Only the current parameters of the statistical regression need to be a secret, if people try to game the system they just change the parameters. They're very effective, I bet they can work out the risk factor of people who have no/missing data by comparing them with those who do. It's scary stuff!