Kelly Criterion for Betfair Trading: How to Size Your Stakes

This page contains affiliate links — if you open an account through them we may earn a commission at no cost to you. It never changes our verdict.

This is a cluster sub of our pillar on Betfair risk management and money strategy. The pillar covers protecting a bank broadly; this page drills into the most mathematically rigorous staking method there is — and, just as importantly, into why applying it naively will hurt you. Read it alongside bankroll management, which is the foundation Kelly sits on top of.

What the Kelly Criterion Is

The Kelly Criterion is a staking formula, developed by John Kelly at Bell Labs in 1956, that tells you what fraction of your bankroll to stake on a bet given your edge and the odds, in order to maximise the long-term growth rate of your bank. It is provably optimal for that goal: stake more than Kelly and you grow slower and risk ruin; stake less and you grow slower but more safely. It is the mathematical answer to “how much should I bet?” — with heavy caveats that matter more than the formula itself.

The key idea is that stake size should scale with edge. A bet where you have a big edge deserves a bigger slice of your bank than a bet where your edge is wafer-thin. Flat staking ignores this; Kelly bakes it in. That’s the genuine insight, and it’s correct — the danger is entirely in the assumption that you can measure your edge accurately, which on Betfair you almost never can.

The Formula, Explained Plainly

The Kelly fraction is: f = (b × p − q) / b, where:

• f = the fraction of your bank to stake
• b = the net odds received (decimal odds minus 1 — so a back at 3.0 gives b = 2)
• p = your estimated probability of winning
• q = the probability of losing (1 − p)

In words: your edge (b×p minus q) divided by the net odds. If the result is zero or negative, Kelly says don’t bet — you have no edge. The bigger your edge relative to the odds, the bigger the fraction. To use it you must first convert the market price to its implied probability and compare it honestly to your own estimate — Kelly is only as good as that estimate, which is the whole problem.

A Worked Betfair Example

Say you’re backing a selection at decimal odds of 3.0 on the exchange, and you genuinely believe its true probability is 40% (the market, at 3.0, implies only 33.3%, so you think there’s value). Plug in: b = 2 (3.0 minus 1), p = 0.40, q = 0.60.

f = (2 × 0.40 − 0.60) / 2 = (0.80 − 0.60) / 2 = 0.20 / 2 = 0.10. Full Kelly says stake 10% of your bank on this bet. On a £1,000 bank, that’s a £100 stake. That already sounds aggressive — and it is, which is the point of everything that follows. Note also that on Betfair you must adjust for commission: the 5% (or 2%) taken from winnings slightly reduces your effective net odds b, which shrinks the Kelly fraction a little. The calculator handles the commission-adjusted figures.

Why You Should Use Fractional Kelly

Full Kelly is theoretically optimal but practically brutal, because it produces enormous bankroll swings — drawdowns of 50% or more are routine even when your edge is real. Most humans cannot psychologically withstand watching half their bank evaporate, and will abandon the method (or tilt) at exactly the wrong moment. So serious practitioners use fractional Kelly: they stake a fixed fraction — commonly a half (½ Kelly) or a quarter (¼ Kelly) — of what the formula says.

The maths behind this is reassuring: half Kelly captures about three-quarters of the growth rate of full Kelly with roughly half the variance. You give up a little long-term growth for a large reduction in the gut-wrenching swings, which is a trade almost everyone should take. In the example above, ¼ Kelly would stake 2.5% of bank (£25 on £1,000) instead of 10% — far more survivable. I use a quarter, and I’d encourage anyone starting out to do the same or go even lower.

The Catch: You Don’t Know Your Edge

Here is the part the textbook formula hides and the part that matters most: Kelly requires your true win probability as an input, and on Betfair you never actually know it. You have an estimate, and your estimate is uncertain. If you overestimate your edge — which losing and even break-even bettors systematically do — full Kelly will tell you to stake far too much, and you’ll go bust faster than flat staking would have allowed.

This is the deepest argument for fractional Kelly: it’s not just about smoothing variance, it’s a hedge against your own overconfidence in your edge estimate. Staking a quarter of Kelly is roughly equivalent to admitting your edge might be a quarter of what you think — which, for most people, is closer to the truth than they’d like. If you can’t estimate your probability with real evidence (a journal and a sample of results), you shouldn’t be using Kelly at all — you should be flat staking small while you find out whether you even have an edge.

From the Desk: Kelly vs Flat Staking, Three Months

Applying Kelly to Trading vs Betting

Kelly was designed for held-to-settlement bets with a clear win/lose outcome and a known payoff, which makes it cleanest for value betting — the kind of held positions in the example above. Applying it to trading (back-then-lay, greening up) is messier, because a trade doesn’t have a simple binary outcome and a fixed payoff — you’re aiming to scratch or green, your “edge” is a distribution of small outcomes, and your risk is defined by your stop, not by losing the whole stake.

