Poker Variance and Downswings

I lost for two straight months once and spent the whole time quietly convinced I had gone bad at the game. I hadn't. I had run into the most ordinary thing in poker, and I treated it as a verdict on me instead of what it was — weather. That gap, between what my results were doing and what I actually was, is the thing this whole piece is about. It has a name. The name is variance, and almost everyone who plays misunderstands it in the exact direction that hurts them most.

What variance actually measures

Start with what it is, plainly, because the word gets thrown around like a curse. Variance is not bad luck. It's not the universe being unfair to you specifically. Variance is just the spread — how far your real-world results swing around your true winrate, the rate you'd earn if you could play forever. You have some honest number buried in you: maybe you win 5 big blinds per hundred hands over an infinite sample. You will almost never see that number. What you see instead is a session up forty buy-ins, a week down twelve, a month that looks like a heart monitor. The true rate is the quiet line underneath. Variance is everything bouncing around it.

The standard way to put a size on that bouncing is standard deviation, measured in the same units as your winrate — big blinds per hundred hands. Your winrate says which direction the line drifts over time. Your standard deviation says how violently it shakes on the way there. In heads-up no-limit, where you play every single hand and the pots get big, standard deviation usually runs somewhere around 100 to 130 bb/100. That's high, and it's high for an honest reason: more hands played, more all-ins, more swings. The number isn't a flaw in the game. It's the texture of it.

Why a real winner still has long losing stretches

Here's the part that took me an embarrassingly long time to feel rather than just nod at. A genuine, established winner — someone whose true rate is positive and would be confirmed over a lifetime — will still spend big chunks of that lifetime losing. Not because they're playing badly. Because the swing is simply larger than the edge over any short stretch.

Think about the two numbers side by side. Your edge accumulates with the hands you play — win 5 bb/100, play a thousand hands, you've earned fifty big blinds on average. But the swing doesn't grow at the same speed. It grows with the square root of your sample, which is much slower. So in the early going, the swing towers over the edge. Over a few thousand hands, a 5 bb/100 winner is drowning in noise that's many times the size of what they're actually earning. The edge is real, but it's a whisper, and variance is shouting over it. Only as the sample stretches into the tens and hundreds of thousands does the whisper start to win out, because the edge keeps climbing in a straight line while the noise crawls up underneath it. The losing months aren't evidence against the winrate. They're what a small edge buried under a large standard deviation is supposed to look like from the inside.

How sample size shrinks the luck

So the cure for not knowing whether you're good is hands. Lots of them. Not because a bigger sample changes your true rate — it doesn't, your rate was always whatever it was — but because a bigger sample shrinks the gap between what you've measured and what's true. The measured rate is a guess, and the guess gets sharper the more hands feed it.

The trap is how slow that sharpening is. Because the noise falls with the square root of the sample, getting twice as confident takes four times the hands. Quadruple your sample, you only halve the uncertainty. This is why a hot weekend tells you almost nothing, and why even a six-figure sample leaves a winrate fuzzier than people expect. You can grind a hundred thousand hands and still not be sure, to within a couple of big blinds, what your real edge is. That's not a failure of effort. It's the arithmetic, and once you see it you stop reading short results as character judgments.

What a downswing of N buy-ins actually implies

Now the question everyone actually asks at the bottom of the swing: I'm down fifteen buy-ins, what does that mean? Honest answer — much less than it feels like. A downswing's depth is governed by your standard deviation, and at heads-up numbers, drawdowns that look catastrophic are routine for players who are genuinely beating the game. A run of bad cards plus a stretch of correct decisions that simply didn't get paid will dig a hole that has nothing to do with your skill changing.

What a downswing does not do is overturn a well-established winrate on its own. If you've got a real edge and a real sample behind you, a fifteen-buy-in slide is the swing doing exactly what swings do, not a signal that you've gone bad. The danger isn't the depth — it's whether the depth exceeds your cushion. A downswing only ends your career if it runs your bankroll to zero before it reverses, which is why the real defense against variance isn't playing scared, it's holding enough buy-ins that the ordinary swing can't bust you. That's the entire logic of bankroll management: keep enough behind you that variance is allowed to be variance.

A worked example

Let me put real numbers to it. Say you've tracked a 5 bb/100 winrate over 100,000 hands, with a standard deviation of 110 — a believable heads-up profile. On average you're up 5,000 big blinds. Good. But the swing around that expectation over the same sample is roughly plus or minus 3,500 big blinds. Sit with that. The noise is nearly the size of the entire result. Your honest 95% range on your true winrate runs from roughly −2 bb/100 to about +12 — meaning even after a hundred thousand hands, you haven't pinned your real edge down to within about seven big blinds either way. On a sample this size you can't even fully rule out that you're a small loser running hot. You're almost certainly a winner; you just can't prove how much of one yet.

I built the Variance Calculator so you can stop estimating this in your head. Feed it your winrate, your standard deviation, your sample, and your bankroll, and it hands you the spread on your results, an honest confidence interval on your true rate, the odds you're a long-term winner at all, and your risk of ruin at the bankroll you're carrying. It's the difference between feeling doomed at the bottom of a swing and actually knowing whether the swing means anything.

What variance is really teaching you

Here's where this stops being math and starts being the only thing that matters. Variance forces a separation most players never make: the quality of your decision is not the same as the quality of your outcome. You can play a hand perfectly and lose it. You can play it terribly and win. Over a night, over a week, over a downswing, the cards are loud enough to sever the two completely, and if you judge yourself by results in that window you will reward your worst habits when they happen to win and punish your best ones when they happen to lose.

So you have to judge the decision by what you could reasonably have known when you made it — not by the card that came after. That's the whole discipline, and variance is what makes it necessary. You're responsible for the quality of your read, your sizing, your fold, given the information in front of you. You are not responsible for the river. The player who internalizes that can take a fifteen-buy-in beating and keep making the right call into it, because they know the call and the result were never the same thing. The one who can't will tilt off their edge chasing a number that was always going to swing. None of this means poker stopped paying — it still pays, and it pays exactly the people who can tell a good decision from a good night apart. Variance is just the tax you accept for the privilege of being right slowly.