Game Balancing Platform

Game Balance is key. The balance of a game, and more specifically, the pieces (unit, fighter, car, etc.) in a game, make it all work in a way that results in the most skilled and the best strategy winning the game - and that is all that matters, well, having fun in the process matters too 😊

Before we dive in too deep, let me clarify the difference between balance and symmetry in very simple terms: Asymmetry can be fun and it indeed makes for really interesting games. Balance is necessary for this to be true. The best strategy can only win in an asymmetrical game only if good balance is present. Symmetry is not so much fun - it is simply boring. Imbalance is never fun - it is simply unfair.


Now that that is clear, I want to present an overview of a tiny little company pivot that led to a plug-&-play product that can balance any game, regardless of its complexity.


Decisive AI, a company that specializes in using Machine Learning (ML) for the training of Intelligent Artificial Players (IAP) in order to replace dumb and cheating NPCs, realized recently that customers were asking for help with game balance. As part of their IAP training process, Decisive AI’s Training Platform generates a lot of very useful QA data and analytics - game balance seems to be a secondary side effect of this process that the company now has decided to package in an easy-to-use and very accessible product.


Hubbed in a web-based UI, Decisive AI’s Game Balancing Platform has three main configurable components which can be accessed by video game companies as users: (i) Specific selection of what needs to be tested / balanced; (ii) the monitoring and management of the engine that communicates to the ‘executor’ which is an SDK that lives in the customer game engine - resulting in precious data that is stored in a database hosted by Decisive AI; and (iii) the visualization of the results in a comprehensive statistical analysis of the data.


The Game Balancing Platform has been designed to tackle two distinct phases of the balancing process with a single product.


1. General: Is the game balanced or not? Yes/no, a little bit: when ‘this’ is used, then this unbalanced situation happens - these are general game balance questions in need of answers based on overall game analysis.


2. Specific: How much should this shotgun cost within the game? Too cheap/expensive - these are specific attribute itemized questions in need of specific recommendations.


The Game Balancing Platform is meant to help answer these questions, however, it is important to note that not all games are suitable for this product. Single player games, for example, are balanced in a completely different way than multiplayer games: by fine tuning the challenge and difficulty levels towards completion of the game. Multiplayer games are the real target for this product, as balancing is key in the sense we have been describing. Luckily, multiplayer games are often designed to be played programmatically - in a headless manner, without a dependency with the UI, graphics, music, etc. This allows the platform to play many millions of episodes in very short amounts of time, reaping the benefits of ML and superb software architecture and design, in the context of game balancing.


Once the Game Balancing Platform is set-up and installed for a specific game, it can be used endlessly, everytime something (like a weapon) is tweaked or added/removed from the game. A few cycles of game balancing can be run and the data analytics reviewed by the development/QA teams, allowing resources to be freed from the tyranny of massive double-entry sheets where everything is painstakingly manual!


Decisive AI believes that this Game Balancing Platform is something the video game industry wants and will benefit from. What do you think?