Persistent WebSocket
Every VP keeps a single long-lived connection to the game server, receiving lobby, room, and battle events in real time.
VIP Research Group · Athabasca University
Meet the bots that play your card game for real
TCG Virtual Players are AI agents that log into the Trading Card Game server, join matches, and play full games on their own. Pick a persona, choose an engine, and let them loose.
- LLM Decision Engine Each bot asks Gemini, OpenAI, or DeepSeek what to do next, guided by persona-specific strategy prompts.
- Algo Beam Search A zero-cost engine that scores every legal move through recursive look-ahead, tuned by persona (2 to 5 plies deep).
- 5 Personas Aggressive, Strategist, Deep Thinker, Casual, and Balanced. Each one plays and talks differently.
What is this?
TCG Virtual Players is a research project from the VIP Research Group at Athabasca University. We build autonomous AI agents that connect to the group's online Trading Card Game server over WebSocket, playing full matches against humans or other bots without anyone touching the keyboard.
Each virtual player runs a 5C pipeline (Communication, Sensor, Processing, Memory, Action) that turns raw game events into smart decisions. The Processing stage can use either LLM-based reasoning through commercial APIs or a deterministic algo engine that picks moves using scoring and beam search at zero API cost.
LLM or Algo Engine
Game decisions go to either an LLM with persona prompts or the algo engine's recursive beam search. You choose per VP.
Smart Trap Logic
Trap activation uses a damage-aware heuristic that saves traps for lethal hits instead of wasting them on survivable attacks.
Dashboard & Analytics
Admin and player dashboards give you real-time VP status, battlefield views, analytics charts, and live log monitoring.
How it works
Every game command flows through a 5-stage pipeline. Click a stage to learn more.
Communication
WS events in
Sensor
Validate & queue
Processing
LLM or Algo
Memory
Store context
Action
Execute move
How to access
Players
Log in with your TCG game account, wake a virtual player, and challenge it to a match or watch it play others.
Administrators
Deploy and monitor all VPs, configure personas and engines, review analytics, and manage keeper bots.
Researchers
Explore battle analytics, JSONL event logs, win-rate trends, and token usage data across all VP sessions.
Terms of Use
The VIP Research Group is led by Dr. Maiga Chang at the School of Computing and Information Systems, Athabasca University. This virtual player platform is one of the research group's projects, with follow-up research planned for the future.
Almost all of Dr. Maiga Chang's work is open access or open source. This service runs on a self-sponsored server, like other research projects listed among Dr. Maiga Chang's advanced projects. It is intended to stay online and accessible as long as operating costs are covered.
If access volume becomes too high, or if a business uses the service for profit, the terms may change to include donations, licensing, or subscription models.
About Us
Our Mission
We build autonomous game-playing agents for education research, exploring how AI can serve as practice opponents, tutors, and engagement tools in game-based learning.
Our Supervisor
Dr. Maiga Chang is a Full Professor in the School of Computing and Information Systems at Athabasca University, Canada, and leads the VIP Research Group.
Research Goal
This project studies how LLM-powered and algorithm-driven virtual players can autonomously participate in a real-time card game with distinct play styles and natural conversation.
Our Team
Dr. Maiga Chang
Supervisor
Full Professor, School of Computing & Information Systems, Athabasca University. Leads the VIP Research Group.
CM
Chater Marzougui
Developer
Lead developer of the 5C bot engine, algo engine, dashboards, analytics pipeline, and production platform.
YBS
Yassine Ben Salah
Developer
Contributing developer for bot implementation, testing, and game integration.
Publications
Publications for this project will be listed here as they become available.
FAQ
What is a virtual player?
A virtual player (VP) is an autonomous AI agent that connects to the TCG game server over WebSocket, logs in with real game credentials, and plays full matches without any human intervention. It places cards, attacks, uses magic, and activates traps on its own.
How do I challenge a virtual player?
Log in with your TCG game credentials on the Player dashboard, then wake a VP. Once it's online, create or join a battle room in the game and the VP will accept match requests and play against you automatically.
Is it free to use?
Yes. The platform is a research project by the VIP Research Group and runs on a self-sponsored server. It is intended to remain free and accessible as long as operating costs are affordable.
What AI models power the bots?
VPs can use commercial LLM APIs (Gemini, OpenAI, DeepSeek) for contextual game decisions, or the built-in algo engine that uses scoring and beam search at zero API cost. The algo engine supports 5 persona profiles with different play depths.
Can I contribute?
The project is maintained by the VIP Research Group. If you're interested in contributing as a researcher, developer, or tester, reach out to the supervisor or the development team through the university.