Simulate 20 or 100 poker result paths with 50, 80 and 95% bands, probability of loss, drawdown scale, drift-equivalent hands and downloadable assumptions.
What does this result mean?
A poker variance simulator shows how many different result paths can come from the same assumed win rate and standard deviation. This version draws 20 or 100 reproducible paths, overlays 50, 80 and 95 percent bands, estimates the probability of finishing below break-even, and reports a positive-edge drawdown scale and drift-equivalent hand count. The lines are not predictions. They are examples under a fixed-edge normal model, designed to show why a good player can run badly and a short-term winner can still be unproven. Save or share the assumptions beside the result, then compare the output with your real session history and hand review before changing stakes.
Assumptions
Build the model
+2,500 BBExpected final result10.7%Probability of finishing behind+810 BBDrawdown scale (model)16200Drift-equivalent hands−1,444 BB95% final low+6,444 BB95% final high
Result paths and confidence bands
How the same edge can look completely different
PathsExpected50 / 80 / 95%
Final-result distribution
Where the model finishes
Accessible data table
Confidence bands through the sample
The table is the same model as the chart. It remains readable without colour, animation or a mouse.
Hands
Expected
50% range
80% range
95% range
Methodology
What the simulator is doing
Each path divides the selected sample into 50 equal blocks. For every block it adds the assumed win-rate expectation plus a seeded normal random draw scaled from the standard deviation. The seed comes from the visible inputs, so the same shared URL reproduces the same paths.
The expected result is win rate × hands / 100. Spread at each point is standard deviation × √(hands / 100). The 50, 80 and 95 percent bands use standard-normal critical values 0.674, 1.282 and 1.960.
The model assumes a fixed edge, independent blocks and a stable standard deviation. Real games change. Table quality, tilt, fatigue, stake mix and non-normal tails all move the result. A confidence band describes outcomes under the assumptions. It does not prove the assumptions are true.
The drawdown scale is standard deviation² / (2 × positive win rate). Under a fixed positive-drift Brownian model this is the expected all-time loss below the starting balance, not the expected maximum downswing from future peaks. The drift-equivalent hands divide this scale by the assumed edge; they are not a predicted recovery time. At a zero or negative edge both figures are unavailable.
Version 2.0.0. Model and copy by Tom Howcroft, reviewed 5 August 2026. Use it for study and planning, never as a profit promise.
Next step
Give your variance range some real context
A 95 percent range shows how widely results can vary under your assumptions. Keep logging real sessions in Overshove to build the history that helps you tell variance from a leak.
The app opens in a new tab. On the sign-in screen, choose "New here? Create an account". Calculator numbers are not transferred to the app. Enter your real session records there to start your log.
Tracking and calculations do not guarantee a profit or prevent losses.
How to use the answer responsibly
Treat the calculator as a planning tool, not a verdict on one session. A live poker input is usually an estimate: your win rate is noisy, your standard deviation changes by game type, and your future volume rarely matches the clean number you type into a form. The useful habit is to save the assumptions beside the result so you know what would need to change before the decision changes.
If the result pushes against what you wanted to do, do not smooth the inputs until the answer feels nicer. Make the conservative version first, using a lower win rate, a higher swing rate or fewer weekly hours. Then run the optimistic version. The space between those answers is the real decision area: where a shot needs a stop rule, where a move up needs more buy-ins, or where a good-looking hourly still needs a bigger sample.
Overshove calculators become more useful when the inputs come from your own logged sessions. Rake, tips, travel, hours, stake mix and session length all change the number. A generic benchmark can teach the shape of the problem, but a clean personal log gives the number you should actually act on.
Variance guide
Poker variance for live cash: reading the range in hours
The simulator above counts in big blinds per 100 hands. Live players count hours. This lesson converts between the two, reads the bands at the sample sizes live players actually reach, explains why tournaments need a different model, and shows how to estimate your own standard deviation from your session records.
What variance and standard deviation mean in big blinds per hour
Your win rate is your average result per hour. Your standard deviation (SD) is the typical distance between a single hour’s result and that average. Variance is the SD squared. Variance is the part that adds up from one hour to the next, which is why the spread of your total grows with the square root of your hours rather than with the hours themselves.
The simulator’s inputs are Win rate (BB/100), Standard deviation (BB/100) and Hands in sample. To move between hourly records and those inputs you need one assumption: how many hands you see an hour. Live full-ring games often deal around 25 to 30. This lesson uses 25.
You have
Simulator input
Conversion at 25 hands an hour
Win rate in BB per hour
Win rate (BB/100)
Multiply by 4
SD in BB per hour
Standard deviation (BB/100)
Multiply by 2
Hours played
Hands in sample
Multiply by 25
The win rate scales with the number of hands, so 100 hands (four hours) hold four times the hourly win. The SD scales with the square root, so four hours hold twice the hourly SD.
