Bankroll Management: The Base of Any Strategy

Define a stake unit

A common practice is to always stake a fixed percentage of your total bankroll (e.g. 1% to 2%) rather than a fixed cash amount, so risk exposure automatically adjusts as the bankroll grows or shrinks.

Use the ROI Calculator in the Tools section to track performance over time, rather than judging by the result of a single bet.

Why flat stakes drift from your real risk

Picture a bankroll of R$1,000 and a bettor who settles on R$20 a bet because it feels like a comfortable, round number, equal to 2% of what they started with. That number doesn't move on its own. If a bad week drops the bankroll to R$600, the same R$20 stake is now roughly 3.3% of what's actually left, which means the bettor is risking proportionally more precisely when the bankroll can absorb it least. Flip it around: if the bankroll grows to R$1,500, that R$20 stake has quietly shrunk to about 1.3%, which under-uses the cushion that improved results just built.

Percentage staking removes that drift by tying the unit to the bankroll instead of to a habit. Recalculate 2% off the current number and the stake contracts automatically during a downswing and expands during an upswing, without any extra decision-making in the moment. That matters, because in-the-moment staking decisions are exactly where discipline tends to break down.

Two ways to set the reference bankroll

There are two common ways to define what '2%' is 2% of. The rolling method recalculates the unit after every settled bet, off the live balance, so risk always stays proportional, bet by bet. The fixed-period method sets a reference number at the start of a week or month and calculates every stake off that fixed figure until the next review, regardless of what happens in between.

Each has a real tradeoff. Rolling staking is the more mathematically 'pure' version, but during a losing run it makes stakes shrink continuously, which can make a drawdown feel slower to recover from because every recovery bet is smaller than the bet that produced the loss. Fixed-period staking is simpler to plan around, since you know exactly what a unit is worth all week, but without a separate rule, it lets you keep staking the same amount through a bad run all the way to the review date.

A workable middle ground many disciplined bettors use is a fixed weekly or monthly reference combined with a hard stop-loss: if the bankroll falls by a set amount, with 20% a common threshold, before the scheduled review, that triggers an immediate recalculation, a reduced unit, or a full pause, instead of waiting for the calendar to catch up.

A worked example across a run of bets

Numbers make the difference between stake sizes concrete. Take a hypothetical bankroll of R$2,000, with a weekly reference unit of 2%, or R$40 per bet. Over a run of 10 bets, suppose 4 win at average odds of 2.10 and 6 lose, a below-average but entirely realistic stretch. The four winners return 4 × (R$40 × 1.10) = R$176 in profit; the six losers cost 6 × R$40 = R$240. Net result: −R$64, or a 3.2% dip in the bankroll.

Now run the identical sequence, same 10 outcomes, same odds, through a bettor staking R$200 per bet, 10% of the same starting bankroll. The four winners return 4 × (R$200 × 1.10) = R$880; the six losers cost 6 × R$200 = R$1,200. Net result: −R$320, a 16% dip from an identical set of results. Nothing about the bettor's picks changed between the two scenarios; only the unit size did, and it turned a manageable dip into a serious one.

Risk of ruin and why bigger stakes don't scale in a straight line

Risk of ruin describes the probability that a losing streak, however unlikely on any single bet, wipes out enough of the bankroll that recovery becomes impractical. The uncomfortable part is that this risk doesn't grow at the same rate as the stake size: it accelerates. A run of losses that barely dents a bankroll staked at 1–2% per bet can threaten the whole bankroll at 8–10% per bet, even though the win rate and the odds are identical in both cases.

This matters even for a bettor with a real, measurable edge, because an edge only plays out over a long enough run of bets. In the short run, variance can produce losing streaks that have nothing to do with whether the underlying judgment was sound. Staking too large relative to that edge means a bad short-run stretch (which will happen eventually to anyone who bets long enough) can end the bankroll before the edge has had the sample size to show up in the results.

The Kelly criterion is the formal version of this idea: it calculates a stake as a fraction of bankroll based on your estimated edge and the odds on offer, using f = (bp − q) / b, where b is the decimal odds minus 1, p is your estimated probability of winning, and q is 1 − p. Say you estimate a 55% chance of a result priced at odds of 2.00 (so b = 1): f = (1 × 0.55 − 0.45) / 1 = 0.10, or 10% of the bankroll on that single bet. Full Kelly is far more aggressive than most bettors are comfortable holding, largely because it assumes the probability estimate is exactly right, which it rarely is. That's why most people who use Kelly-style sizing apply a fraction of it, a quarter or a half, precisely to buffer against the fact that a real-world edge estimate always carries some error.

