Kelly criterion calculator for prediction markets
Explore how the Kelly model translates your own probability estimate into a bankroll fraction for a binary $1 contract. The result depends on that estimate and is not a recommendation to place a trade.
Kelly stake
How much of a bankroll to put on a position, given the price and your own probability. Full Kelly maximises expected logarithmic growth under its assumptions. Fractional Kelly scales down that model output; neither validates your probability estimate.
Positive edge at your own numbers. Kelly assumes those numbers are right, which is the whole risk.
How to use this calculator
- Enter the contract's execution price.
- Enter your own probability estimate separately from the market price.
- Set an illustrative bankroll and compare full and fractional Kelly outputs.
- Change your estimate to see how sensitive the result is to uncertainty.
Formula
- c = price in cents / 100; p = your estimated probability
- Full Kelly fraction = (p − c) / (1 − c)
- Displayed stake = bankroll × max(0, Kelly fraction)
- Half Kelly = full stake / 2; quarter Kelly = full stake / 4
Worked example
At a 40¢ price and your estimated 50% probability, the full Kelly fraction is (0.50 − 0.40) / 0.60 = 16.67%. On an illustrative $1,000 bankroll, the model gives $166.67 full Kelly, $83.33 half Kelly and $41.67 quarter Kelly, before costs.
Assumptions and limitations
Kelly maximizes expected logarithmic growth under its model assumptions. Errors in probability, transaction costs, correlated positions and changing opportunities can make the output inappropriate. This single-position calculation does not optimize a portfolio or limit drawdowns.
Common questions
Why does the calculator show no bet?
When your probability is at or below the purchase price expressed as a probability, the calculated long-position fraction is non-positive. The displayed allocation is then zero.
Does quarter Kelly remove the risk of a loss?
No. It reduces the displayed stake to one quarter of the full calculation. It cannot correct a wrong probability estimate or guarantee a positive outcome.
Examples are our arithmetic illustrations. Linked references explain the underlying market mechanics or model; they do not endorse a result or a trade.