Size Your Stakes Like a Pro: The Real Bankroll Playbook

Staking is the part of betting where most serious operators hand back whatever edge their selections produce. Flat stakes leave money on the table in favourable spots; full Kelly blows the bankroll in bad ones. The working middle ground is smaller than most guides admit and requires honesty about how calibrated your probabilities actually are. Here is the real playbook: what flat, Kelly, and fractional Kelly do under the hood, how broker wallet floors shape the calculation, and a drawdown model that tells you whether your bankroll is actually survivable.

Why this is where serious bettors fail

A 3 percent expected-edge bettor placing flat 2 percent of bankroll per bet will typically realise about 70 percent of their theoretical long-run growth over a year. The same bettor applying full Kelly will realise closer to 100 percent of the growth on the median path, but the 5th-percentile path is a 60 percent drawdown. Flat is safe and slow; full Kelly is fast and reckless. The interesting operational space is everything in between, and the exact choice of fraction is a far bigger P&L driver than most bettors give it credit for.

The tool: a live Kelly calculator

Enter your current bankroll, the decimal odds you are about to take, your honest estimate of the true probability of the outcome, and the fraction of Kelly you intend to apply. The tool returns the full-Kelly percentage, your applied percentage at the chosen fraction, the absolute stake in euro, and a verdict on whether the edge is positive at all. Everything runs locally in your browser.

How to read the result

The full-Kelly number is the mathematically optimal growth-rate fraction under one strict assumption: your probability estimate is exactly right. Under real conditions, probability estimates drift. A quarter-Kelly stake captures roughly 75 percent of the theoretical long-run growth with dramatically smaller drawdowns, and a half-Kelly stake captures approximately 94 percent. The extra percent or two of expected growth bought by moving from quarter to full Kelly is almost never worth the variance it introduces; the 5th-percentile path gets brutal very quickly.

The three staking plans compared

Flat staking

Bet the same absolute amount on every ticket, typically 1 to 2 percent of starting bankroll. Simple, variance-friendly, and it quietly under-bets your strongest spots while over-betting your weaker ones. The 70th-percentile bettor does fine on flat staking; the top-decile bettor loses growth to it. Use flat if you have no reliable probability estimates, or if you want one decision (unit size) that does not require recalculation per bet.

Full Kelly

Risk the fraction of bankroll equal to (b·p - q)/b, where b is decimal odds minus 1, p is your estimated probability, and q is 1 minus p. Maximises long-run geometric growth under the assumption that p is right. In practice nobody's p is right, and full Kelly on an overestimated edge compounds losses as fast as it compounds wins. Serious operators almost never run it.

Fractional Kelly

Run quarter Kelly, half Kelly, or another fraction between. Quarter Kelly is the standard for value bettors because it is robust to a 10 to 20 percent overestimate of edge. Half Kelly is defensible when your probability model is backtested and calibrated over a large sample. Anything above half Kelly is a choice to prioritise theoretical growth over realised survivability, and it rarely pays back in practice.

Worked examples through the calculator

Example 1, value bet with a modest edge

Bankroll 5,000 euro, decimal odds 2.10, estimated probability 52 percent. Implied book probability 47.6 percent; your estimated edge is 4.4 percentage points. Full Kelly output 8.57 percent; applied at quarter Kelly that is 2.14 percent, a stake of 107 euro. A flat 2 percent bettor would place 100 euro, within rounding. The fractional Kelly path is slightly larger on high-edge bets and slightly smaller on low-edge ones, which is exactly the pattern that improves long-run growth.

Example 2, sharp arb leg

Bankroll 10,000 euro, decimal odds 2.05 on one side of an arb, estimated probability equal to the book implied of 48.8 percent (because arbs have no directional edge, only a structural one). Full Kelly output is zero. Staking by the arb mechanic, not by Kelly: place the proportional split returned by an arb calculator, not a Kelly-sized stake on either leg alone. This is why arbs and value bets use different staking rules; Kelly does not apply to zero-variance positions.

