IterativeGreedy locks itself into a direction choice early in the protein chain

[JoelvanIngen]

After choosing a direction for the first few nodes, the algorithm will not return to these nodes in a realistic amount of iterations. We could consider to drop the few directions considered worst by the heuristics to cut off certain directions.

If our heuristics are good enough, this should allow us to get great results because we won't even consider the worst ones.

We could consider increasing the likelyhood of a branch being cut the higher the current search depth is.

[JoelvanIngen]

Idea: Add confidence value and confidence threshold based on heuristic values

• We pass a confidence value to every next depth of a move based on how the heuristics rate the move.
• Next moves can increase or decrease the confidence value further based on the ranking of every move
• If the confidence value decreases below a threshold, we return without looking at child moves

This should let the algorithm freely investigate potential good moves, but quickly drop bad ones.