In 1956, Bell Labs scientist John L. Kelly Jr. published a formula for maximising the long-term growth rate of a bankroll. Originally applied to information theory, it was quickly adopted by gamblers and investors who realised it answered one of the most critical questions in any positive-expectation game: how much of my bankroll should I stake on this bet?
Too little and you leave compounding growth on the table. Too much and you risk catastrophic drawdowns that no winning strike rate can rescue you from. The Kelly Criterion finds the mathematical sweet spot - but only if you feed it accurate inputs. The most critical input is your true edge percentage, and that is precisely where most punters fall short.
DegenToPro is built for Australian punters who want to run a data-driven staking system. The platform calculates and displays live edge percentages across 100+ bookmakers, giving you the exact number Kelly requires. Let's walk through how it all works.
What Is the Kelly Criterion?
The Kelly Criterion is a staking formula that tells you what fraction of your bankroll to wager in order to maximise the long-run geometric growth rate of your capital. It assumes you have a positive expected value (EV) bet and asks: given the odds and your true win probability, what is the optimal bet size?
The key insight Kelly identified is that betting too much is just as harmful as betting too little - just in a more dramatic way. If you consistently over-stake a bankroll, even a genuine edge will eventually destroy you through variance. The formula finds the fraction that produces the highest long-term compounding without risking ruin.
Kelly staking is not suitable for negative EV bets. If you are placing bets where the bookmaker has the edge, Kelly will correctly tell you the optimal stake is zero. It is a tool for disciplined, edge-aware punters - which is exactly the kind of bettor that DegenToPro is designed to produce.
The Formula: f = (bp − q) ÷ b
The standard Kelly formula for a simple win/lose bet is:
| Variable | Meaning |
|---|---|
| f | Fraction of bankroll to stake (the Kelly output) |
| b | Net odds received per unit staked (decimal odds minus 1) |
| p | True probability of winning (as a decimal, 0 to 1) |
| q | True probability of losing (q = 1 − p) |
So in full:
f* = (bp − q) ÷ b
Which can also be written as:
f* = p − (q ÷ b)
Or, using edge and decimal odds directly:
f* = Edge ÷ (Decimal Odds − 1)
Where Edge = (True Probability × Decimal Odds) − 1.
That last form is the most practical for punters, because DegenToPro displays your edge percentage directly. If the tool shows a +5% edge on a bet at 2.50 odds, your Kelly fraction is 5% ÷ (2.50 − 1) = 5% ÷ 1.50 = 3.33% of bankroll.
Notice that the Kelly output is entirely driven by two things: your edge and your odds. Higher odds with the same edge produce a smaller Kelly stake. Lower odds with the same edge produce a larger stake. This is intuitive - a small edge at short odds is highly reliable, so you stake more; the same edge at long odds is more volatile, so you stake less.
Worked Examples
Let's walk through several real-world scenarios so the formula becomes second nature. In each case, assume your bankroll is $5,000 AUD.
| Scenario | Decimal Odds | True Win Prob (p) | b (odds − 1) | q (1 − p) | f* = (bp − q) ÷ b | Stake ($5,000 bankroll) |
|---|---|---|---|---|---|---|
| NRL - short favourite, small edge | 1.60 | 0.65 | 0.60 | 0.35 | (0.60 × 0.65 − 0.35) ÷ 0.60 = 6.7% | $333 |
| AFL - even-money, solid edge | 2.10 | 0.52 | 1.10 | 0.48 | (1.10 × 0.52 − 0.48) ÷ 1.10 = 8.4% | $420 |
| Soccer - underdog, thin edge | 3.20 | 0.34 | 2.20 | 0.66 | (2.20 × 0.34 − 0.66) ÷ 2.20 = 4.1% | $205 |
| NBA - favourite, strong edge | 1.80 | 0.60 | 0.80 | 0.40 | (0.80 × 0.60 − 0.40) ÷ 0.80 = 10.0% | $500 |
| Racing - longshot, small edge | 8.00 | 0.14 | 7.00 | 0.86 | (7.00 × 0.14 − 0.86) ÷ 7.00 = 1.7% | $85 |
Notice how the longshot racing scenario, despite having a genuine edge, produces a very small Kelly stake. That is because the variance at longer odds is enormous - Kelly correctly downsizes the bet to protect the bankroll from the high probability of losing on any given event.
Also notice that a 10% Kelly recommendation on a $5,000 bankroll is $500 - a significant wager. This is full Kelly, and most professionals do not use it. We will explain why shortly.
