A plain-English guide to full Kelly, fractional Kelly, qKelly stake sizing, rounded stakes, and why model error matters.
Written by Charlie Marsh, founder and editor of Matched Betting Trading. Last reviewed 24 April 2026.
Kelly staking is designed to link stake size to edge size. qKelly, or fractional Kelly, takes that idea and deliberately scales it down so the stake is more cautious when the fair price is still an estimate.
Full Kelly uses the estimated edge and the net odds to calculate a bankroll percentage. The larger the edge and the more favourable the odds, the larger the theoretical stake.
That sounds tidy, but the assumption is demanding. Full Kelly works from the idea that the probabilities are known accurately. In betting, fair prices are usually model estimates, so that assumption needs caution.
Full Kelly links stake size to calculated edge.
It can become aggressive when the model shows a strong edge.
It is highly sensitive to fair-price errors.
Why fractional Kelly is calmer
Fractional Kelly multiplies the full Kelly stake by a smaller fraction. Quarter Kelly uses 25% of the full Kelly stake. Half Kelly uses 50%. qKelly is a practical way to make that fraction explicit.
The purpose is not to make losing impossible. It is to reduce volatility, soften the damage from model error, and keep the bankroll from being pushed too hard by one estimate.
Quarter Kelly means 25% of the full Kelly stake.
Half Kelly means 50% of the full Kelly stake.
Lower fractions can be appropriate when the model has limited proof.
Why rounded stakes still matter
A calculator might produce an exact stake like GBP 12.47. That is useful for reviewing the model, but practical stake planning often needs cleaner increments such as GBP 1 or GBP 5.
Rounding down is the conservative option because it avoids increasing exposure above the model result. The exact stake still matters, because it shows what the model produced before practical handling.
Keep the exact model stake visible.
Round down to the selected practical increment.
Record the practical stake actually used, not only the theoretical stake.
When the answer should still be no stake
A qKelly calculator can show a neat stake, but the correct decision may still be no stake if the fair price is weak, the market has moved, the bankroll is too small, or the model has not been tested enough.
The better habit is to treat qKelly as the final sizing step after the price edge has already survived scrutiny. If the edge itself is unclear, a smaller stake does not fix the problem.
Do not stake from a fair price you cannot explain.
Do not increase stake size to recover a previous loss.
Pause when the calculated stake feels uncomfortable relative to the bankroll.
qKelly stake check
Confirm EV is positive before reading the stake output.
Choose the Kelly fraction before seeing whether the stake feels exciting.
Check whether a max stake cap is needed for the bankroll.
Use the rounded practical stake for records and review.
Set the stake to zero if the fair price, odds, or bankroll comfort are not clear.
Relevant next step
Open the qKelly calculator with a known fair price and current odds. Compare full Kelly, selected qKelly, exact model stake, and rounded practical stake before recording anything.
