Updated November 2026
Variance is the difference between what you win on average and what actually happens. A real winner can lose for tens of thousands of hands, because a win rate of a few big blinds per 100 hands is tiny next to the swings around it.
Don't take that on trust. Run the simulator and watch six players with the same skill end up in very different places.
Every poker result is a mix of skill and luck. Skill sets the average, and luck moves you around it. In one session luck is nearly everything, and over a very long stretch skill takes over.
Measured in big blinds per 100 hands, a good win rate might be 5 and the swing around it about 90. The noise is eighteen times bigger than the signal.
The win rate is the thin red slice.
The swing is a standard deviation. It tells you how far one block of 100 hands usually lands from your average. About 2 blocks in 3 finish within one swing of the average, and about 19 in 20 within two.
With a win rate of 5 and a swing of 90, 100 hands will usually end somewhere between −85 and +95 big blinds. A big losing session is entirely normal for a winner.
More than most people play in a year. Your measured win rate gets tighter around your real one very slowly, and you need four times the hands to halve the error.
The range containing 95 of every 100 players
At a live table with 30 hands an hour, 100,000 hands is 3,333 hours. Playing 20 hours a week, that's more than three years, and at that point your win rate is still only known to within about 5.6 big blinds per 100 hands either side.
That's why a live player's results over a year say so little, and why it's so easy to think you're better or worse than you are.
Bad enough to make you doubt everything. We simulated 400 players with a real edge, each playing 100,000 hands, and measured the biggest fall from a high point for each one.
Half the players had a worse downswing than the middle figure, and they were all winners.
In buy-ins of 100 big blinds, for a true 5 bb/100 winner.
You have ace-king and he has a pair of queens. He wins 57 times in 100, so he is the favorite.
An ace comes and you're now way ahead. You did nothing to deserve it.
You won a pot you were behind in. It counts the same on the scoreboard as a pot you played perfectly, and that's why short-term results fool people.
Then it takes much longer to show. With a win rate of 2 big blinds per 100 hands, you'd still be behind about 3 times in 10 after 50,000 hands, and with a win rate of 5 it's about 1 time in 9.
Small edges are normal at tougher stakes, and that's one reason players keep moving down and up as their results move around.
Tracking software shows a number called all-in EV. When chips go in before the river, it credits you with your share of the pot in place of the actual result.
With ace-king against queens, you get 43% of the pot whether you hit or not. Over time, that line shows how you've been running compared with your real results. It only removes the luck from all-ins, and everything else stays in.
Yes, and it's the more expensive way. A player on a heater moves up in stakes too soon, starts to believe in reads that happen to work, and stops studying. A big winning month feels like proof, and it's about as much proof as a big losing month.
Keep a simple record for each session: date, stake, hours or hands, and the result. A few months of notes tells you far more than your memory, which keeps the big wins and forgets the small leaks.
A little. Playing fewer marginal all-ins and fewer big bluffs makes your results smoother, usually at some cost to how much you win.
Games with more players in each pot tend to swing more. Tournaments swing the most of all.
Twelve spots in three rounds.
You usually can't tell from results alone. Review the decisions.
Tens of thousands at the very least, and more for a precise answer.
Long losing runs are exactly what normal variance looks like. Run the simulator a few times and see.
A stretch where you lose steadily from a high point in your results.
A run of good results, usually made better by luck.
Because a month is far too few hands to measure it. Most of that change is the swing.
Results per 100 hands are treated as normal with a fixed mean and spread, a common simplification. The range is the mean plus or minus 1.96 standard errors. The downswing figures come from 400 simulated players on a fixed random seed. The swing of 90 is an assumption, and you can change it in the simulator.
Send them the simulator.