Bankroll management is not a moral position about discipline. It is a response to a measurable fact: over any stretch of poker you are able to observe directly, the variance in your results is far larger than your edge.
The buy-in figures above are the ranges widely used by serious players rather than anything derived from your own results. They are a starting point; a player with a smaller edge, a higher-variance style, or no income outside poker should sit at the conservative end of every one of them.
How long noise hides the answer
In online six-max no-limit hold'em, the standard deviation of results is typically reported in the region of 90 to 100 big blinds per 100 hands. A strong win rate in the same units might be 3 to 5 big blinds per 100.
Read those two numbers together. The swing around your result, per 100 hands, is something like twenty times the size of the result itself. Which means:
- A 10,000-hand sample tells you almost nothing about whether you are a winning player.
- A losing month is entirely compatible with being good.
- A winning month is entirely compatible with being bad, and this is the direction people fail to check.
Tournaments are worse by a wide margin. When only a small fraction of the field is paid and most of the prize money sits at the very top, a skilled player loses their buy-in in the large majority of entries. Hundreds of tournaments can pass without a result that reflects the player's actual standard.
The bankroll exists to keep you at the table long enough for this to resolve. That is its only function.
Rules that survive contact with a bad run
Set the roll aside, in a real sense. A bankroll is a defined sum that is not also the rent. Money you might need for something else is not a bankroll, it is a countdown.
Move down without negotiating. The rule has to be written before the downswing, because during one you will find persuasive reasons why this particular case is different. If the roll drops below the minimum for your stake, drop a level until it recovers.
Take shots on a fixed budget. Moving up on a defined number of buy-ins, with a pre-set point at which you go back down, is a reasonable way to progress. Moving up because you just won is how a roll disappears.
Cap what any one session can cost. A stop-loss is not about tilt in the abstract; it is about the specific and well-documented fact that decision quality falls after losses, at exactly the moment when the stakes feel highest.
Do not count the money you are winning while you are winning it. Session-by-session accounting encourages quitting winning sessions early and playing losing ones long, which is precisely backwards for anyone with an edge.
Why the tournament numbers are so much larger
The buy-in guidance jumps from a few dozen for cash to a few hundred for large-field tournaments, and the reason is the shape of the payout rather than the difficulty of the game.
A cash game returns your result continuously: you win a pot, the money is yours, and a session's outcome is roughly the sum of many small independent results. That averages out reasonably quickly. A tournament returns nothing at all for the great majority of entries and everything at once for a handful. When most of the prize money is concentrated in the top few finishing positions, a player's entire expected return depends on outcomes that are rare by construction.
The practical effect is that a tournament downswing is measured in months rather than sessions, and it looks identical from the inside whether you are good or not. The bankroll is what stops that ambiguity from being fatal.
A useful discipline: if you play a mix of formats, size against the one with the highest variance, not the average. A player splitting time between cash games and large-field tournaments who is rolled only for the cash games is, in practice, funding the tournaments out of the cash roll.
The uncomfortable part
Bankroll rules are also a filter. Applied honestly over a long enough period, they will tell some people that they are not beating the game they are playing — because a roll that only ever shrinks, across every stake and every format, is data.
Most people who play poker lose money at it. That is the arithmetic of rake and fees before anyone's skill is considered, and no staking scheme changes it. If the honest answer is that the game is entertainment rather than income, the correct bankroll rule is the one you would apply to any other entertainment: a fixed amount you are content to spend, and no reload.
Related: tournament variance and ICM explain why tournament rolls need to be so much larger than cash rolls.