The SCG (Specker Challenge Game) is a game
to drive innovation in algorigthmic domains.
In the SCG the agents deal with uncertainty
because they need to assign a confidence in [0,1]
to every belief they offer. The confidence
determines how much reputation is gained or lost
when the belief is supported or discounted.

The confidence depends on:
the amount of resources spent on trying to discount
the belief;
the knowledge the other agents may have;
the likelihood that new algorithmic techniques exist;
etc.

How can a universal probabilistic programming
language help to program the agents?
How can inductive learning help to improve the agents?