I'm Ross. I'm a chemical engineer and I make tiny games in my spare time. I recently made a strategy game with Chess-like AI from scratch in Pico-8. In this series of posts I'll walk you through my design process and share what I learnt throughout.

The aim of this write-up is to highlight the key design decisions I made with my strategy game, Ziege. By sharing my process, I hope this will provide some insight for anyone else looking to build a similar chess-like AI opponent for their strategy games. I’ve learned a lot from the experience, and while my approach isn’t definitive, I think it could be useful for others facing similar challenges. 

Hopefully, this overview will help you avoid a few headaches during your own similar developments!

Inspiration

Earlier this year, I was introduced to the ancient board game, Tafl; a Nordic strategy game often referred to as Viking chess (or siege chess). Having had a long-standing interest in chess, I immediately fell in love with the game's rich strategic depth. Unlike chess, with Tafl being a niche game, there was very little documentation available covering winning strategies and tactics. This made for an incredibly rewarding experience figuring out what worked (and what didn’t) through experimentation.

In a game of Tafl, two asymmetrical sides face off. One side, typically known as the defenders, starts at the centre of the board and includes a king piece, while the other side, the attackers, begins positioned around the edges of the board. The objective of the defenders is to help the king escape to either one of the corner squares of the board, while the attackers aim to capture the king before he can escape.

Despite it's long-standing history, one aspect that makes Tafl less widespread is the wide variety of rules, capture mechanics, win conditions, and even board sizes, within the Tafl family of games. Each variation feels quite different to one another, making the game significantly harder to standardise. 

As a result, I wanted to create my own Tafl-inspired game using a simplified ruleset on a smaller board, making it more accessible for newcomers to the game.

Game Design

Before diving into Pico-8 to start prototyping, I decided it would be better to create a paper prototype first. I hand-drew a grid of squares to represent the board, and used small rocks as stand-ins for game pieces. 

This approach allowed for significantly quicker iteration, allowing for ease of experimentation with different board sizes, number of pieces and rulesets in order to discover what worked best.

From my introduction to Tafl, I was used to playing on an 11x11 board. However, larger boards introduce additional complexity, as a game starts with more pieces, each having more legal moves in any given position, making the game more daunting to a beginner. To reduce this complexity and make the game more accessible, I settled on a 7x7 board.

Furthermore, in traditional Tafl games, the king piece often needs to escape to a corner square to win. I instead opted for a simpler victory condition in which the king piece escapes to any edge square on the board, as this avoids additional complexity that comes with corner escapes, often requiring additional rules to handle captures in the corner of the board and balance gameplay. 

Lastly, in most Tafl games, the king is captured differently than the other pieces. I chose to simplify this aspect by having all pieces, including the king, be captured in the same manner.  A piece is captured when it is trapped between two opposing pieces.

These decisions were all made to streamline the rules and keep the game more accessible, allowing new players to quickly pick the game up without needing to learn special capture rules for different pieces.

Design Goals

While my overall goal for this project was to design my own Tafl-inspired game, one feature I was particularly eager to implement was a chess-like AI opponent. Traditional Tafl games often face the challenge of game balance, as the two sides start with an unequal numbers of pieces and have entirely different objectives. This inherent asymmetry makes it difficult to ensure a perfectly balanced experience for both players.

However, by implementing an AI opponent to play against, balance becomes less of a concern. The focus shifts to creating an enjoyable experience for the player (even if the game itself isn’t perfectly balanced). As long as the AI opponent offers a fun and challenging match, slight imbalances can be overlooked.

With this in mind, developing an AI opponent was essential to crafting a Tafl experience that could offer a pleasant and accessible introduction to the game.

The Pieces

Similarly to chess, traditional Tafl games are games of war; one side of pieces represents attackers, and the other defenders. 

Instead of sticking to the traditional war theme, I wanted to take things in a different direction by "gamifying" the board and pieces. My goal was to give them a unique design that would stand out visually and make the game feel more immersive.

Objectively analysing the gameplay, where one side swarms the other while the opposing side attempts to escape, it became clear that this dynamic lent itself perfectly to the theme of a zombie outbreak. I settled on an outbreak in a scientific laboratory where the attacking side took the form of the infected horde, whilst the defending side took the form of a team of scientists. With this in mind, I took to aseprite and mocked up some pieces for the scientists and the infected.

Instead of a traditional king piece, the scientists are tasked with escorting the head scientist, who carries the cure to the infection, to an edge square on the board, so that they can escape the laboratory with it safely.

At this point, it dawned on me that I needed a clear way to distinguish the head scientist from the other scientist pieces. In traditional Tafl games, the king is often made taller than the other defenders, making it easy to identify in crowded positions. 

I applied a similar approach, having the head scientist hold the cure (a laboratory flask filled with red liquid) high above his head to make him taller than the other game pieces. However, since the game is in 2D, this alone wasn’t enough to differentiate him from the crowd easily. A simple edit to further set the head scientist apart from the other scientist pieces was changing the head scientist's hair colour.

Lastly, I added a set of outlines to all of the pieces to ensure they could be easily distinguished from the background.

The Board

With the pieces finished, I shifted my focus to the board. Thematically, the squares were designed to represent the tiled floor of the laboratory. I marked the edge squares on the board with crosses to clearly signify the escape points for the head scientist - a trait also seen on traditional Tafl boards.

For a a 7x7 board, I decided to make the individual squares 16x12 pixels. With these dimensions, the squares were shorter than the game pieces. This caused the pieces to overlap slightly, giving the game a sense of depth that would be lost otherwise.

With the board layout complete, this left a lot of empty space available at the top of the screen. I opted to fill this space with decorative computer monitors to fit the lab theme. This provided the perfect opportunity to incorporate a diegetic UI.

Put simply, a diegetic UI is when a game’s interface is built into the world itself, rather than as overlays or menus outside of the game’s environment, making it feel more immersive. GoldenEye 007 for the Nintendo 64, for example, uses James Bond’s watch as the in-game menu, keeping the UI within the world of the game. 

With this concept in mind, I added a larger monitor to the back wall, that stretched across the length of the screen, on which I could display important game information like which side's turn it was to move.

With game design covered, next up I'll be going over the techniques I used to make Ziege feel more engaging.