Aww... I was so excited that I had an idea, I forgot to read everyone else's suggestions. Also, I'm not good at reading C++ and I don't know enough about the existing CGA code to know exactly what the hell you're talking about, so can you tell me if your code is determining the reward before or after the block is hashed? Sorry if that's a stupid question, but that's the only possible problem I can see right now-- if the reward outcome of block 2 is based on the hash of block 1, it's just as easy to predict whether block 2 will pay out as it was before.