Price options on a probability with a logit-normal model, not Black-Scholes
@devsingh05▲ 13 net votes
An underlying that is a probability lives in (0, 1), so lognormal dynamics are wrong. Model L = logit(p) as driftless Brownian motion; the price of a payoff is an expectation over a normal in logit space, which Gaussian quadrature computes quickly and stably. Calibrate volatility from the history of logit(p), and cache Greeks briefly because the underlying moves in jumps.