Common mistakes in practice

Mixing a betting bankroll with day-to-day money is the most common leak. Without a hard boundary, there's no honest number to calculate a percentage from, losses start to feel like they're coming out of rent or grocery money, and wins get spent immediately instead of compounding the bankroll for the next stake.

Increasing stakes after a winning streak, with no predefined rule for when or by how much, is another. A streak raises confidence faster than it raises actual edge, and confidence is a poor input for a staking decision, since it's exactly the moment overconfidence tends to creep in undetected.

Sizing bets by 'gut feeling' about confidence, rather than by a defined method, is a related problem: it's not that confidence-weighted staking is wrong (Kelly-style sizing is a formal version of exactly that), it's that an undefined method produces inconsistent stakes that are hard to review or correct afterward. And betting the same unit across sports or markets where your actual track record differs sharply hides useful information: tracking results by category (using the ROI Calculator or even a basic spreadsheet) often reveals that one category is carrying the whole record while another has been roughly break-even the entire time.

Setting the bankroll itself, and reviewing on a schedule

Before any of the staking math matters, the bankroll figure has to be honest: money that could be lost completely without touching rent, bills, or savings targets. That's not a moral instruction: it's a practical one, because money that isn't truly spare changes decision-making under pressure. A losing stretch is easier to sit through calmly when the money at risk was always meant to be at risk.

Review the bankroll size and the staking plan on a fixed schedule (weekly or monthly) rather than adjusting it in reaction to a single result. Deciding whether the unit should change is a calmer, better decision made by looking at tracked numbers after the fact than one made in the minutes right after a big win or a rough loss, which is usually when the reasoning is least reliable.

Bankroll math changes with multiples and correlated bets

The percentage-of-bankroll approach gets more complicated once accumulators enter the picture, because a single multiple bet concentrates several independent risks into one stake. Treating a 5-leg accumulator the same as a single bet at the same stake size understates the real risk: the probability of all five legs landing is the product of each leg's probability, which drops fast even when each individual leg looks reasonably likely. A stake sized for a single bet is often too large for a multiple with the same odds, precisely because the odds compress a much lower combined probability into a headline number that looks attractive.

A workable adjustment is to treat multiples as a separate, smaller sub-bankroll with its own, lower, percentage per stake, rather than applying the same unit used for single bets. Some bettors cap total exposure to multiples at a fixed share of the overall bankroll each month, so that a bad run of multiples (which is more likely than a bad run of singles, given how the probabilities compound) can't do disproportionate damage to the whole bankroll.

Tracking, not just staking

A staking plan is only as useful as the record kept alongside it. Logging stake, odds, result and category for every bet (even a plain spreadsheet works) turns vague impressions ('I've been doing alright with football this season') into a number that can actually be checked. The ROI Calculator does this arithmetic for you once the raw data (stakes and returns) is entered, but the discipline of logging every bet, including the ones you'd rather forget, is what makes the resulting number mean anything.

Without a log, the mind keeps a selective record: wins and near-misses stay memorable, ordinary losses fade fast, and the resulting impression of overall performance tends to run more optimistic than the real numbers support. A written record is the correction for that, and it's also what makes it possible to spot, over time, whether the staking plan itself, not just the picks, needs adjusting.

Matching the percentage to the market's variance

The 1–2% range isn't an arbitrary default: it's calibrated to the kind of variance found in mainstream markets like match result or over/under goals, where odds sit in a fairly narrow band and outcomes are frequent enough to generate a reasonable sample size within a season. Markets with more extreme odds (outright winner markets, first goalscorer, correct score) carry much higher variance per bet, and applying the same 2% unit there ignores that the swings between individual bets are far larger even when the long-run math still works out.

A practical adjustment is to scale the unit down for higher-variance markets and up, within reason, for lower-variance ones, rather than using one flat percentage everywhere. Some bettors keep two or three separate unit sizes: one for mainstream match-result markets, a smaller one for outrights and long-odds markets, and occasionally a slightly larger one reserved for markets they've tracked closely and shown a consistent, measurable edge in over a meaningful sample of results.