I think I understand one way to do this now.

For example I classified favorites with odds of 1.00-1.20 as a one group. In my sample dataset there are 121 matches with a favorite that has odds between 1.00 and 1.20. Of these 113 won their match. The average odds for these 121 matches was 1/0.8655 and thus the average 'subjective probability' was 0.8655. The 'objective' probability of winning is 113/121.

Now I am dealing with a Bernoulli variable Y that takes the value of 1 if the favorite won and 0 otherwise. E(Y)=µ and Var(Y)=µ(1-µ)/n where n is the sample size. Y is binomially distributed since matches are independent (or at least very close to independence).

Hence using Central Limit Theorem:

H0: subjective probablity - objective probability = 0

z= (subjective pr - objective pr) / (µ(1-µ)/n)

where µ signals the objective probability 113/121.

----------------------------------------------------------

Now I will have to figure out how to do this for the rates of return (RR) since the above mentioned method suffers from aggregation.