Poker Variance and Downswings: The Complete Guide (2026)

Poker Variance and Downswings: Why Winning Players Run Bad (2026)

Take a NL50 6-max regular who has logged two years of solid, winning poker with a legitimate 5 bb/100 win rate. Real money in the bank. Then, over six weeks, he drops 40 buy-ins. Two thousand dollars gone, his graph looks like a ski slope, and there's one thought on loop: "am I broken, or is this just not my month?"

The answer is almost always the same, and it's uncomfortable precisely because it's so mundane: variance. Not punishment, not comeuppance, not "poker is dead." Just the math baked into the game from hand one. The bad news is it can throw anything at you over any stretch. The good news is you can quantify it, project how wide the swings get, and size your bankroll so a cold streak doesn't knock you out of the game. Thirteen minutes from now you'll understand your swings better than 90% of the players at your tables.

TL;DR: Variance in 30 Seconds

Variance is the gap between what your game deserves and what you actually receive over a given sample. Measured through standard deviation (SD) in bb/100. Here are the key numbers we'll come back to throughout the guide.

Key Numbers Worth Memorizing

Three figures answer half of all variance questions. SD in hold'em typically runs 80–110 bb/100. A 20+ buy-in downswing for a winning player is routine, not a red flag. A win rate only becomes genuinely reliable somewhere north of 500,000 hands. Keep those in your head and you won't panic at buy-in number 30 on the way down.

What Variance Actually Is

Variance isn't "bad luck" or some mystical force. It's an unavoidable property of any game where decisions happen under incomplete information and outcomes depend on the cards. You can get the money in as a mathematical favorite a thousand times in a row and still lose half of them over a short stretch. In the long run, edge collects. In the moment, the cards decide.

A simple image: flip a coin 50 times. You know roughly 25 heads should come up. In practice you'll see 21 one time, 29 the next. That deviation from the expected 25 is variance. The more flips, the closer the heads percentage drifts toward 50%. Poker works exactly the same way, except you need hundreds of thousands of "flips," and the "coin" is your win rate competing against table variance.

Variance vs. Standard Deviation

These two terms get mixed up constantly, but the distinction is straightforward. Variance is a mathematical quantity: the average squared deviation from expectation. Standard deviation is its square root. Formally:

$\text{SD} = \sqrt{\text{variance}}$

In practice, nobody works with raw variance because its units (bb squared per 100 hands) are impossible to reason about intuitively. Everyone uses SD in bb/100 because it lives in the same units as win rate, making the two directly comparable. When your tracker shows "SD 92 bb/100," that's the practical measure of your swings. Everything from here on deals in SD.

Why Downswings Feel Like They Last Forever

The core reason a cold streak feels endless, and it's something most people never explain properly. Your expected profit grows linearly with the number of hands: play twice as many, earn roughly twice as much on average. But the spread of possible outcomes only grows as the square root of hands played.

$\text{SD}_{N} = \text{SD}_{100} \times \sqrt{N/100}$

In plain terms: play 100 times more hands and your spread only grows by a factor of 10. Profit eventually outpaces spread, but slowly. That's why "the long run" is actually long. Over a 50,000-hand stretch, spread comfortably swamps your win rate, and your graph can sit below zero for weeks while you're playing just as well as you did during your winning months.

Standard Deviation in bb/100

SD tells you how widely your results scatter around your win rate. Low SD means a smooth graph with rare dips. High SD means a roller coaster: bigger upswings, deeper downswings. High SD isn't inherently good or bad. It describes your style and format. But it directly dictates how much bankroll you need.

Typical SD by Format

These ranges come from data across two major industry simulators and align closely with each other. Use them as a benchmark, but remember: your personal SD only shows up in your own tracker over a large sample.

The pattern is clear: fewer players at the table and a wider playing style both push variance higher. PLO swings nearly twice as hard as hold'em because Omaha hand equities run close and money goes in more often on the flop and turn. Heads-up is also high variance: you play every hand, spots come up constantly, and pots are large.

How Playing Style Shifts Your SD

The same format can produce very different SD numbers depending on how you play. A tight-passive nit waiting for the nuts will land near the bottom of the range. A loose-aggressive maniac who 3-bets light and calls down bluffs will push SD toward the top, or beyond it. Neither is "correct." It simply means the higher your SD, the more buy-ins you need to survive the swings, even if your win rate is identical.

How Deep Downswings Actually Get

This is the question most people open articles like this to answer. Two things determine the answer: your win rate and your sample size. The higher the win rate, the less frequent and shallower the dips. The longer the run, the deeper the worst stretch inside it can go.

