Kelly Criterion Calculator

Flat betting stakes the same amount regardless of edge size. Kelly scales stakes proportionally to edge - bet more when the price is much better than fair, less when the edge is thin. Over many bets, this produces higher geometric growth than flat betting at the same average stake.

The catch is sensitivity. Kelly is extremely sensitive to your win probability estimate. A 5% overestimate on win probability can recommend a stake twice what is actually correct. That is why fractional Kelly exists.

Professional sports bettors typically use quarter Kelly as a practical baseline. It bets 25% of what full Kelly recommends, accepting lower theoretical growth in exchange for much better variance control and protection against model error.

Kelly fraction = (b x p - q) / b, where b is the decimal profit on a winning bet (decimal odds minus 1), p is your estimated win probability, and q is your loss probability (1 - p). Multiply this fraction by bankroll to get the suggested wager.

At +150 odds (b = 1.5) with a 45% win estimate: Kelly = (1.5 x 0.45 - 0.55) / 1.5 = (0.675 - 0.55) / 1.5 = 0.125 / 1.5 = 8.3%. On a $5,000 bankroll, full Kelly suggests $417.

Quarter Kelly on that same example: $417 x 0.25 = $104. If your 45% estimate is actually 41%, full Kelly drops from $417 to $83 - an 80% stake reduction from a 4-point probability error. Quarter Kelly on the 41% estimate: $21. The reduced multiplier gives a large safety margin.

Set Bankroll to dedicated betting capital only - money you have set aside for wagering, not total savings or investment accounts. Kelly math assumes you maintain the same strategy through drawdowns.

If the result suggests a wager above the book's posted limit, cap at the limit and treat the excess edge as unavailable for this bet.

Full Kelly: fraction 8.3%, wager $417. Quarter Kelly: wager $104. Half Kelly: wager $208.

Now suppose your estimate is off and the true probability is 41%. Full Kelly drops to 1.67% - wager $83 instead of $417. A 4-point error cut the recommended stake by 80%. Quarter Kelly at 41%: $21. The reduced multiplier keeps stakes manageable even when model certainty is low.

If Kelly returns zero, the math says there is no edge at these inputs. Either the win estimate is too low, the odds are not favorable enough, or both. Do not force a bet.

Edge estimates in sports betting carry real uncertainty. Your model might say 45% but the true probability could reasonably be 42-48%. Full Kelly acts as though your estimate is exact, which leads to over-betting when the estimate is high.

Quarter Kelly cuts the risk significantly. You never bet more than 25% of what full Kelly says, which means a 10-point probability error does not blow up your bankroll. The cost is roughly 25% lower geometric growth rate in expectation - a worthwhile trade.

Correlated bets are another reason to go fractional. If you have three related team props and Kelly each independently, the sum of fractions might be 20-30% of bankroll. Fractional Kelly keeps total exposure manageable.