Following my post on the evaluation function, a few words about MLA's search algorithm (the algortihm names are the names found in the litterature) :

- the alpha beta search type is MTD(f)
- it uses R=2 null move forward pruning
- MLA ignores broadside moves from depth=4 (1 is the top of the tree ie the starting position)
- MLA uses the killer heuristic (if in some position, a move is proven good, try it first if it's legal for other positions in the analysis tree). MLA stores uo to three killer moves in the hash table for each position
- MLA uses the history heuristic, the history record being divided by 2 after each move
- MLA uses enhanced transposition cut-offs at the first top 5 levels

node odering and pruning :

- killer moves are examined first
- ejections are examined before pushes and pushes before simple moves
- some moves are eleminated because they decrease connectivity too much
- some broadside moves are not examined at the top levels of the tree if they decrease connectivity too much.