Downswing Depth by Win Rate

The table below shows the probability of hitting a downswing of a given depth at least once over a long stretch. Numbers assume SD of 90 bb/100, the middle of the hold'em range. These are simulation estimates, order-of-magnitude figures, not guarantees.

Read the top row again. Even a solid 5 bb/100 winner hits a 10+ buy-in downswing in roughly 60% of long stretches. A 20+ buy-in shot happens to one in four players like that. A marginal 2 bb/100 player hits a 30 buy-in hole one time in four. Check your own numbers in the simulator. Don't guess.

Normal Downswings by Format

How deep a downswing goes depends heavily on where you play. Live cash is the calmest. Tournaments are the most brutal.

In dollar terms at NL50: a 10 buy-in hole is minus $500 over a few weeks (happens often), a 30 buy-in hole is minus $1,500 over a couple of months (less common), a 50 buy-in hole is minus $2,500 over four-plus months (rare, but a real scenario). The numbers look scary on paper. With a properly built bankroll, none of them kill you.

Why 100+ Buy-In Downswings in MTT Are Normal

Tournaments are their own planet. You pay a buy-in every time and only collect meaningful money when you go deep, which doesn't happen often. Hundreds of tournaments can go by between significant scores. A 100–200 buy-in downswing in MTT isn't a disaster or a sign of poor play. It's just normal tournament life. Tournament variance is measured not in bb/100 but in ROI spread (typically ±150–260% for large fields) and buy-ins. Don't mix cash game SD with tournament variance. They're different scales entirely.

Want to see your own personalized spread rather than table averages? Run your own variance simulation: plug in your win rate, SD, and number of hands, and get the result range and typical downswing depth calibrated to your exact numbers.

Confidence Intervals: Your Real 95% Corridor

A downswing measures the deepest single drop. A confidence interval is something else: where your result can actually land after N hands. It's the most honest picture of how slowly your win rate reveals itself.

How to Read a 95% Interval

The core formula of this guide, and it's simpler than it looks:

$\text{Win rate} \pm 1{.}96 \times \frac{\text{SD}}{\sqrt{N/100}}$

Take your win rate and add or subtract a corridor that narrows as sample size grows. The 1.96 multiplier is the standard 95% boundary (many people just round it to 2). The table below is calculated for a win rate of 5 bb/100 and SD of 90. Check the math yourself if you want; it's straightforward.

Look at the 100,000 row. A genuine 5 bb/100 winner can still show almost zero after a hundred thousand hands. A hundred thousand hands is a serious sample, yet the win rate still swings across an 11 bb/100 range. That's why you can't draw conclusions about your skill level from a couple of months of play.

Can a Winning Player Finish a 100k Stretch in the Red?

Yes, and it's not rare. The probability of ending a stretch in the red is calculated through the normal distribution using $\Phi(-\text{Win rate} \times \sqrt{N/100} / \text{SD})$. Here are the recalculated figures for SD 90.

A borderline 2 bb/100 winner goes into the red in almost one out of every four 100k stretches. Even a solid 5 bb/100 winner finishes in the red over 100k hands around 4% of the time.

One important note on these numbers. A popular variance calculator puts the probability of a losing 100k stretch at around 11% for a 5 bb/100 player. That's inflated. The correct calculation through the normal distribution gives 3.9% (check: $z = 5 \times \sqrt{1000} / 90 = 1{.}76$, and $\Phi(-1{.}76) = 3{.}9\%$). That's nearly a threefold discrepancy. We use recalculated figures rather than numbers copied from someone else's site, because when it comes to bankroll questions, being off by a factor of three means seriously misjudging your risk in either direction. If you see scarier percentages somewhere else, run the formula yourself.

Running Bad or Playing Bad

The most agonizing question mid-downswing: is this bad luck, or have I actually lost my edge? The difference matters, because the fix is completely different. Variance you wait out with the right bankroll. A leak you have to find and plug, otherwise the losses never stop.

How Many Hands Until Your Win Rate Means Something

Short answer: a lot more than your gut tells you. At 10,000 hands your graph is basically a random walk. At 50,000 you can see a trend, but the corridor is still ±8 bb/100. At 100,000 the picture starts to clear, but you're nowhere near proof. A reasonably reliable read on your win rate takes around 500,000 hands, and real confidence takes a million. Until that point, any boastful or panicked conclusion drawn from your graph is self-deception. Your sample is lying to you, and it will keep lying for a long time.

Variance or Leak: The Diagnostic

Run your last few sessions through this table. If the signals land in the left column, you're running bad. If they pile up on the right, you have a leak and need to study right now.

