Information Technology Reference
In-Depth Information
early development stages of the game, the city population would not increase fast
enough. Sooner or later, the city would bankrupt since the population is too low
for the city to develop. The player must balance the tax rates and expenses to
make the citizens happy. While the city is developed, the tax rates should not be
too low in the later stages. If the tax rates are too low, there would be insucient
budget to maintain the development of the city even though the citizens are
happy. In addition, there are many custom laws that are used for collecting
maintenance fees. Players may come up with some sophisticated strategies so as
to get as much budget as possible in a short time. In management games, as
there are different combinations of strategies to develop virtual items, players
enjoy very much in playing this kind of games. Furthermore, if the simulated
environment of the games is based on examples in real life, the players also
acquire new knowledge while they play the games. However, the major problem
of this kind of games is that there is no clue about how long it would take to
develop the virtual items to a certain level.
In this paper, we propose to use a neural network to evaluate the diculty of
management games. Based on the evaluation score, we can adjust the diculty
level of the game to suit for the players expectation. There are a vast amount
of different management games that have a diverse set of features. We have
developed our in-house management game called Wonders of Seabed. We have
built a manageable set of features so that we can easily evaluate the progress
of the game. While players develop the cities, we evaluate the players financial
strategy and adjust the diculty level of the game. If the players cannot earn
enough money at early stages, the game diculty level would be lowered. On
the other hand, if the players could find out a good strategy to play the game,
our system would adjust the settings to increase the di culty level. Thus, our
game would achieve a better gaming experience for players.
The organization of the remaining sections of this paper is as follows. Section 2
reviews the previous work. Section 3 covers our game system architecture, im-
plementation and the game rules. Section 4 presents the method for evaluating
the game diculty. Section 5 presents our results and finally Section 6 concludes
this paper.
2 Related Work
There are techniques which have been developed for balancing game diculty.
For example, the artificial intelligence technique was adopted to switch the dif-
ficulty of the game based on a finite state machine [10]. The range of game
diculty could also be dynamically adjusted according to the fitness values of
players [15]. An extending reinforcement learning technique can provide dynamic
game balancing [5]. The basic idea is to employ reinforcement learning policy
to monitor the game diculty level that fits the players. In [3], a reinforcement
learning technique evaluates the players satisfaction level to achieve challenge-
sensitive game balancing. The dynamic game balancing scheme is designed for
fighting games. A quantitative modeling technique was proposed to satisfy or
entertain players [1]. A challenge-sensitive technique was present to balance the
 
Search WWH ::




Custom Search