How in the holy Universe would one ever distinguish that this "system" over time was working anymore than random chance?
The system itself must be publicly accessible so that anons can fund different corrective, rehabilitative, warning and a variety of other goals and levels as needed. Predictors come from all walks and must be able to see the board to select their jobs. The predictions must also be public in order for the awarders to hand out to matching predictors. Funders need to see correct awards to trust putting their funds into new objectives. And any derivatives layer would also need everything public. Predictions could be private if an AI factbot is awarding to the correct predictor. And news media would cover changes of heart in high profile figures anyways. Initially, until a new far lower equilibrium count is reached, politicians and evil corp types will be resigning in large numbers. Given effectively zero of them resign in history, that will be more than random. Other than identities, locations, fintech, the system is open.
Also, what is the difference between your system that Assassinates Politicians and a duplicate system in which one donates small amounts of cash to sway peoples opinions away from said politicians views, call it Political(instead of Character) Assassination.
The former solves the problem at the source by modifying the selected person etc, the latter tries to route around them which may have limited success.
death?
Anyone who understands how AP works will know that even the most stubborn target will stand down well before then. You should make some budget estimates on how much it might take for various type of people in various roles to wise up and correct their ways before then.
encryption breaks and your "anonymity" just went out the window.
Security is rarely a function on a single variable S(x). And most breaks are not in the cryptography itself.