Game Development Reference
In-Depth Information
FIGURe 8.3
a machinations chart
showing the result of
100 runs
Designing Artificial Player Strategies
Artificial players cannot replace real players, and the scripting options will not
allow you to create clever artificial intelligence (AI). The purpose of artificial players
is to test relatively simple strategies to see how the mechanics operate, not to build
an adaptive AI player. To get the most out of artificial players and automated play
tests, it is best to design artificial players that represent caricatures: Design them to
consistently follow a particular strategy, no matter the consequences. For example,
if you want to find out what mix of unit types is best for a real-time strategy game,
design a few different artificial players to build the various mixes that you want to
try and have them play each other. Artificial players can be passive or automatic
just like any other node. This allows you to switch between different artificial play-
ers quickly and thus try different artificial players simultaneously. While you are
running a diagram, you can click an artificial player to switch it between automatic
and passive modes.
It often takes several tries to get the strategy for an artificial player right. This is to
be expected, especially if you are creating artificial players for your own designs.
Designing artificial players is a good way to explore and test your design. Ideally,
you should be able to find many valid strategies for your artificial players to follow.
If you can script an artificial player that consistently beats other artificial players (or
yourself), you probably have found a dominant strategy, and you need to change
the mechanics to reduce the effectiveness of that strategy.
remOvinG aLL randOmness
As we discussed in the section “Randomness vs. Emergence” in Chapter 6, “Common
Mechanisms,” random factors can obscure the operation of a pattern that might
create a dominant strategy. When testing for balance, it is important to identify
dominant strategies, and this will be easier to do if you remove the random factors
from the mechanics.
