July 28, 2026
Everything below uses the same standard assumptions: a five big blind per hundred hands win rate, a standard deviation of ninety big blinds per hundred, and a hundred big blind buy-in. If you disagree with those, the arithmetic still works, it just moves. The point is the shape of the answer, not the decimal.
A fair objection first: those are hand-based numbers, and live poker doesn’t count hands. Nobody at your table is logging 25 hands an hour by hand.
That’s exactly why the app measures in hours and dollars instead. Big blinds per hundred is the unit the published research uses, so it’s the honest unit for the arithmetic on this page. Where it lands in practice is your hourly rate, which is a thing you can actually observe.
29% of the time, a winner is losing
Take a genuine five big blind per hundred winner and give them ten thousand hands.
Expected profit over that stretch is five hundred big blinds. The standard deviation over the same stretch is nine hundred.
The noise is nearly twice the signal. Work through the normal distribution and a real winner is in the red after ten thousand hands about twenty-nine percent of the time. Not a bad run. Not tilt. Just what the distribution does.
Live, ten thousand hands is most of a year. So a winning player being stuck at the end of a year isn’t evidence of anything, and this is worth sitting with, because the instinct to change something after a bad stretch is exactly what that arithmetic says you should resist.
Three million hands to be sure
The obvious follow-up is: fine, how many hands until I actually know my win rate?
To pin it down to within one big blind per hundred with any confidence you need roughly three million hands.
At live speed, call it twenty-five hands an hour, that’s a hundred and twenty thousand hours at the table. That isn’t a sample. That’s a career, and then some.
Which means the honest position is that nobody knows their real win rate. Not you, not the regular at your table who quotes his to one decimal place, not us.
The wrong conclusion to draw is that measuring is pointless. The right one is that everybody is working with an estimate, so the only question is whether yours is built from what actually happened or from what you remember. A rough number beats a feeling. It isn’t close.
24 buy-ins for a thin edge
Here is the one that surprises people.
Take that same five big blind per hundred win rate, a ninety big blind per hundred standard deviation, and say you will accept a five percent chance of going broke. The bankroll that requires is twenty-four buy-ins.
Now halve the win rate. Same game, same swings, you’re simply beating it half as hard. The requirement goes to forty-eight.
The roll doesn’t scale with how good you are. It scales with how thin your edge is. Halve the edge and you double the bankroll, because the swings haven’t moved and you now have less pull against them.
This is why the answer to "am I rolled for this game" can’t be a fixed number of buy-ins. It depends on a win rate you don’t precisely know, which is the whole problem, and it’s why the Bankroll Advisor in the app leans on textbook assumptions early on and shifts toward your own numbers as sessions accumulate rather than pretending to certainty it doesn’t have.
And when your win rate at a stake isn’t clearly positive, it doesn’t produce a bankroll requirement at all. It says the game hasn’t been shown to be beatable yet, because a losing game isn’t a bankroll problem and no amount of roll will fix it.
Why this is an argument for tracking
Read together, the three numbers say something uncomfortable. Your results over any stretch you can actually play are mostly noise. Your win rate will never be pinned down. And the decision that matters most, whether you can afford the game you’re sitting in, depends on the number you’re least sure of.
None of that’s an argument for not measuring. It’s the argument for measuring properly, and for treating the output as an estimate that gets less bad over time instead of a score.
The players who get this wrong in both directions are easy to spot. One quotes a win rate from thirty sessions as though it were settled. The other says the whole thing is variance and keeps no records at all. Both end up running on memory, and memory keeps the good nights.