The Kelly Criterion Is Brilliant Until Your Probability Estimate Is Wrong
fractional Kelly, bankroll growth and the danger of overconfidence in the edge estimate.
The Kelly criterion offers an answer to one of betting's favourite questions: how much should you stake when you believe you have an edge?
Its answer is mathematically elegant and emotionally dangerous.
Kelly aims to maximise long-term bankroll growth. It increases the stake when the estimated edge is larger and reduces it when the price is less attractive. The problem is that the formula trusts your probability estimate. Bettors are rarely as accurate as their spreadsheets look.
A simple Kelly example
Suppose decimal odds are 2.20 and you estimate the outcome has a 50% chance of winning.
The net odds are 1.20. The Kelly fraction is:
(1.20 × 0.50 − 0.50) ÷ 1.20 = 0.0833
Full Kelly recommends staking about 8.3% of the bankroll.
That is aggressive. A €1,000 bankroll would place roughly €83 on one uncertain event.
If your 50% estimate is correct, the bet has positive expected value. If the true probability is only 44%, the apparent edge disappears and the formula has confidently instructed you to overbet a bad price.
The formula is not the difficult part
Anyone can type odds and probability into a calculator. Producing a calibrated probability is the real work.
A model may omit an injury, overfit recent results or underestimate the bookmaker margin. A human may increase a team's chance because the narrative feels persuasive. Tiny errors matter because Kelly responds directly to the edge.
At odds of 2.20:
- 50% estimated probability produces a meaningful stake
- 46% produces only a small edge
- Below about 45.5%, there is no positive edge at all
A few percentage points can transform “bet 8%” into “do not bet.”
Why many people use fractional Kelly
Half Kelly stakes half the recommended fraction. Quarter Kelly uses one quarter.
This sacrifices some theoretical growth in exchange for lower volatility and greater protection against estimation error. In the example above, half Kelly would stake about 4.2% and quarter Kelly about 2.1%.
Even those numbers can be uncomfortable during a losing run. Kelly does not promise a smooth upward line. It accepts substantial drawdowns because its objective is long-run logarithmic growth, not emotional comfort.
Kelly can make confidence look scientific
A bettor says, “My model gives this 57%.” The spreadsheet returns a precise stake. The decimal places create authority.
But the precision belongs to the arithmetic, not necessarily to the forecast.
A probability estimate should be tested for calibration. When the model says 60%, do outcomes occur around 60% of the time across a large comparable sample? Does performance survive new seasons and changing markets? Does it beat the price after transaction costs and realistic limits?
Without that work, Kelly is a multiplier for self-belief.
Do not use it to chase losses
Kelly staking falls when the bankroll falls because every stake is a fraction of current capital. It does not increase stakes to recover previous losses.
That separates it from progression systems such as Martingale. Still, a person can misuse Kelly by inflating the estimated edge after losing or treating the bankroll as money that must be recovered.
A bankroll is not protection if it contains rent, debt payments or money needed for ordinary life.
My view
Kelly is one of the best ideas in betting mathematics because it connects edge, price and survival. It is also one of the easiest ways to turn an uncertain opinion into an oversized wager.
Before calculating the stake, challenge the probability. Reduce it. Ask what the model missed. Compare the available odds with the closing market. Consider fractional Kelly even when confidence feels high.
The formula cannot know whether your estimate came from disciplined forecasting or a strong feeling in a nice spreadsheet. It will optimise either one with equal enthusiasm.
Read [expected value in betting](/articles/expected-value-betting-plus-ev) before treating any positive percentage as an instruction to stake.
Questions readers usually ask next
What is the Kelly criterion in betting?
It is a staking formula designed to maximise long-term bankroll growth by linking stake size to the estimated edge and offered odds.
What is fractional Kelly?
It uses a fraction of the full Kelly stake, such as half or quarter Kelly, to reduce volatility and damage from inaccurate probability estimates.
Can Kelly guarantee bankroll growth?
No. It depends on accurate probabilities and still produces losing runs. Overestimating an edge can lead to repeated overbetting.
Is Kelly the same as chasing losses?
No. Proper Kelly staking uses a fraction of the current bankroll and generally shrinks after losses. It should never be used with money needed for living costs.