Kelly criterion in sports betting: a full worked example

Punters across Melbourne and Sydney are not short on football to bet on, and the modern Australian sportsbook menu runs from Thursday night AFL through to Saturday afternoon NRL and full A-League fixtures on a Sunday evening. With so many markets available, the discipline is rarely about finding a bet — it is about sizing one correctly. That is where bankroll modelling earns its keep, and the Kelly Criterion remains the most respected framework for working out how much to risk on any given wager. Learn more about Kazinovostokskachat.xyz.

The headline attraction of the Kelly Criterion is its promise of mathematically optimal bet sizing, but in real life it demands a level of certainty that even sharp punters rarely enjoy. This guide walks through the formula using a realistic AFL match, shows where the calculations break down, and explains why fractional staking is the only version most bettors should ever touch.

What the Kelly criterion actually calculates

The Kelly Criterion was originally developed by John Kelly at Bell Labs in the 1950s to deal with noisy long-distance phone signals, but it was almost immediately picked up by horse-racing syndicates and casino teams. In its simplest sports betting form, the formula tells a bettor what percentage of bankroll to wager so that the expected log-growth of that bankroll is maximised, given the price you have and the probability you have assigned to it.

The standard formula is f* = (bp - q) / b, where b is the decimal odds minus one, p is your estimated probability of winning, and q is the probability of losing. The output is a fraction of bankroll, so 0.05 means 5%. If the result is negative or zero, the formula is telling you to leave the bet alone because the price does not beat your true estimate.

The most common mistake newcomers make is treating Kelly as a money-making scheme rather than a money-management tool. It does not generate tips and it does not predict winners; it scales a known edge. Anyone using it on bets they have not researched has skipped the part that matters.

Why the standard formula is risky alone

The math behind Kelly is unsettlingly aggressive. If your probability estimate is even slightly too generous, the recommended stake can blow up the bankroll during a bad run. Imagine a punter in Brisbane who thinks their model is 60% confident when the true figure is 52%; Kelly treats that 8% overestimation as free money, and the recommended stakes will be roughly double what they should be.

This sensitivity is why almost no serious professional uses full Kelly on its own. Instead, bettors apply a fraction — half-Kelly, quarter-Kelly or eighth-Kelly — to dampen swings while keeping most of the long-term growth. A half-Kelly approach sacrifices about 25% of the theoretical growth rate but roughly halves the variance, which most bankrolls cannot tolerate losing.

There is also an assumption baked into the formula: that the bettor can place the full recommended size at full decimal odds without restriction. In Australia, where Sportsbet and the state TABs actively cut down winning punters, this assumption breaks almost immediately, so any worked example must account for stake limits and price compression, not just the textbook arithmetic.

A concrete Australian football match

To make this less abstract, consider a real-shape AFL fixture between Collingwood and Carlton at the MCG on a Friday night. Suppose the bookmaker is offering $2.20 for the Magpies, which implies about 45.5% once the overround is removed. Your own ratings model, however, gives Collingwood closer to a 52% chance based on recent form, the venue, the expected midfield matchup, and Carlton's injury list.

Plugging the numbers in: b = 1.20, p = 0.52, q = 0.48, so f* = (1.20 × 0.52 - 0.48) / 1.20 = 0.144 / 1.20 = 0.12, or 12% of bankroll. On a $5,000 Australian-dollar bankroll that means a recommended stake of $600, a serious commitment for a single Friday fixture and a clear sign that Kelly is reserved for long-term portfolios rather than weekend punting.

If the same Collingwood price is available tighter elsewhere, the picture changes. At $2.10 the implied line is closer to 47.6%, and the formula gives f* = (1.10 × 0.52 - 0.48) / 1.10 = 0.0282, or 2.8% of bankroll. The single most powerful tool for an Aussie punter is therefore not a better model but a better price.

Picking fractional Kelly in real life

The sensible interpretation of a 12% full-Kelly signal is to drop down to somewhere between a quarter and a half. A quarter-Kelly on the same numbers says to risk 3% of the bankroll, or $150 on the Collingwood example. That keeps the long-run growth curve attractive while giving the bettor room to be slightly wrong on the probability estimate.

