Models of risky choice have attracted much attention in behavioural economics. Previous research has repeatedly demonstrated that individuals' choices are not well explained by expected utility theory, and a number of alternative models have been examined using carefully selected sets of choice alternatives. The model performance however, can depend on which choice alternatives are being tested. Here we develop a non-parametric method for estimating the utility map over the wide range of choice alternatives. The estimated maps are compared against the three of the most well-known models of risky choice: expected utility theory, cumulative prospect theory, and the transfer of attention exchange model. Model comparison indicates that cumulative prospect theory provides a better prediction of individuals' choices, but the estimated maps show that the overall shape of utility map is different from what the model predicts.