Full Kelly vs Fractional Kelly
Full Kelly maximises long-run bankroll growth rate in theory. But in practice, it produces extreme variance and psychological discomfort that causes most bettors to abandon the system during inevitable downswings. Fractional Kelly scales the Kelly output down by a fixed factor.
| Kelly Variant | Multiplier Applied to f* | Stake on 10% Full Kelly ($5,000 bankroll) | Volatility Profile |
|---|---|---|---|
| Full Kelly | × 1.0 | $500 | Very high - major drawdowns expected |
| Half Kelly | × 0.5 | $250 | Moderate - most popular professional choice |
| Quarter Kelly | × 0.25 | $125 | Low - suitable when edge estimates are uncertain |
| Tenth Kelly | × 0.10 | $50 | Very low - close to flat staking |
The mathematical relationship between fractional Kelly and bankroll growth is well-established. Half Kelly achieves roughly 75% of the long-run growth rate of full Kelly, but with dramatically reduced volatility. The trade-off is highly favourable for most real-world bettors, which is why half Kelly has become the default recommendation among serious punters.
Quarter Kelly achieves about 44% of full Kelly's growth rate but produces very smooth bankroll curves. It is particularly appropriate when you are uncertain about the accuracy of your edge estimates - a situation we address in detail below.
Why Most Pros Use Fractional Kelly
Professional bettors who use Kelly almost universally apply a fractional variant - typically between half-Kelly and quarter-Kelly. The reasons are practical, psychological and mathematical.
1. Edge Estimates Are Never Perfect
Kelly is only optimal when your probability estimate is precisely accurate. In the real world, your devigged fair probability is an estimate - it might be off by 1-3%. If you overestimate your edge and run full Kelly, you are effectively over-betting, which destroys long-run growth more than under-betting. Fractional Kelly provides a buffer against estimation errors.
2. Full Kelly Drawdowns Are Brutal
Even with a genuine edge, full Kelly can produce drawdowns of 30-50% of bankroll over realistic sample sizes. Most bettors - even experienced ones - cannot stomach a 40% drawdown without making irrational decisions. Half Kelly cuts typical maximum drawdowns roughly in half, keeping you mentally and financially in the game.
3. Simultaneous Bets Require Adjustment
The Kelly formula assumes sequential, independent bets. When you are betting on multiple events simultaneously - which is common for volume EV bettors using DegenToPro - the Kelly fractions compound. Betting three 5% Kelly positions simultaneously means you have 15% of your bankroll at risk at once, which is aggressive. Many professionals divide the Kelly output by the number of simultaneous bets, or simply use a lower fractional multiplier across the board.
4. Account Limits Change the Picture
Australian soft bookmakers restrict sharp accounts. Once you have a $50 maximum stake at a book, Kelly sizing becomes academic for that outlet. Fractional Kelly keeps your absolute stake sizes modest for longer, helping extend the lifespan of each bookmaker account.
Kelly with Uncertain Edges
The single greatest risk in Kelly staking is feeding it an inflated edge estimate. If you think your edge is 8% but it is actually 2%, full Kelly will recommend a stake four times too large - putting you in over-betting territory where the Kelly formula itself predicts long-run bankroll destruction.
There are several ways to handle edge uncertainty:
The Conservative Estimate Method
Deliberately underestimate your edge before plugging it into Kelly. If DegenToPro shows a 6% edge, use 3-4% in your Kelly calculation. This builds in a margin of safety and effectively approximates half-Kelly without requiring you to think in fractional terms.
The Confidence Adjustment Method
Weight your edge estimate by your confidence in the underlying model. If the edge is being driven by a very liquid, well-modelled market (NRL head-to-head at peak volume), use 80-100% of the displayed edge. If it is a niche player prop from an early-morning market with thin liquidity, use 30-50% of the displayed edge.
The CLV Calibration Method
Track your Closing Line Value (CLV) over hundreds of bets. If your average CLV is consistently lower than your average displayed edge at the time of bet placement, your edge estimates are systematically overstated. Adjust the Kelly multiplier down accordingly. DegenToPro's bet tracker records both figures automatically, giving you the data to calibrate your staking over time.
| Edge Confidence Level | Suggested Kelly Fraction | Rationale |
|---|---|---|
| Very high (liquid market, proven model, strong CLV history) | Half Kelly (0.5 × f*) | Variance buffer only |
| Moderate (standard EV bet, average-volume market) | Quarter Kelly (0.25 × f*) | Covers estimation error and variance |
| Low (thin market, early-season, limited CLV history) | Tenth Kelly (0.10 × f*) or flat unit | Prioritise capital preservation |
Get the Edge% Kelly Needs - Automatically
DegenToPro scans 100+ Australian bookmakers in real time and displays the live edge percentage for every +EV opportunity. No manual devigging. No guesswork. Just the number you plug straight into Kelly. World Cup Offer: 20% off all plans right now.
