Two punters find the same set of +EV bets. One is up 30% after a year. One has lost half their bankroll. The bets were identical. The only difference was how much each one staked per bet.
Bankroll management doesn't get talked about as much as picking the right bets because it's not glamorous. Nobody tweets a screenshot of a disciplined 1.2% stake. But I've watched more good bettors go broke from reckless sizing than from reckless picks. Sizing is the part of the game that determines whether your edge actually pays you.
What a bankroll actually is
Your bankroll is the pool of money you've decided to use for betting, segregated from the money you use for anything else in your life. It is not your savings. It is not the emergency fund. It is an amount of money you have specifically earmarked for wagering, the complete loss of which would be annoying but not financially harmful.
Be honest about that last part. The single most common sizing mistake is running a bankroll that is secretly too big — money that you're actually relying on for other purposes, mixed in with money you're gambling with. When a drawdown hits, and drawdowns always hit, you panic because the loss matters in a way it shouldn't. Panic is how sharp bettors turn into reckless ones.
Pick a number you can genuinely afford to lose. That's your bankroll. Treat it as a closed system.
The three sizing approaches
Flat staking
Every bet is the same fixed dollar amount. $50 per bet, always. $100 per bet, always. The simplicity is the feature. You don't have to calculate anything and your sizing doesn't react to streaks.
Flat staking is fine for beginners and for anyone whose edge-estimation isn't yet reliable enough to trust. It is not optimal for long-term growth because it ignores the fact that some of your +EV bets have more edge than others, but it is robust. You won't blow up.
Percentage staking
Every bet is a fixed percentage of current bankroll. 2% per bet, recalculated after every result. When you're winning, stakes grow. When you're losing, stakes shrink. The percentage stays constant.
This is what most serious punters use when they haven't committed to Kelly. It's self-correcting through drawdowns (you automatically bet less when the bankroll is smaller) and self-compounding through upswings.
Kelly staking
Every bet is sized according to the Kelly Criterion: a formula that takes your edge and the odds and returns the fraction of your bankroll you should wager for optimal long-term growth.
Kelly is mathematically the best sizing scheme for maximising geometric bankroll growth, provided you know your edge accurately. That last caveat is where everything gets interesting.
The Kelly formula
kelly_fraction = edge / (decimal_odds − 1)
Where edge is your EV expressed as a decimal — 4% becomes 0.04. The fraction is the portion of your current bankroll to stake.
A worked example
You identify a bet on Collingwood at $2.10 where you estimate the true probability of a Pies win at 52%. Your EV is (0.52 × 2.10 − 1) = 0.092, or 9.2% edge.
Kelly fraction is 0.092 / (2.10 − 1) = 0.092 / 1.10 = 0.0836, or 8.36% of bankroll. If your bankroll is $10,000, full Kelly says bet $836 on this game.
That's a lot of money on a single sports bet. It's also what the maths says to do if you are absolutely confident in your 52% probability estimate.
Why full Kelly is a trap
Kelly's optimality depends on a big assumption: that your edge estimate is exactly right. Not approximately. Exactly.
In reality, everyone's edge estimate has error bars. You think the true probability is 52%. It might actually be 50%. If you size at full Kelly on a 52% estimate and the real probability is 50%, you are now betting 8% of your bankroll on a coin flip, and your expected geometric growth rate is slightly negative. You are going broke on average.
That's the cliff. Full Kelly is only optimal when your edge estimates are unbiased and accurate. The second your estimates overshoot — and they will, especially in your first year — full Kelly tips from optimal to catastrophic.
The other problem is variance. Even with perfectly accurate edge estimates, full Kelly produces brutal bankroll swings. Drawdowns of 50% are normal. Recovery from a 50% drawdown requires a 100% gain just to break even. Most people can't stay disciplined through that. They tap out or they tinker with the strategy, and either move destroys the compounding.
Fractional Kelly: the practical version
The sensible fix is to bet a fraction of Kelly. Half Kelly. Quarter Kelly. Eighth Kelly. You sacrifice some theoretical growth rate for massively lower variance and much higher tolerance for edge-estimation error.
Quarter Kelly on the Collingwood example above is 0.0836 / 4 = 2.09% of bankroll, or $209 on a $10,000 bankroll. That still feels like a real bet. It's not life-ruining if it loses. Over a hundred bets of this size with genuine +EV, the expected growth is significant. Over a thousand, it compounds meaningfully.
My honest recommendation for most AU advantage bettors, including myself most of the time: start at quarter Kelly. Move to half Kelly only after you have a long enough track record (measured in months and hundreds of bets) that you're genuinely confident your edge estimates aren't inflated.
If quarter Kelly feels too aggressive during a drawdown, it's not the sizing. It's you. The maths is right. Your emotional tolerance is the constraint, and that's legitimate — there's no prize for optimal sizing if you quit halfway through because you can't sleep. Eighth Kelly is a perfectly respectable place to live.
The weak link: estimating your edge
Every sizing scheme more sophisticated than flat staking relies on knowing your edge. This is harder than it sounds.
The cleanest way to estimate edge on a single bet is to compare your price to the market consensus fair odds. If the market implies a true probability of 50% on Collingwood and you're getting $2.10, your edge against the consensus is 5%. The Krok Odds +EV Finder calculates this automatically for every bet it surfaces.
The uncomfortable truth is that “edge against market consensus” is the best single estimate most punters have, but it assumes the market is efficient. If you have genuine private information — a contact at a club, a proprietary model — your edge might be higher than consensus suggests. If you're betting a slower bookmaker whose prices haven't caught up to the market, your edge might be lower than it looks. Know which situation you're in.
The way to sanity-check your edge estimates across time is closing line value. If your average CLV is close to your average claimed edge, your estimates are roughly calibrated. If your claimed edge is +5% and your CLV is -1%, you're kidding yourself and your Kelly sizing is going to hurt you.
Planning for drawdowns
A drawdown is a peak-to-trough decline in bankroll. They are unavoidable. The question is how big, how often, and how you behave during one.
A 30% drawdown on quarter Kelly sizing with a genuine 3% edge is a normal event. It happens every year or two. It is not a sign your edge has disappeared. It is what variance looks like. If you change your strategy in response to it, you are probably making things worse.
Before your first real drawdown, write down what you'll do when it happens. Not “trust the process” as a platitude. Specifically: at -20% bankroll I continue sizing as planned. At -30% bankroll I audit my bets for the quarter to check for process errors but do not change sizing. At -50% bankroll I reassess whether my edge estimation is still calibrated.
Having this rule written down before you need it is worth a lot. Making a decision about sizing while you're down 40% is like making a decision about relationships at 3am. You are not the right person to be making that call in that moment.
When to reassess the bankroll
A bankroll review once a quarter is reasonable. Review means: was my sizing appropriate for the variance I experienced? Is my bankroll still the right size relative to my total financial picture? Has my edge estimation changed materially?
If the bankroll has grown 50% or more, you can raise the base stake up, but I'd argue it's better to withdraw the surplus than to compound it indefinitely. At some point you have more money at risk than you're comfortable with, even if the percentage sizing is the same.
If the bankroll has shrunk 30% or more, don't top it up from outside. The percentage sizing will automatically shrink your stakes with it, which is the correct response. Topping up mid-drawdown is how people go from losing 30% of their betting bankroll to losing 30% of their actual savings.
Good bet sizing is boring. That's the feature. The bettors who last aren't the ones with the best picks — they're the ones whose picks survive twelve months of variance intact. Everything in this guide is in service of that.