Poker Variance Calculator

How much swing is normal? Monte Carlo your next 100,000 hands or 200 tourneys and see where you actually land — downswings, risk of ruin, and what your "Am I a winner?" math really says.

I win bb/100 over hands of
. SD 140 bb/100
020k40k60k80k100k Hands
Probability of profit
74.9%
Expected profit
+29.6 BI
95% confidence range
−53 BI … +117 BI
1,000 runs simulated
More analysis
Outcome distribution Median run: +29.7 BI

Sorted from best to worst. The median is the middle — the "typical" outcome — not the EV.

Top 1% luckbox
+128 BI
Top 5%
+103 BI
Top 25%
+59.5 BI
Median the typical run
+29.7 BI
Bottom 25%
−48 bb
Bottom 5%
−39.7 BI
Bottom 1% nightmare
−72.5 BI
Bankroll & risk of ruin

How deep can the cumulative profit line dip, and how big a starting bankroll do you need so it (almost) never dips past zero?

Risk of ruin
15.4%
over 100,000 hands

Required bankroll by risk tolerance

Click a row to re-run the sensitivity table below at that target.

If you're wrong about your winrate (5% RoR target)

ScenarioAssumed winrateBankroll needed
Your assumption3.00 bb/10097.9 BI
Slightly worse1.50 bb/100196 BI
Much worse0.75 bb/100391 BI
Breakeven0.00 bb/100∞

A small misread of your true winrate balloons the bankroll you actually need. If you're unsure, read the Am I a winner? card below.

Downswings

How often a deep downswing shows up, and how long losing stretches drag on, under +3 bb/100 · 100k hands · SD 140.

How deep, and how long

Probability of a downswing ≥
5 BI >99.9%
10 BI >99.9%
25 BI 89.3%
50 BI 28.8%
100 BI 1.0%
Probability of a slump lasting ≥
2.0k >99.9%
5.0k >99.9%
10k 99.6%
25k 85.7%
50k 42.1%

Peak-to-trough inside a window

Your worst drawdown inside any given stretch of hands — pick the threshold, see how often it shows up in a month / quarter / full sample.

Threshold
In any 8.3k-hand stretch
94.2%
chance of dropping ≥20 BI
In any 25k-hand stretch
96.6%
chance of dropping ≥20 BI
Over the full 100k hands
96.6%
chance of dropping ≥20 BI
Am I a winner? Bayesian credible intervals on your true skill

If your results so far are +3.00 bb/100 over 100k hands, what can you honestly claim about your true skill?

Probability you are actually a winning player
73.2%
Credible interval on your true winrate
ConfidenceLowerUpper
60%-0.90+5.92
75%-2.15+7.17
90%-4.15+9.17
95%-5.43+10.44
Hands needed until you know your winrate to…
within bb/100 at confidence 7.5M hands
Your year, month by month One randomly drawn year with lucky/unlucky months
Total for the year · 1,200,000 hands +348 BI
100,000 hands /
Jan
+39.2 BI
+0.2σ
Feb
−17.6 BI
-1.1σ
Mar
+18.4 BI
-0.3σ
Apr
+77 BI
+1.1σ
May
+4.5 BI
-0.6σ
Jun
+22.4 BI
-0.2σ
Jul
+17 BI
-0.3σ
Aug
+67.4 BI
+0.8σ
Sep
+6.5 BI
-0.5σ
Oct
+56.4 BI
+0.6σ
Nov
+15.8 BI
-0.3σ
Dec
+41.3 BI
+0.3σ
🔥 ≥+2σ blessed 🍀 ≥+1σ lucky ❄️ ≤−1σ unlucky 💀 ≤−2σ disaster
Best month
Apr
+77 BI
Worst month
Feb
−17.6 BI

A single year rolled from your exact distribution. Hit roll enough times and you'll see how wild a "normal" year can look.

Backing deal Simulate staker + player with makeup

Model a stake deal. The backer absorbs the losses; you split the profits at each chop event (after makeup clears).

Your EV per 100,000 hands
0 bb
Backer's EV
0 bb
% of sims ending in makeup
<0.1%
Avg makeup owed at end
0 bb
Avg worst makeup (any point)
0 bb
Backer risk of ruin
<0.1%

Backer absorbs losses one-for-one; takes 50% of cleared profit at each chop. Chops happen every 10 sessions (25,000 hands). Your EV is strictly less than solo play — but you play bigger without risking your own cash.

Zoom out — stake progression Auto move-up and move-down through the stake ladder

Play 2.5M hands, moving up when you clear the threshold and back down on downswings. Every setting below is editable — test your own assumptions.

Derived winrate per stake
PLO10 +9.0
PLO20 +8.0
PLO25 +7.0
PLO50 +6.0
PLO100 +5.0
PLO200 +4.0
PLO500 +3.0
PLO1K +2.0
PLO2K +1.0
PLO5K 0.0
PLO10K −1.0

What this calculator is showing you

Variance is the thing poker players blame and the thing poker players misread. This tool runs Monte Carlo simulations on your exact numbers — a thousand alternate futures where nothing changes except the cards — then shows you the shape of the distribution.

In cash mode, each 100-hand block is drawn from a rescaled Student-t(df) distribution — same mean as your winrate and same marginal SD as your input, but with heavier tails than a Normal because poker has coolers, runouts, and bad beats. PLO 6-max defaults to df = 5; NLH 6-max to df = 10. Override df in Advanced if you've fit your own data.

The chart is layered: the faint gold band is the 95% confidence interval, the darker band inside it is the 70% (where roughly two thirds of runs land), the solid gold line is expected value, and the thin cyan lines are 20 actual sampled paths. The brightest cyan is the luckiest of your trials; the red is the unluckiest.

In tournament mode, each tourney is a discrete outcome — either you bust (most common), min-cash, deep-cash, or win. We derive a finishing-rank distribution from the payout structure and your target ROI. Toggle "realistic finish distribution" in Advanced to weight finishes by skill (better players cash deeper more often than not).

Every other analysis on this page — outcome distribution, bankroll and risk of ruin, downswings, the Bayesian "am I a winner?" check, your year month-by-month, backing splits, and stake progression — is derived from the same underlying simulation. Hit the 🎲 Re-roll button beside the inputs to see a different draw. All your settings encode into the URL, so you can share a specific simulation as a link.

One small footnote on the "year, month by month" card: it treats the "hands" you typed up top as your monthly volume, then rolls 12 months from your distribution. So at 100k hands the year shows 1.2M played — what a real grinder's calendar year of that workload looks like, lucky and unlucky months and all.

None of this math tells you whether you should gamble. It tells you how much noise is normal around your true skill — and how long it takes before your results actually mean something.