Start Free →Kelly vs Flat Staking
Flat staking - betting the same fixed amount or percentage of your bankroll on every bet - is simpler than Kelly and has real advantages in certain situations. Understanding when to use each approach is important.
| Staking Method | Long-Run Growth | Drawdown Risk | Complexity | Best When |
|---|---|---|---|---|
| Full Kelly | Maximum theoretical | Very high | Medium (needs accurate edge%) | Edge is well-established, bets are sequential, stomach for volatility |
| Half Kelly | ~75% of full Kelly | Moderate | Medium | Most professional EV bettors - best risk/return trade-off |
| Quarter Kelly | ~44% of full Kelly | Low | Medium | New to EV betting, uncertain edge estimates, high volume simultaneous bets |
| Flat % (e.g. 1% per bet) | Sub-Kelly for most edges | Low | Very simple | High volume, very small edges, beginners building CLV track record |
| Flat unit (e.g. $50 per bet) | Sub-Kelly, not bankroll-proportional | Very low | Simplest | Starting out, tiny bankroll, focus on process not growth |
For most Australian EV bettors starting out with DegenToPro, the recommended path is: begin with flat 1% of bankroll staking for your first 200-300 bets while building your CLV track record. Once you have data showing consistent positive CLV, migrate to quarter Kelly or half Kelly. This approach prevents blowing up before your edge has been confirmed, while still allowing compounding growth once you have proof of process.
The critical difference between Kelly and flat staking becomes apparent as your bankroll grows. Flat staking at a fixed dollar amount does not scale - your stake stays the same even as your bankroll doubles. Kelly-based staking grows your absolute stake in proportion to your bankroll, compounding your edge over time.
The Dangers of Over-Betting Kelly
Over-betting Kelly is one of the most common ways skilled, genuine +EV bettors destroy their bankrolls. The mathematics are unforgiving: bet above the Kelly fraction and you are in a zone where expected bankroll growth is actually negative, even if each individual bet has positive expected value.
The Over-Kelly Destruction Zone
If full Kelly says stake 10% and you stake 20% (double Kelly), the mathematics of compounding work against you. You will win bets at the correct rate, but your bankroll will trend downward over time. This is sometimes called "over-betting ruin" - you have a genuine edge on each bet but lose money because your staking destroys the geometric growth that Kelly optimises.
Practical Warning Signs You Are Over-Betting
- Your Kelly fractions exceed 15-20%: A genuine, safely-bet edge rarely justifies a Kelly fraction above 10-12%. If Kelly is recommending 20%+ stakes, your edge estimate is almost certainly inflated.
- You are using full Kelly on uncertain edges: As discussed above, uncertain edges require fractional Kelly. Full Kelly on a potentially-overstated edge is double over-betting.
- You are not accounting for simultaneous bets: Running five separate 8% Kelly bets at once means 40% of your bankroll is live simultaneously. That is catastrophically aggressive.
- Your CLV is consistently lower than your edge display: If your bets close at prices significantly worse than your displayed edge, you are systematically over-estimating edge and therefore over-betting Kelly.
The 5% Rule of Thumb
Many professional bettors use a simple cap: never stake more than 5% of bankroll on any single bet, regardless of what Kelly recommends. This acts as a sanity check against inflated edge estimates and simultaneous bet exposure. Combined with half Kelly, this cap is rarely triggered - but when it is, it prevents catastrophic single-event drawdowns.
How DegenToPro Powers Your Kelly Staking
The Kelly Criterion is only as good as the edge percentage you feed it. That is its fundamental constraint - and it is exactly what separates bettors who run Kelly successfully from those who blow up trying.
DegenToPro is designed specifically for Australian punters who want to run a systematic, data-driven staking approach. Here is what the platform provides that makes Kelly staking viable at scale:
| DegenToPro Feature | How It Supports Kelly Staking |
|---|---|
| +EV Finder with live edge % | Displays the exact edge percentage for every identified opportunity - the primary input Kelly needs |
| 100+ bookmaker coverage | More books means more +EV opportunities and more data points to verify your edge estimates |
| Sharp market devigging | Edge% is calculated against sharp reference markets, not soft book lines - giving you a reliable probability baseline |
| Bet Tracker + verified CLV | Records your edge% at bet placement and closing line - lets you calibrate whether your Kelly multiplier is correctly sized over time |
| Arbitrage Finder | Guaranteed-profit arb bets have a definable edge% and work well with fractional Kelly - ideal for low-risk bankroll building |
| Market EV view | See the EV profile of entire markets, helping you assess confidence in individual edge estimates before applying Kelly |
| Racing software | Australian thoroughbred and harness markets with edge% displayed - apply Kelly to racing as well as sport |
| Discord community (6,000+ members) | Community discussion on staking approaches, Kelly sizing and real bankroll results from members using the platform |
The practical workflow is straightforward: DegenToPro flags a +EV opportunity and shows you, for example, a 4.8% edge on Melbourne City at 2.20 with Sportsbet. You take that 4.8%, plug it into your Kelly formula, apply your chosen fractional multiplier, and place the bet. After the market closes, the bet tracker records your CLV automatically. Over hundreds of bets, that CLV data tells you whether to adjust your multiplier up or down.
