This blog post has been written based on the presentation "AI That Makes Fun Games" by Sentience CEO Hyeyon Kwon at Gamescom Asia 2023.

Intro: A real case of AI applied to games

In our last blog, we discussed how implementing a player-centered hero's journey affects the fun of the game. In the process, if you use AI to analyze data and improve the game, you can not only increase efficiency but also ensure that you don't miss the golden hour that will revive the game. In this blog, we will explain the AI technology for fun games that Sentience has been researching and developing, using our game ‘South Pole Bebop’ as an example.

AI x Game = South Pole Bebop

South Pole Bebop

First of all, you need to know what kind of game it is. Let me briefly explain South Pole Bebop, a turn-based strategy game developed by Sentience Game Studio. South Pole Bebop is a PvPvE format where players battle on a 9x9 grid. Each player chooses three of a variety of Antarctic animals, including penguins, seals, and cormorants, to participate in battle. You can choose the most attractive character by appearance. Still, each has different abilities, such as movement range and attack method, so it is helpful to choose a character that fits your strategy well.

The third-party, zombies’ attacks are added to this player-to-player combat. Zombies are on no one's side and will emerge through the tiles to attack the player's base and characters. Zombies preview what attacks they will make during their turn, so players can avoid zombie attacks or use them to inflict damage on their opponents. In addition, terrain features such as water and mountains can be used as a means of attack and defense.

Example #1: Develop characters' skills with AI

SPB_TonySealprano example

You may want to assign special skills to characters in a game, but if the skills do not match the characters' attributes, the game progress may be hindered and the player's immersion may be broken. Let's take the character, Tony Sealprano, from South Pole Bebop as an example.

South Pole Bebop's characters are equipped with one of the following attack skills: melee, artillery, or ranged attack. As Tony is a ranged attack character, he cannot use skills such as hitting with a bat or flipping over an enemy. He could instead fire his weapon in a straight line, roll for a headbutt, or pull up for a headbutt, and the AI would apply Tony's character attributes to choose the appropriate skill. Despite Tony's tough appearance, he is actually an adorable seal with sparkling eyes. Therefore, we can match Tony with the skill of attracting enemies with charm, pulling them in, and then attacking them with a headbutt, to give the player an immersive gameplay experience.

Example #2: Stimulating the challenge of games with AI

SPB_character balance

The feeling of challenge becomes more intense when situations seem surmountable but still difficult. Opponent players and the characters they choose are also challenges. If one of the opponent's characters or a specific character combination is too powerful, the game's balance will be disrupted. The user experience will also be negative. South Pole Bebop utilizes AI to simulate all combinations of characters and their skills and dynamics, making adjustments to ensure that certain characters or combinations do not stand out too much. Even if there are changes to the game system, such as new characters or skills, AI finds the perfect balance and enhances user experience.

SPB_Terrains example

Terrains on the South Pole Bebop map are also obstacles that increase the challenge. For example, mountain tiles can block melee and ranged attacks, and if a character is immersed in a water tile, they lose their ability to attack. Here, AI is also responsible for creating maps by exploring all possibilities to find the most optimal feature placement and provide the right level of challenge.

Example #3: Deliver personalized experiences with AI in games

SPB_zombie spawn example

AI can also provide personalized game experiences that are tailored to the player's skill level. In the case of South Pole Bebop, the zombies' skills and attack powers are all different, and the game progress varies greatly depending on the zombies' abilities and spawn location. In this case, AI is utilized to provide a personalized experience that matches the creation and placement of zombies to the player's skill level.

SPB_ai bot mode example

Also, we think the biggest concern that PvP game developers face in the early development stages is the lack of players. If you can't find the other player and don't match the game, you'll be left out. To overcome this issue, South Pole Bebop put a lot of effort into developing AI players that can take on each new match just like real-world opponents. This is done using deep reinforcement learning technology. By employing deep reinforcement learning algorithms, AI players can adapt to the skill level of real players and provide a much more enjoyable and exciting experience. (→ Learn more about How Sentience Game Studio Applies Deep Reinforcement Learning)

Conclusion: Future of AI and Games

So far, we have looked at how Sentience Game Studio utilizes AI technology when developing games. We hope that by reading this article, you will gain a better understanding of the capabilities of AI beyond just creating game assets. AI has the potential to contribute to the core aspect of games, which is 'fun', as we mentioned in our previous blog. Provide personalized experiences and gaming fun to your players with AI and Sentience, a company to bring revolution to games with AI!