Game Development Reference
In-Depth Information
The words are stirring enough, and the year it was written—1599—predates
modern copyright law by a good 110 years, but the utter familiarity of the
St. Crispin's Day speech from Shakespeare's Henry V makes it a poor choice for
most computer games. The good news is that the game AI need not compute
brilliance; it only needs to be able to import it and use it for its own.
All that being said, computer games lack the benefits of 3,000 years of written
history and scholarship. Cutting-edge games sport novel forms of gameplay,
employing rules of play that are less than two years old and known only within
the development studio working on them. These rules might not stabilize until
the gold master disk is burned. Where does the AI programmer get expertise
when no experts exist? One potentially risky place is from early reviews of the
core gameplay.
Games are supposed to be fun. Because games are financially quite risky to
produce—indeed, few games break even—many studios review the core
gameplay early and often to validate that the fun is there and stays there. These
reviews and play-test sessions are where the earliest expertise with the game will
be forged. This expertise may be rendered useless as the game evolves, but some
of it may survive. Play-test sessions can be mined for good moves if care is taken
from the outset. An observant set of players willing to write things down might be
all that is needed. If the games are all logged and the scores are reported, pro-
mising candidates can be found and replayed for in-depth analysis. As the game
develops, special AI can be used to probe and experiment. This AI need not be
limited by the constraints of the regular AI; it can take longer to think or use
more memory or even use a farm of machines to help. Recall from Chapter 6 that
na¨ve AI methods applied to simple games can generate run times that compare
poorly to the heat decay death of the universe; no farm of AI machines is large
enough to explore a space that large. For any particular new game and the studio
developing it, there will be a balance point between investment and return.
A final consolation in the search for brilliance is that the AI does not have to be
brilliant all of the time. As long as it is rarely stupid, the AI can thrive with
occasional flashes of brilliance punctuating otherwise solid play. As we shall see,
sometimes the topic of moves itself enables the solid play.
Having great moves is only part of the task. The code that uses the topic has to
ensure that any move selected is good for the current situation. In sports,
selection failures are characterized as ''Good team, bad coach.'' Recall Horatio,
our opera singer who broke into song at a funeral? What if he was supposed to
Search Nedrilad ::




Custom Search