Read the other way, the Live NLH preset (6 BB/100, SD 100 BB/100, 50,000 hands) describes a player winning 1.5 BB an hour with an SD of 50 BB an hour, over 2,000 hours. In a £1/£2 game that is £3 an hour on average against a typical hourly swing of £100. At four hours a week, 2,000 hours is nearly ten years of play. If your game deals 30 hands an hour, 50,000 hands is about 1,667 hours instead. The poker variance guide covers the same ideas in more depth.
How to read the range at live sample sizes
Choose Live NLH, then set Hands in sample to 5,000. That is 200 hours at 25 hands an hour, or about a year at four hours a week. The simulator shows:
Expected final result: +300 BB.
95% final low and high: −1,086 BB and +1,686 BB.
Probability of finishing behind: 33.6%.
Under these assumptions, a player whose true edge is exactly 1.5 BB an hour finishes that year behind about one time in three. Nineteen years in twenty land between −1,086 and +1,686 BB, which at £2 a big blind is about −£2,170 to +£3,370. The expected +300 BB (£600) sits in the middle, and on its own it is the least informative number on the screen.
The narrower bands are useful for planning. The 50% band, from about −177 to +777 BB, is where the middle half of paths finish. The 80% band runs from about −606 to +1,206 BB. If a plan only works when you finish in the top half of that picture, it depends on luck rather than edge.
Switching between 20 and 100 Simulated result paths changes how many example lines you see, not the bands or the probability. Watch how many lines cross zero and come back.
Why short samples mislead
Expected profit grows in step with your hours, but the spread grows only with their square root. Early on, the spread dominates. Here are the Live NLH inputs at different sample lengths:
Live hours (25 an hour)
Hands in sample
Expected result
95% range
Probability of finishing behind
40
1,000
+60 BB
−560 to +680 BB
42.5%
200
5,000
+300 BB
−1,086 to +1,686 BB
33.6%
400
10,000
+600 BB
−1,360 to +2,560 BB
27.4%
1,000
25,000
+1,500 BB
−1,599 to +4,599 BB
17.1%
2,000
50,000
+3,000 BB
−1,383 to +7,383 BB
9.0%
Two consequences follow.
A good start shows little. Over 200 hours with an SD of 50 BB an hour, a player with no edge at all finishes within about 1,386 BB either side of break-even 95% of the time. About one break-even player in 13 would be up 1,000 BB (£2,000 at £1/£2) or more.
Small edges take a long time to see. For a measured win rate’s 95% band to clear zero, you need roughly (1.96 × SD ÷ edge)² hours. With an SD of 50 BB an hour, that is about 1,067 hours for an edge of 3 BB an hour and about 4,268 hours for 1.5 BB an hour. Halving the edge multiplies the hours by four. The win-rate confidence calculator runs this check on your own record; its inputs are already hourly (Win rate (BB/hr), Hours played and Std dev (BB/hr)).
Cash swings and tournament swings are different shapes
In a cash game, each hour’s result can land on either side of your average, and over many hours the total is roughly the bell shape this simulator draws. Tournament results are lopsided. Most entries lose the whole buy-in, a minority cash, and a few deep runs carry most of the profit. A model built on an average and a symmetrical spread understates how long a tournament player can go without a meaningful result.
A simple illustration: if you cash 15% of the time and each event is independent, the chance of ten events in a row without a cash is 0.85 to the power of 10, about 19.7%, or roughly one in five. Over twenty events it is still 3.9%. Each of those events costs a full buy-in.
Measure tournament results in buy-ins and ROI, not big blinds per hour. The MTT variance guide linked from this page explains tournament standard deviation by field size, and the tournament variance and bankroll simulator models a schedule from your buy-in, fee and number of events. Keep cash and tournament records separate so that neither distorts the other.
Estimate your own standard deviation from session records
You need each session’s result and its length. Live sessions vary in length, so a four-hour session and an eight-hour session should not count equally. This method weights each one by its hours:
Convert each result to big blinds by dividing by the big blind.
Work out your win rate: total big blinds ÷ total hours.
For each session, find the gap: result − (win rate × hours).
Square each gap and divide it by that session’s hours.
Add those figures, divide by the number of sessions minus one, and take the square root. That is your SD per hour.
Here it is with six invented £1/£2 sessions (£2 big blind):
Session
Hours
Result
In BB
Expected at 2 BB an hour
Gap
Gap² ÷ hours
1
5
+£200
+100
+10
+90
1,620.0
2
4
−£220
−110
+8
−118
3,481.0
3
6
+£150
+75
+12
+63
661.5
4
5
−£100
−50
+10
−60
720.0
5
4
+£260
+130
+8
+122
3,721.0
6
6
−£170
−85
+12
−97
1,568.2
Total
30
+£120
+60
11,771.7
Win rate: 60 ÷ 30 = 2 BB an hour. SD: 11,771.7 ÷ 5 = 2,354.3, and its square root is about 48.5 BB an hour. For the simulator at 25 hands an hour, that is a win rate of 8 BB/100 and an SD of about 97 BB/100. The Standard deviation input moves in steps of 5, so run it at 95 and again at 100.