For pure trading, most people are better served by a simpler rule: risk a small fixed percentage of bank per trade (defined by your stop distance), rather than forcing Kelly onto a structure it wasn’t built for. Kelly’s principle — size up when your edge is bigger — still applies (a high-conviction swing trade can carry more than a marginal scalp), but the literal formula belongs to betting, not to trading positions you exit early.

The Honest Verdict on Kelly

Kelly is the right idea — stake in proportion to edge — wrapped in a formula that’s more precise than your inputs deserve. Used well, by someone with evidence-based probability estimates and the discipline to take a fraction of it, it genuinely outperforms flat staking over the long run. Used naively, by someone who overrates their edge and stakes full Kelly, it’s a faster road to ruin than almost any other staking method.

My practical advice: if you can’t demonstrate an edge with a real sample of results, don’t use Kelly — flat-stake small and find out if you have an edge first. If you can, use quarter-Kelly with a hard cap on any single bet (3% of bank is a sensible rail), and never let the formula talk you into a stake that makes you feel sick. The discipline to under-bet Kelly is, paradoxically, what makes Kelly work. Pair it with the broader risk-management framework and you have a staking plan that respects both the maths and the uncertainty.

Simultaneous Bets and the Kelly Cap

The textbook Kelly formula assumes one bet at a time, settled before the next. Real betting isn’t like that — you often have several positions live at once, and naively applying full Kelly to each ignores that your total exposure across simultaneous bets can balloon past what any single Kelly figure intends. If you have five “10% Kelly” bets running at once, you are not risking 10% of your bank, you are risking up to 50%, which is reckless even if each individual edge is real.

The practical guard is a total-exposure cap: a hard ceiling on how much of your bank can be at risk across all open positions at any moment, regardless of what the per-bet Kelly figures say. I cap any single bet at around 3% of bank and total simultaneous exposure well below the sum the formula would permit. This is another reason fractional Kelly is sane — quarter-Kelly fractions are small enough that even several at once stay within a survivable total. Whenever the maths and your gut disagree about whether an exposure is too large, side with the cap.

A Practical Kelly Workflow

Here’s how I’d actually apply Kelly without the theory eating you alive. One: only use it where you have evidence-based probability estimates (a model, or a documented edge from your records) — otherwise flat-stake small. Two: convert the market price to implied probability and confirm a genuine gap to your estimate; no gap, no bet. Three: compute full Kelly, then take a quarter of it. Four: apply your per-bet cap (e.g. 3%) and your total-exposure cap. Five: log the bet and, later, grade your probability estimates against results to keep yourself honest.

Step five is the one that turns Kelly from a formula into a discipline. The criterion is only as good as your edge estimates, so the long-term job is calibrating those estimates — checking whether things you call 40% actually happen about 40% of the time. If they don’t, the answer isn’t a different staking formula, it’s a more honest probability, or smaller stakes until your estimates earn trust. Kelly rewards accurate humility and punishes confident error, which is exactly the right incentive. Treat it as a way to express edge you can prove, sized down for the edge you can’t, all sitting on top of solid bankroll management.

Kelly vs Level, Percentage and Fixed-Profit Staking

Kelly isn’t the only staking plan, and it helps to see where it sits among the alternatives. Level staking — the same fixed amount every bet — is simple, robust and ignores edge entirely; it’s the right starting point while you’re establishing whether you even have an edge, because it can’t blow you up through overconfident sizing. Percentage (or proportional) staking — a fixed percentage of current bank — grows and shrinks stakes with the bank and is effectively a simplified, edge-blind cousin of Kelly; it’s a sensible middle ground that captures Kelly’s compounding benefit without requiring an edge estimate.

Fixed-profit staking — varying stake to target the same profit regardless of odds — is generally a poor idea, because it forces large stakes on short prices where a single loss hurts most, and it bears no relationship to edge. Against this field, Kelly’s distinguishing feature is that it’s the only one that sizes by edge, which is its strength and its danger: it rewards accurate edge estimates and punishes inflated ones harder than any of the others. The honest hierarchy for most people: start at level staking, graduate to percentage staking as your bank and discipline mature, and only adopt fractional Kelly once you have evidence-based probability estimates worth sizing from. All of them sit on top of the same non-negotiable foundation — a properly sized bank and the bankroll rules that protect it.

FAQ

Related Reading

Stay in the cluster: risk management pillar, how to manage risk, losing-streak recovery. Concepts: implied probability, expected value. Foundations: bankroll management, calculator, glossary.