Home / Risk Management / Stake Sizing

Risk management guide · Last reviewed November 2026 · The Betting Site

Risk management · Stake sizing

Stake Sizing: Flat, Percentage and Kelly

Stake sizing decides how much of your bankroll each bet risks. Flat staking bets the same amount every time; percentage staking bets a fixed share of the current bankroll; the Kelly criterion sizes bets from your estimated edge. None of them changes the bookmaker's margin — they only change how bumpy the ride is.

Why stake size decides survival

Losing runs are not bad luck; they are arithmetic. Even a bettor whose selections are sound will hit streaks of eight, ten, twelve straight losses sooner or later — that is what randomness produces over hundreds of bets. Whether those streaks are an annoyance or a catastrophe depends almost entirely on stake size. Bet 2% of your bankroll per bet and a ten-loss run costs about a fifth of the pot. Bet 20% per bet and the same ordinary run ends your betting — and tempts you to refill from money that was never meant for it. Most bettors obsess over picking winners; the quieter truth is that stake size determines whether you are still betting next season.

Flat staking

Flat staking means every bet risks the same fixed amount: one unit, every time, whatever the match and however confident you feel. Its virtues are simplicity and emotional flatness: no single result can tilt you into a reckless next one. Its weakness is that the amount does not adjust as your bankroll changes: after a good month the unit becomes a smaller share than intended, and after a bad month a larger one. The standard fix is to recalculate the unit at your monthly bankroll review and leave it untouched in between. For most recreational bettors, flat staking at a small unit is the method we would point to first — not because it wins, but because it is the easiest to actually stick to while losing.

Percentage staking

Percentage staking bets a fixed share of the bankroll as it stands today — say 2% of whatever the current balance is. Stakes automatically shrink after losses, which slows the fall, and grow after wins. Because the stake is always a fraction of what remains, the bankroll mathematically never quite reaches zero; in practice it can dwindle to amounts too small to bother with, which amounts to the same ending spread over more time. Percentage staking demands more discipline than flat staking, because the stake changes after every result and the temptation is to round up after a win or fudge the maths after a loss.

The Kelly criterion, in plain words

The Kelly criterion is a formula from information theory that answers a narrow question: if you genuinely know a bet's true probability, what fraction of your bankroll should you stake to grow it fastest over the long run? In words, the Kelly fraction is your edge divided by the odds on offer: for decimal odds O and your estimated win probability p, it is (p × (O − 1) − (1 − p)) / (O − 1). The idea is elegant; the problems are practical and serious. First, Kelly needs a true probability, and almost nobody has one — the bookmaker's price implies a probability, and beating it requires estimates genuinely better than the market's. If your estimate is even slightly too optimistic, Kelly instructs you to overbet, and overbetting a negative-edge bet loses money faster than flat staking. Second, full Kelly is violently volatile: swings that halve the bankroll are a normal part of its path, and few people can watch that without intervening. That is why professionals who use Kelly at all stake a fraction of it — half Kelly, quarter Kelly — accepting slower theoretical growth in exchange for surviving their own estimation errors. Kelly is a tool for people with proven probability models. For everyone else it is a way to bet too big with mathematical confidence.

A worked example (illustrative only)

These numbers are arithmetic, shown only so you can see how the formulas behave. They are not a recommendation and not a prediction.

Bankroll: 1,000. A bet at decimal odds of 2.00 (evens). Suppose you estimate the true chance of winning at 55%. The Kelly fraction is (0.55 × 1 − 0.45) / 1 = 0.10, so full Kelly stakes 100, half Kelly stakes 50 — while the flat 1–2% convention from our bankroll guide stakes just 10 to 20. Now the catch: if the true chance was actually 50%, the bet had no edge at all. Your 100 Kelly stake was then pure overbetting, placed with total confidence in a number that was simply wrong. That gap between the estimate and the truth is where Kelly does its damage.

Why betting systems fail: Martingale

The most famous staking system, Martingale, says: after every loss, double your stake, so that the first win recovers all previous losses plus one unit of profit. It sounds like a loophole. Follow the sequence: 1, 2, 4, 8, 16, 32, 64. After seven straight losses — a run that will happen to anyone who bets long enough — you have already staked 127 units and must now stake 128 more to continue, all to win back a single unit. Two walls end the story. Bookmakers and tables cap maximum stakes, so eventually you are forbidden from doubling. And your bankroll is finite, so eventually you cannot. When either wall is hit, the loss is enormous and final. Meanwhile every individual bet carried the bookmaker's margin all along, so the system's expected result is identical to betting flat: it simply converts many small wins into rare, catastrophic losses. Fibonacci and every other progression system rearrange the same losses into different shapes. Rearranging is all any of them do.

Choosing a method

Pick the method you can still follow calmly on your worst day, at the smallest size that keeps betting interesting. For most readers that is flat staking at 1–2 units of a 100-unit bankroll. Write it down, set it before the month starts, and change it only at a scheduled review. A staking plan you abandon after three losses is not a plan; it is a mood.

Stake sizing FAQ

What is the safest way to size bets?

Flat staking at a small percentage of a separate bankroll — 1 to 2 percent per bet is a common convention — is the simplest and steadiest option. Safest here only means losses arrive slowly and predictably. No staking method makes betting safe from loss, because every bet still carries the bookmaker's margin.

Should I use the Kelly criterion?

For most bettors, no. Kelly only works if your estimate of a bet's true probability is accurate, and most people cannot estimate probabilities better than the market does. Full Kelly is also extremely volatile. Professionals who use Kelly at all typically stake a half or a quarter of the full amount, precisely because their estimates might be wrong.

Do staking systems like Martingale work?

No. Martingale and similar systems do not change the odds or the margin on any individual bet. They produce frequent small wins paid for by rare, very large losses when a losing run meets a stake limit or the end of your bankroll. The expected result is the same as betting flat — it just arrives in a more dangerous shape.

18+ and responsible gambling: Betting is for adults only and always carries risk. Never bet more than you can afford to lose. See our responsible gambling page for the warning signs, budget rules and limit-setting tools every operator should offer.

Reader comments

Comments show on this page. No account needed — please keep it about your own experience; we remove spam and abuse, and never edit real comments.

Loading comments…

Something wrong or out of date on this page? Tell us — we read every message and fix what we can.