An honest review of ten big losing pots answers the question almost every time. Did you get the money in as a favorite and lose on the runout? Variance. Sit it out. Did you call down second pair three streets against obvious strength? That's not bad luck, it's a decision, and it needs fixing.

Variance and Your Bankroll: Risk of Ruin

Most players get as far as "you need a bankroll" and stop. Bankroll size flows directly from your SD and win rate through Risk of Ruin, and you can actually calculate it.

Risk of Ruin in Plain Terms

Risk of Ruin (RoR) is the probability of losing your entire bankroll before your edge pulls you into profit. The classic continuous approximation:

$\text{RoR} \approx e^{-2 \cdot B \cdot \text{Winrate} / \text{SD}^2}$

Here B is your bankroll in bb, with win rate and SD both in bb per 100 hands. Rearranging for a target risk level, you get the required bankroll:

$B = -\ln(\text{RoR}) \times \frac{\text{SD}^2}{2 \cdot \text{Winrate}}$

Higher SD and lower win rate mean you need a bigger cushion. Double your win rate and you need half the bankroll. Let SD grow and the required bankroll scales with the square of that growth, which is fast. That's exactly why PLO demands several times more buy-ins than hold'em at the same win rate.

Formula vs. Simulator Numbers

There's an honest wrinkle here that rarely gets discussed. The closed-form formula for 3 bb/100 at SD 90 with a 5% risk target gives roughly 40 buy-ins. Simulators, including ours and industry-standard tools, recommend 50 to 75 buy-ins for the same parameters. Who's right?

Both. They're just answering different questions. The formula is the theoretical minimum, assuming you know your exact win rate and play forever at the same stake. Simulators build in uncertainty: your win rate is only an estimate, you may be underestimating your SD, and they add a buffer. The practical takeaway: the formula gives you the floor (around 40 buy-ins), while a sensible working bankroll sits at 50 to 75. Playing with fewer than 40 is mathematically dangerous. More than 75 is generally overkill, unless you're chasing near-zero risk.

Bankroll by Win Rate and SD

The closed-form formula recalculated for SD 90. This is the theoretical minimum; add a buffer in practice.

The logic is immediate: a thin win rate with high SD demands a massive cushion, while a strong edge lets you play on a smaller bankroll. Want an exact number for your own parameters? Model your risk of ruin in the simulator or size your stack with the bankroll calculator. We've gone deep on the math of ruin in a dedicated guide to risk of ruin and bankroll, and covered the fundamentals in the article on bankroll management in poker. If you size your bankroll in buffer units against drawdown, check out the breakdown of how to calculate bankroll units.

How to Reduce Variance (and How to Survive It)

You can't eliminate variance. It's baked into the game. But you can shrink its range and build your own resilience to it. Both work at the same time.

Table Selection, Number of Tables, and Edge

Three levers actually move your SD down. First: table selection. A soft table full of recreational players gives you more edge, and more edge cuts swings relative to profit. Second: number of tables. Every extra table beyond what you can genuinely handle drops decision quality and inflates variance through mistakes. Play fewer tables and play them well. Third, and most powerful: edge itself. The higher your win rate, the shorter your downswings in buy-ins and the faster profit outruns variance. Working on your game is the single best anti-variance tool you have.

Another way to offload some variance is staking. By giving a backer a share of your swings, you smooth out your personal graph at the cost of a portion of your profits. How that math works is covered in the staking calculator. The connection between edge and the long-run win rate that tames variance comes through clearly when you look at pot odds and implied odds. For tournament-specific variance, which sits at the high end of the scale, see the breakdown of ICM in poker.

Mental Game During a Downswing

Variance gets compounded by something the formulas don't capture: tilt. A downswing by itself doesn't cost you anything beyond what the math already accounts for. Tilt does. You start calling wider, chasing losses, playing stakes that are too high for your emotional state. At that point variance gets amplified by actual leaks, and the hole deepens through your own doing.

What actually works: take breaks when you feel yourself boiling over; move down in stakes if your bankroll has dropped below a comfortable threshold (there's no shame in that, it's management); measure yourself by decision quality, not by results from a single session. Understanding that a 20-to-30-buy-in downswing is statistically normal takes away half the tilt on its own. You already know from the tables above that it's a routine event, not a verdict. To work through the emotional mechanics of one specific hand, check out the bad beat calculator, and we've covered win rate in bb/100 and the specifics of tournament format in the glossary.

Variance isn't your enemy. It's the price everyone pays so that weaker players keep sitting down believing today is their day. Your job isn't to beat variance, it's to survive it with the right bankroll and a clear head. The numbers above give you the map. The simulator plots the route for your specific parameters. After that it's volume and discipline.