Half-Kelly fans argue that 6% on a fixture with strong edge is the sweet spot, and there is solid mathematical backing: it typically captures around 75% of full-Kelly's growth at about a quarter of its variance. The choice depends on the bettor's life outside betting; a wager that swings the bankroll by 10% will keep a Sydney punter up at night, even if theoretically correct.

Fractional Kelly also produces more bets at meaningful sizes. Full Kelly generates a tiny number of well-sized wagers, while quarter Kelly generates a larger sample of useful ones. In Australian sport, sample size matters, so a Sydney-based AFL model running on around 60 matches a season will be more reliable if it places 120 small bets than if it sits on the couch waiting for the rare 12%-of-bankroll monster.

Staking banks in Australian dollars

Running Kelly in AUD is straightforward because the percentages do not change, but the practical frictions are Australian. Most local bookmakers settle in dollars to the cent, and minimum stake rules effectively place a floor on fractional Kelly; if your full-Kelly signal is 0.6% on a $3,000 bankroll, the implied $18 wager is below the typical minimum and rounds up to the nearest dollar, slightly inflating the risk.

Bankroll separation matters as well. Many savers run a separate betting account, topped up monthly in line with disposable income, so the Kelly percentages line up cleanly with the running total, while mixing the bankroll with rent money introduces skew at the worst moment.

Tools that automate the calculation help, and Australian bettors are well served by platforms that publish transparent staking guides. Sites such as Goldenbet sit at the intersection of casino and sportsbook content, useful for punters who prefer a single account. The discipline of sizing bets to edge is well established in poker too, where pot odds reasoning follows almost identical logic to a Kelly framework.

Common pitfalls when betting with Kelly

The first pitfall is bias in your own probability estimate. Anyone who watches a few NRL matches a weekend and thinks their read of a game is "about 55%" is kidding themselves; most trained analysts operate under the discipline of writing the number down before looking at the market. Without that discipline, Kelly is feeding on garbage and the recommended stakes will not save the punter.

The second pitfall is bookmaker restrictions. Australian operators are not shy about limiting accounts, and a staking strategy that throws 12% of bankroll into single markets is exactly the kind of behaviour that lands on a limit list. Spread action across multiple agencies instead, accepting smaller stakes and more price comparison.

A third pitfall is treating Kelly as a single-bet decision. The formula assumes the bettor is making many similar bets over time, and any single result is almost irrelevant to the long-run picture. A player who loses three straight full-Kelly bets will be out of the game emotionally long before the math catches up, which is another argument for fractional scaling. Offshore platforms remain a temptation, often pitched through sites like this offshore platform, and the Australian Communications and Media Authority has been actively blocking such venues.

Comparing prices across bookmakers

By the time the math is done, the real work in Australian sports betting is line shopping, because operators across the country — from TAB to Sportsbet, Bet365 Australia to Betfair — frequently disagree on AFL and A-League prices by more than a full point. A Collingwood-Carlton line at $2.20 at one agency can be $2.30 or $2.35 elsewhere, and a punter who applies full Kelly using the worst available price gives up edge to laziness. Licensing and verification matter as well: before sending any bankroll anywhere, it is worth checking the licence that backs a given platform, and the Northern Territory Racing Commission and the New South Wales Liquor and Gaming authority are the main Australian regulators for locally licensed betting.

Keeping a routine helps. Going through a short list of trusted agencies before each round, jotting down the best available price for the bet intended, and applying the Kelly formula to that number is a discipline that separates bettors who grow a bankroll from bettors who drain one.

The trick to remember is simple. The Kelly Criterion does not promise a steady stream of winners or a happy run on the punt — it promises a way of sizing bets so that, in the long run, your bankroll grows at the highest possible rate given the edge you actually hold. Treat the formula as a tool that protects the edge rather than manufactures it, never stake full Kelly unless you are confident your probability estimate is well calibrated, and round down to a quarter or a half to give yourself room to be wrong. Australian punters who combine fractional Kelly with disciplined line shopping across the major operators are the ones who stay in the game long enough to see the compounding work in their favour.