Six sessions is far too few to trust either number; the method is the point. The SD estimate settles faster than the win rate, but recheck both as your log grows. Use each session’s cash-out minus buy-ins, which already reflects rake and any tips paid from your stack. Keep stakes separate, or convert every result to big blinds first. Record time at the table, not time at the venue.
What variance cannot tell you
Your edge. The simulator takes a win rate as an input and never estimates one. Its output is only as good as the number you typed.
Whether today’s game is the game in your record. A new venue, a deeper-stacked table or a straddle can change both your win rate and your SD.
How far a losing run can go. The bands describe final results under a normal model. Real results can have heavier tails, and the bands say nothing about the fall from a previous peak. The downswing guide covers that separately.
Whether to keep playing. A loss that sits inside the 95% band is still a real loss of real money. Nothing in a variance model is a reason to play longer or at a higher stake to win it back.
Check yourself
Your records show 1 BB an hour with an SD of 40 BB an hour over 400 hours, at about 25 hands an hour. What do you enter?
A win rate of 4 BB/100, a standard deviation of 80 BB/100 and 10,000 hands. The win rate scales with hands (× 4), the SD with the square root of hands (× 2), and 400 hours × 25 hands is 10,000. The simulator then shows an expected result of +400 BB and a 30.9% probability of finishing behind.
With the Live NLH preset over 5,000 hands, why is the probability of finishing behind as high as 33.6%?
The expected result is 6 × 50 = +300 BB, but the spread is 100 × √50, about 707 BB. Break-even sits only 0.42 spreads below the expected result, and about a third of a normal distribution lies further below than that.
A friend is up 1,000 BB after 200 live hours, with an SD of about 50 BB an hour. Does that show an edge?
Not on its own. Their measured win rate is 5 BB an hour, but its 95% band is about ±6.9 BB an hour, from roughly −1.9 to +11.9. About one break-even player in 13 would be up at least 1,000 BB after 200 hours.
Sources and method
Original Overshove writing and arithmetic. Every figure uses the same fixed-edge normal model as the simulator above: expected result = win rate × hands ÷ 100, spread = standard deviation × √(hands ÷ 100), and standard-normal values for the bands and the probability of finishing behind. Hourly conversions assume 25 hands an hour, and money is shown at a £2 big blind. The weighted SD method treats each session's variance as proportional to its hours. The sessions, the friend and the tournament cash rate are hypothetical, and the tournament example assumes independent events. No figure is a benchmark for your game or a forecast.
Questions
How do I convert BB per hour to BB/100 for live poker?
For the win rate, divide by your hands per hour and multiply by 100. For the standard deviation, multiply by the square root of 100 divided by your hands per hour. At 25 hands an hour that means multiplying the win rate by 4 and the standard deviation by 2, so 2 BB an hour with an SD of 50 BB an hour becomes 8 BB/100 with an SD of 100 BB/100.
How many hours of live poker do I need to know my win rate?
It depends on the size of your edge compared with your swings. With an SD of 50 BB an hour, a measured edge of 3 BB an hour needs about 1,067 hours before its 95% band clears zero, and an edge of 1.5 BB an hour needs about 4,268. Halving the edge multiplies the hours needed by four.
Is higher or lower variance better?
Neither is better on its own; what matters is your win rate compared with your swings. Lower variance makes results easier to read and a bankroll safer at the same win rate: with an edge of 3 BB an hour, an SD of 50 BB an hour needs about 1,067 hours before the 95% band clears zero, while an SD of 25 needs about 267. A game with bigger swings and a bigger edge can still be the better game if your bankroll can absorb them.
What does a high variance tell you?
That single results sit far from the average, so short samples say little. In poker, a high standard deviation means bigger swings in both directions: you need more hours before your win rate means much, and a deeper bankroll for the same risk of ruin. It does not tell you whether you are a winning or losing player. Only your win rate, measured over enough hours, can show that.
Can a winning player be down after a year of live poker?
Yes. Using the Live NLH preset, 1.5 BB an hour with an SD of 50 BB an hour, a player who plays 200 hours in a year finishes behind about one time in three. The edge in that example is real; the sample is simply too short for it to show reliably.
Why is my result range so wide?
Because variance is large relative to most win rates. Even a clear winner sees a wide 95 percent band over a finite sample, which is why short-term results mislead.
Does a bigger sample narrow the range?
Yes, but slowly. Spread grows with the square root of the sample while expected profit grows linearly, so the edge becomes clearer over long samples rather than overnight.
What standard deviation should I use?
Use your own tracked figure when possible. As rough starting points, online NLH is often lower than live NLH, while PLO is usually higher. The presets are examples, not claims about your game.
Are the 20 and 100 paths predictions?
No. They are seeded examples drawn from the visible assumptions. They show how different plausible paths can emerge from the same edge and volatility.
Can this simulator tell me whether I played well?
No. It models result variance only. Review hands, table selection, fatigue and decision quality separately before changing stakes or rewriting your view of your edge.