Game Development Reference
In American football, teams do not just play football. They execute predefined
plays. Coaches and players do not just make it up as they go, except in ''broken''
plays, which are usually bad and only occasionally glorious. Indeed, the word
''playbook'' is common in sports. Likewise, it has a place in game AI. In fact, a
book of moves is applicable to game AI beyond the sports games that require one.
At best, a book of moves encapsulates brilliant play and makes it available to the
AI when conditions are right. At other times, the topic can provide a selection of
reasonable, if not brilliant, starting points when the cost of computing them is
too high. Opening moves in Chess fit this description. A book of moves can guide
look-ahead AI to search potentially fruitful paths.
The first thing we will do in this chapter is clear up any confusion between a book
of moves, heuristics, and rule-based AI. Once that is done, we will examine the
rationale for a book of moves, with the idea of encapsulating killer moves leading
the charge. A book can provide more than killer moves; as part of a hybrid AI, we
will see how a book can make it possible to have a reasonable AI at all in a
complex game like Twixt . We will also show how a book of moves can empower
our already effective Minesweeper AI with low-risk opening moves. After sum-
marizing the advantages and disadvantages of the method, we have two projects:
a simple book of opening moves for Fox and Hounds and a more complex one for