Bankroll Management
Nothing in betting is less glamorous or more decisive than bankroll management. Two people can follow identical picks, and staking alone will make one of them solvent and the other broke. This guide covers the three decisions that matter — how much money, how much per bet, and what to write down — plus the maths of losing runs that makes those decisions non-negotiable.
The bankroll itself
A bankroll is money set aside for betting that has no other job. Not this month's spare cash, not money that will be missed if it goes to zero — because it can go to zero, and treating that possibility as real from the start is the entire foundation. Pick the figure, move it somewhere separate from everyday money, and top it up only on a schedule you set in advance, never in response to losses. If losing the whole amount would change a single decision in your actual life, the amount is too large. This is both the responsible-gambling position and, not coincidentally, the position from which people bet best; scared money makes terrible decisions.
The maths of losing runs
Here is the calculation most bettors have never done. Suppose you back selections around 55% — solid, short-priced picks like the ones on the banker page. The chance of five consecutive losses is about 1.8%, which sounds comfortably rare. But across a season of 300 bets, a run of five losers is not just possible; it is close to inevitable, and a run of seven is unremarkable. At longer odds it gets sharper: back 30% draws and a losing run of fifteen is something a season can easily produce with nothing whatsoever wrong with the picks.
Losing runs are not a signal, they are a certainty on a long enough sequence, and stake sizing exists to make certainty survivable. The question is never whether the cold streak comes; it is whether your stakes let you still be standing, and still following the process, when it ends.
Stake sizing that survives
Level stakes, 1 to 2 percent of the bankroll per bet. That is the whole recommendation, and its virtue is that it turns the mathematics above into an inconvenience instead of a catastrophe. At 1%, a ten-bet losing run costs a tenth of the roll; painful, survivable, and recoverable by the same edge that produced it. At 10% stakes the identical run is ruin, with the same picks.
A few refinements earn their keep. Recalculate the stake monthly against the current roll rather than after every bet, so stakes drift down in drawdowns and up in good runs without daily fiddling. Consider a half-stake tier for speculative bets — longshots, correct scores, experiments — so curiosity has a price ceiling. And treat accumulators as their own small pot, as the accumulator guide argues, because multiplied variance deserves quarantine.
What about staking proportional to edge — the Kelly criterion and its fractions? The idea is sound and the practice is treacherous, because Kelly punishes overestimating your edge brutally, and everyone overestimates their edge. If the maths appeals, use a quarter-Kelly at most, and only once you have a hundred-plus recorded bets proving the edge you are sizing for actually exists. Until then, level stakes are not the beginner option; they are the honest option.
What never works deserves its own line: doubling after losses. Martingale staking converts a survivable losing run into a guaranteed eventual disaster, with the special cruelty that it works right up until it doesn't. Any system whose safety depends on the losing run ending soon is a system that has misunderstood the previous section.
The record, or it didn't happen
Every bet, written down, with five fields: date, selection, odds, stake, and the probability you (or the model) assigned when placing it. The last field is the one nobody keeps and the one that matters most, because it lets you audit your judgement separately from your luck. If your 60% bets are landing 45% of the time over a large sample, the problem is the estimates, not the variance — and no staking plan fixes broken estimates.
Review monthly, never nightly. A month is the shortest period in which signal begins to outweigh noise, and nightly reviews mostly teach people to abandon good process during normal cold streaks. This is the same reason our own results archive publishes rolling windows rather than daily verdicts: short horizons lie, in both directions.
The quiet payoff
Proper bankroll management will not make bad picks profitable. What it does is subtler and more valuable: it keeps stakes small enough that decisions stay rational, it makes variance emotionally boring, and it guarantees that if you do have an edge — from the model, from your own judgement, from wherever — you will still be in the game long enough to collect it. Betting's graveyard is not full of people who picked badly. It is full of people who picked adequately and staked like the losing run would never come. Set a budget, stick to it, and if the habit ever stops being fun, stop.