A working roguelike
A complete 2D roguelike game, playable from start to finish.
Project · 2026
Roguia
An intelligent 2D roguelike whose monsters learn how YOU play.
- Reinforcement Learning
- Procedural Generation
- MongoDB
- PostgreSQL
- Python
- Project Management
Team
5 people
Duration
6 months
Context
ESEO · DAIA option
Summary
Roguia is a 2D roguelike, a genre inspired by the 1980 game Rogue, where the player clears as many dungeons as possible without dying. We, 5 friends, decided to mix our passion for video games with AI and data. Our touch: our evil little monsters learn through reinforcement learning to adapt to each player's strategy.
My role. I managed the reinforcement learning algorithm crafted to make you loose ;)
Timeline
How it unfolded.
Step 1
Defining the player
Defining the player character, the starting point of the game.
In pictures
A closer look.
Tools
What I worked with.
Python
The language of most of the game and of the monsters.
R
Reinforcement learning
To train the monsters, so that their difficulty adapts to each player.
P
Procedural generation
To generate the rooms of every map randomly.
MongoDB
A NoSQL database that stores the monsters' training logs.
PostgreSQL
A SQL database that stores the player's information, inventory and more.
Git
Each of us worked on a personal branch of the shared code.
Challenges
What got in the way.
Monsters that start out dumb
At first, the monsters were not smart and players could get bored. We pre-trained them to give them a base of knowledge before the player even downloads the game, then let each monster adapt to the player's style: the more the player attacks, the more the monsters dodge.
Five people, one codebase
We each worked on our own branch, held weekly meetings and communicated a lot.
Results
What came out of it.
Smart monsters
Monsters driven by reinforcement learning that adapt to each player.
Procedural generation
Maps whose rooms are generated procedurally.
Databases
MongoDB and PostgreSQL set up for the training logs and the player data.
Next project
Rerailing Simulation