Document Type

Conference Proceeding

Publication Title

MODSIM2021, 24th International Congress on Modelling and Simulation


Modelling and Simulation Society of Australia & New Zealand


School of Science




Snell, J., Masek, M., Lam, C.P. (2021). An evolutionary approach to balancing and disrupting real-time strategy games. In Vervoort, R.W., Voinov, A.A., Evans, J.P. and Marshall, L. (eds) MODSIM2021, 24th International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand, December 2021, pp. 967 - 973.


In most computer games, the level of challenge experienced by a player is dependent on a range of variable factors defined within the game environment. When the end goal of such games is entertainment, the variables are carefully tuned by the designers to achieve a sense of fair, balanced gameplay. For military force design and wargaming applications, where the purpose is to explore elements of a real scenario, the question turns to how the variables can be exploited so as to provide the maximum advantage, and to disrupt the balance in favour of a particular side. In this paper, an automated approach to explore the impact of game variables on game balance is presented and evaluated. Based on an evolutionary algorithm, the approach explores a user-defined set of variables in order to determine optimal combinations of variable values to achieve a defined level of game balance or game disruption. The approach also provides the ability of biasing the search towards a set of user-defined values of the game variables, providing insight into how the most disruption can be achieved with the least amount of deviation from an existing strategy. Two scenarios were developed in a Real-Time Strategy game environment with a focus on verifying the developed approach. Both scenarios were adversarial, with two opposing teams, Red and Blue, with the goal of each team to eliminate the opposition. The level of balance/disruption for a particular set of Blue Team variables was measured as a function of the difference between a target blue win rate and the actual win rate, with a tuneable bias to favour solutions where the solution deviated the least from a ‘fair’ solution (where the Blue Team had the same strength as the Red Team). The first scenario was designed so that both teams were evenly matched in terms of number of units. The scenario was used to explore how the game balance could be disrupted by manipulating variables associated with the Blue Team units. The second scenario was developed so that one team was given significantly more fighting units. This was to test if the approach was able to achieve a particular desired level of balance or disruption despite the starting balance skew. A series of experiments were performed using the developed scenarios to evolve a set of game variables tied to the Blue Team to achieve a range of balance levels while the red team's variables were kept static. The experiments show that it is possible to use the developed approach to balance or disrupt the variables of a game so that the desired level of game balance is achieved. A separate series of experiments also showed that the evolution process could be biased to find game variables that required the smallest amount of change. This is particularly important for balancing video games where designers often prefer only to make slight changes to the variables of a game even when desiring a large difference in a game's balance.



Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Research Themes

Securing Digital Futures

Priority Areas

Artificial intelligence and autonomous systems