# You estimate 53% on a bet priced -105 (decimal 1.952) b = 1.952 - 1 = 0.952 f = (0.952 × 0.53 - 0.47) / 0.952 = 0.0346 / 0.952 = 3.6% of bankroll
Full Kelly is famously volatile and assumes your probability estimate is exactly right - it never is. Most practitioners bet fractional Kelly (a quarter to a half of the computed stake) to shrink drawdowns and hedge estimation error. Kelly with an inflated edge estimate over-bets catastrophically, which is why honest edge measurement matters more than the sizing formula.
Kelly needs two inputs and ParlayAPI supplies both: the offered price (any odds endpoint) and a fair probability estimate (devigged sharp-anchor prices, as used by the +EV scanner). On reading edge estimates skeptically before sizing into them: reading +EV outputs honestly.