 95% confidence, 960,400 for ±2, and 426,900 for ±3.

So a 20,000-hand sample is compatible with being a small winner and with being a small loser at the same time. That is not a reason to ignore your results; it is a reason to judge your play by your decisions, which give you feedback immediately, instead of by your graph, which does not.

Variance or a leak?

The practical version of all of this is one question: is this stretch of bad results the cards or me? You can answer it properly rather than by feel. Enter your result and your hand count in the variance calculator and it places you in one of three brackets.

  • Inside the middle 70% of outcomes. Nothing to explain — this is what your win rate looks like from the inside. Keep playing, keep studying.
  • Outside 70%, inside 95%. A cold run. Variance still covers it, but this is the right moment to review hands rather than simply wait.
  • Outside the 95% band. One result in twenty lands here by chance, so it is not proof — but a leak is now the simpler explanation, and you should go looking for it.

The reason the distinction matters is that the two situations demand opposite behaviour. Variance punishes players who rebuild their strategy to chase it; leaks punish players who sit and wait for them to pass. Knowing which one you are facing is worth more than any single strategic adjustment.

The emotional part is also maths. Variance does its real damage through tilt: the downswing costs you 20 buy-ins, and the way you play during it costs you another ten. Knowing in advance that a 20 buy-in drop is a 61% event for a good player is not consolation — it is preparation.

Related

Variance you cannot control. The decisions inside it you can — drill the spots where the money actually goes.

Ready to train?

Play real spots, see the solver’s answer and what your choice cost. Free to start, no real-money play, 18+.

iPhone: coming soon

Frequently asked questions

What is variance in poker?

Variance is the spread of results around your expected result. Your win rate is the slope of your graph; variance is everything that makes the graph wobble around that slope. It comes from the cards, not from your play, and it is the reason a good player can lose for months and a bad player can win for weeks. In numbers it is usually quoted as standard deviation in big blinds per 100 hands.

What is a normal standard deviation in poker?

For no-limit Hold’em cash games, roughly 75–120 bb/100 online 6-max, 60–80 bb/100 online full ring, and 70–100 bb/100 live. Pot-limit Omaha is much higher at around 120–160. Your own number depends on your style: more three-betting, more bluffing and more big pots all push it up. Any hand tracker reports it next to your win rate.

How big a downswing is normal?

Bigger than almost anyone guesses. Over 100,000 hands at 5 bb/100 with a standard deviation of 100, the typical worst drop from a peak is around 22 buy-ins, one run in ten falls 35 buy-ins or more, and one in a hundred falls over 50. A 20 buy-in downswing is not a sign that anything is wrong; it is what a winning player’s graph looks like from the inside.

Why do I keep losing with the best hand?

Because being a favourite is not the same as winning. A hand that is 70% to win loses three times in ten, and over a few hundred such spots a stretch of losing most of them is entirely ordinary. The useful question is never "did this hand win" but "was this the right decision given everything I knew". Outcomes are noisy; decisions are not.

How many hands before my results mean anything?

At a standard deviation of 100 bb/100 it takes roughly 3.8 million hands to know your win rate within ±1 bb/100 at 95% confidence, 960,000 for ±2, and 430,000 for ±3. Practically, 100,000 hands tells you whether you are in the right ballpark and nothing finer. This is why studying beats results-watching: your decisions give you feedback immediately, your graph does not.

Is variance the same as bad luck?

Variance is the mechanism; bad luck is the half of it you notice. The same distribution that produces a 30 buy-in downswing produces a 30 buy-in heater, and players remember those differently. If you are tempted to describe your results as cursed, run them through the calculator — most "impossible" runs sit comfortably inside the normal range.

Does variance go away if I play more hands?

The total swing in big blinds grows as you play more, but it grows with the square root of the hands while your expected win grows in a straight line. So variance never disappears in absolute terms, but it shrinks relative to your profit. Over 10,000 hands the noise swamps the edge; over 500,000 hands the edge wins the argument. Volume is the only cure.

Does tournament variance work the same way?

No, and we do not model it here. Tournament results are skewed rather than bell-shaped: most of your profit comes from rare deep runs, so the normal-distribution maths on this page would understate both the swings and the bankroll you need. The calculators on this site are cash-game only.