Bankroll management does not turn a losing game into a winning one. What it does is decide how long the money lasts and whether one bad month ends the whole thing — here is the arithmetic.
July 4, 2026 • Reading time: 6 minutes • By Isabelle Grant, writer on bankroll and casino mathematics
Bankroll management has an unusual property: it is the only widely recommended practice in gambling that does not claim to improve results. It cannot. The expected cost of play is turnover multiplied by edge, and no staking rule changes either term.
What it changes is the distribution — how long the money lasts, how deep a losing run can go before it ends the activity entirely, and whether the decisions made during that run are ones you would recognise afterwards. That is a smaller claim than the usual one and it happens to be true.
The separate balance
The first rule is structural rather than mathematical: gambling money lives in its own place, funded on a schedule, and nothing else is ever moved into it mid-month.
This does one useful thing. It makes the amount visible. Spending drawn from a general account has no natural total — each deposit is small, and the sum is never assembled anywhere. A dedicated balance assembles it automatically, which is why the figure people produce from memory is almost always lower than the one the statement produces.
It also removes the decision that does the most damage, which is the one about whether to add more after a loss. If the answer is fixed in advance by the funding schedule, the question does not arise at the moment it would be answered worst.
Unit sizing and what it buys
A unit is the standard stake, expressed as a share of the bankroll. Smaller units mean more sessions before the balance is gone and a smaller chance that a normal losing run ends the activity.
| Unit size | Bets in a €500 bankroll | Effect of a 20-bet losing run |
|---|---|---|
| 10% (€50) | 10 | Bankroll gone twice over |
| 5% (€25) | 20 | Bankroll gone exactly |
| 2% (€10) | 50 | 40% of bankroll remaining |
| 1% (€5) | 100 | 80% of bankroll remaining |
A twenty-bet losing run is not unusual. In any long series of independent outcomes, runs of that length appear routinely — their absence would be the surprising result. Sizing that cannot absorb one is sizing that depends on the run not happening, which is not a plan.
What the table does not show, and must not be read as showing, is a longer bankroll producing a better outcome. The expected cost is unchanged; a hundred bets at €5 and ten at €50 have the same expected loss. The arithmetic does not care about the schedule. Smaller units buy time and stability, and nothing else.
Rules that only work set in advance
Every rule worth having shares a property: it is decided at a moment when nothing is at stake, and it does not require agreement from the person who will encounter it later.
- Account-level deposit limits. The strongest available tool, because a limit that takes 24 hours to raise cannot be raised at the moment it becomes inconvenient. Licensed operators are required to offer them.
- A session loss limit. A fixed figure, decided before opening a game, at which the session ends regardless of what it feels like.
- A session time limit, timed externally. Interfaces are deliberately poor at conveying elapsed time, and a session that runs long without being noticed is a documented state, not a lapse of attention.
- A fixed funding schedule. One amount, one date. Every top-up outside it is the decision the whole structure exists to prevent.
After a bad month
The single most important rule in this article is also the shortest: stake size does not increase because of losses. Ever, and for a reason that is arithmetic rather than temperament.
Money already lost has no bearing on the expected value of the next bet. A larger stake placed to recover it carries the same negative expectation applied to a bigger number, so the effect of recovery staking is to increase the expected loss and widen the spread at the same time. Doubling progressions make this explicit — they convert many small wins into one loss large enough to erase them all, and they run into the table maximum well before the mathematics rescues them.
The correct response to a losing month is to change nothing, or to reduce. Both are stable; only the third option is not.
One caveat on the whole framework, because bankroll writing tends to imply otherwise. None of it produces profit, and a bankroll spread across a thousand small bets at a 4% edge is expected to be gone in the same way as one spread across ten large ones — just more slowly and with less drama. Time and stability are the deliverables. The claim that discipline turns a negative-expectation game positive is the same claim loyalty ladders make, and it is false in both places.
The monthly review
Two numbers, once a month: total deposited, total withdrawn. Not per session, not per game — those invite pattern-hunting in noise. The monthly pair is the only honest measure of what the activity costs, and the only one that cannot be improved by selective recall.
Add a third if you keep one: how many sessions ended on a pre-set limit versus how many ended when the balance did. That ratio describes a month better than the profit figure, and it responds to the rules above in a way the profit figure never will. It is also, over a long enough period, the difference between a hobby with a known cost and one with an unknown one — which is what the small decisions add up to.