Example 3, overconfident edge

Bankroll 2,500 euro, decimal odds 3.50, estimated probability 33 percent. Implied book probability 28.6 percent; estimated edge 4.4 percentage points. Full Kelly output 6.80 percent, quarter Kelly 1.70 percent, stake 42.50 euro. If your true probability was actually 30 percent rather than 33 percent (a 3-point miscalibration), full Kelly would still be 2.86 percent and quarter Kelly 0.71 percent. Quarter Kelly survives the miscalibration comfortably; full Kelly drifts down but remains a meaningful stake, which is how overconfidence compounds into real losses.

Broker wallet minimums and how they shape the plan

Asianconnect↗'s 10 euro minimum deposit corresponds to a practical minimum working bankroll of roughly 500 euro if you are running 2 percent unit stakes (the 10 euro figure itself is a floor, not a working unit). MadMarket↗'s 100 euro minimum implies a working bankroll floor closer to 5,000 euro at similar unit sizing; the broker is structurally aimed at higher-volume operators. A bettor with a 1,500 euro working bankroll fits Asianconnect↗ cleanly and fits MadMarket↗ only for larger tickets, which is one of several reasons the broker choice and the bankroll size are entangled.

Ring-fencing exchange liquidity

A back-and-lay workflow needs liquidity on both sides of the hedge at the moment the hedge opportunity appears. If all your capital is on a sportsbook wallet and an exchange hedge becomes profitable, you either skip the hedge or take a slower cross-wallet transfer that introduces timing risk. The discipline is to carve the exchange allocation out of the working bankroll at the start of each month and refuse to redeploy it to sportsbook tickets mid-month. A 30 percent exchange allocation is typical for mixed strategies; arb-only operators often run 60 percent exchange, 40 percent sportsbook.

A rare tip: the monthly re-base with variance check

A drawdown model you can actually use

Assume a 3 percent expected-edge value bettor placing 50 bets per month at quarter Kelly, average odds 2.00. Under those parameters, Monte Carlo simulation produces the following approximate drawdown distribution over a one-year horizon. The 50th percentile outcome is a bankroll 1.35 times starting size; the 25th percentile is 1.05 times; the 5th percentile is a drawdown to 0.72 times starting size, from which recovery takes roughly four to six months on the median path. These are not catastrophic numbers, which is the point; quarter Kelly is designed to make the 5th percentile path survivable. At full Kelly the same model's 5th percentile drops to 0.28 times starting size, which is operational ruin for most part-time bettors.

The takeaway is not the specific numbers; your parameters will differ. The takeaway is that a staking plan without a drawdown model is a staking plan you have not tested. Run the simulation with your own numbers before you commit to the fraction, not after.

Tax-aware bankroll splits for Irish-based players

Gambling winnings are not taxable for Irish residents under current Revenue guidance, so the tax calculation is usually zero on the individual side. What is not zero is the operational cleanliness of your banking, which matters for AML reviews and for any conversation with a mortgage underwriter or similar. The working split used by many disciplined Irish operators is: a dedicated current account for betting cashflows (deposits to brokers, withdrawals back), monthly reconciliation against the broker statement, and an annual summary kept for records. If you run value-betting at a scale that resembles a business, talk to a qualified tax professional before Revenue does; the treatment is nuanced and this page is not a substitute.

When the staking rule breaks

• Your edge estimate is noise. Without a calibrated model, any Kelly output is theatre. Default to flat staking at 1 percent until you have a backtested probability source.
• Correlated bets. Three bets on the same match are not three independent Kelly calculations; size as one combined exposure or you will over-stake.
• Position sizing inside a live in-play window. Prices move fast; by the time you compute Kelly, the odds have shifted. Pre-compute stake ladders for the in-play markets you run, or cap in-play stakes at a fixed 1 percent regardless of edge.
• Bankroll below broker minimum. If the per-bet stake implied by Kelly is below the broker's minimum ticket size, the plan does not execute cleanly. Either increase bankroll or move to a broker with a smaller unit floor.
• Emotional re-sizing. Doubling your unit after a bad week is the textbook signal of tilt. The staking rule exists to remove that decision; respect it or replace it with a smaller flat rule.

Responsible gambling note

Bankroll management is the operational form of self-restraint. If your staking plan requires repeated re-topping from outside capital, the plan is failing and no better formula fixes it. Problem Gambling Ireland and GambleAware publish frameworks for honest self-assessment, and both brokers covered here expose deposit limits, cool-off periods, and self-exclusion inside the account settings. Use them before you need them.

Frequently asked questions