Model uncertainty and prediction markets
in a Bayesian setting, how can I express known estimation errors? I am
working on a problem where I want to model estimation volatility, but
also biases on the estimation and estimation volatility.
So say I one want to guess who will win the NBA basketball game Boston
Celtics vs Detroit Pistons this evening. If I am an expert, not only
will my belief match the objective probability (if it exists at all),
but I will have a good understanding of how good my guess will be.
Furthermore say I am trading a contract in a prediction market and am
estimating how good guesses of other people is. Currently the contract
trades at .32$ which means the implied probability of the event BC,
'Boston Celtics wins', is 32%. What is a good algorithm to derive
other agents' estimation error? I might form the belief that the
estimation error of the implied probability of BC is Â±10%.