Iterative Deepening

• These games are typically time bound.
• So, the agent selects a move and then the player does.
• Each gets say one minute.
• The agent builds his tree, and then selects his move.
• This throws away a large part of the tree, but he can still keep the branch he selected.
• So, if he had 12 plys, he now has 11.
• The player is going to select one branch so he'll be down to 10 ply.
• Of course, he can keep calculating while the player thinks.
• Perhaps the agent will be back to 12 or even 13 ply by then.
• Once the user selects the agent is back to 11 or 12, but can continue from there.
• Each move you keep what you have and iteratively deepen.