This is the difference between running Kelly properly and running Kelly blindly. The formula works - but only if the edge input is accurate and continuously validated. DegenToPro handles the accuracy. The bet tracker handles the validation. You handle the discipline.
Frequently Asked Questions
Can I use the Kelly Criterion for negative EV bets?
No, and you should not try. When you plug a negative edge into the Kelly formula, the output is a negative fraction - meaning Kelly recommends betting zero, or taking the other side of the bet. Kelly only produces a positive stake recommendation when the bet has genuine positive expected value. If your edge is zero or negative, the correct Kelly stake is zero.
How accurate does my edge estimate need to be for Kelly to work?
For full Kelly, quite accurate - within 1-2 percentage points. For half Kelly, the formula tolerates more estimation error. This is why most practitioners use fractional Kelly: it provides a buffer that keeps you in safe growth territory even when your edge estimates are slightly off. The more fractional your Kelly multiplier, the more forgiving it is to estimation error.
Does Kelly work for horse racing?
Yes, provided you can estimate a reliable true probability for each runner. Australian racing is particularly suited to Kelly staking because DegenToPro's racing tools display edge percentages derived from sharp tote and fixed-odds market comparison. The same formula applies - the key is having a credible probability estimate, which is harder to obtain in racing than in two-outcome sport markets.
What Kelly fraction do you recommend for beginners?
Quarter Kelly (0.25 × f*) or flat 1% of bankroll - whichever is smaller. This keeps individual stake sizes very modest while you build a CLV track record over your first 300-500 bets. Once you have 500+ bets of data showing consistent positive CLV, you have the evidence to justify migrating to half Kelly. Do not run half or full Kelly before your edge is validated by data.
Should I recalculate my Kelly stake after every bet?
Strictly speaking, yes - Kelly is a continuous function of your current bankroll. Practically speaking, most bettors recalculate weekly or after significant bankroll movements (more than 10%). Daily recalculation is a reasonable middle ground for active volume bettors. DegenToPro's bet tracker maintains your current bankroll position automatically, making recalculation straightforward.
What happens if two Kelly bets are on at the same time?
You need to reduce each individual Kelly stake to account for simultaneous exposure. A simple approach: if you regularly have N bets open at once, multiply your Kelly fraction by (1/N) or use quarter Kelly across the board. Some advanced practitioners use a portfolio Kelly approach, but for most punters, capping simultaneous exposure at 10-15% of total bankroll is a practical rule of thumb.
Is the Kelly Criterion legal in Australia?
Absolutely. Kelly is a staking method - a mathematical formula for sizing your bets. There is nothing illegal about how you choose to calculate your stake. Betting with Australian-licensed bookmakers is legal for adults 18+ and using analytical tools like DegenToPro to identify edges and calculate stakes is standard practice among professional punters.
The Bottom Line
The Kelly Criterion is the most powerful bet-sizing tool available to sports bettors - but it is only as good as the edge percentage you feed it. Run it with an accurate edge and appropriate fractional multiplier, and your bankroll compounds at the optimal rate. Run it with inflated edge estimates or at full Kelly on uncertain inputs, and even a genuine positive-expectation approach can destroy your bankroll.
The formula itself is simple: f* = (bp − q) ÷ b. For most Australian EV bettors, half Kelly (0.5 × f*) is the right starting point once your edge is validated by CLV data. Quarter Kelly is the right starting point while you are building that track record.
The missing piece for most punters is not the formula - it is the edge percentage input. That requires accurate, real-time devigging against sharp market data across 100+ Australian bookmakers. That is exactly what DegenToPro provides: live edge percentages for every +EV opportunity, a bet tracker that records your CLV automatically, and the data you need to calibrate your Kelly multiplier over time.
More than 6,000 Australian punters are already using DegenToPro to find edges and track results. The live profit dashboard shows community performance in real time. The maths work. The only question is whether you give them the right inputs.
Start Betting with the Edge Kelly Needs
Join 6,000+ Australian punters who use DegenToPro to find +EV bets, measure live edge percentages and track verified CLV across 100+ bookmakers. Start free with no card required - or unlock the full Pro Odds Screener, EV finder and bet tracker from $59.99/week AUD. World Cup Offer: 20% off all plans.
Start Free →