Introduction
Create your own AI Pokémon Trainers and make them battle with randomly generated Pokémon like creatures. In the VGC AI Competition, you make your AI agents command the Pokémon in battle and at the same time they also must decide the strongest team against the current Meta. You can compete in one or two tracks:
1) Battle Track: Here, the Pokémon and teams are randomly generated at each battle, so your AI Pokémon Trainer must know how to command any unpredictable team. For fairness, the teams of both players are switched during the rounds.
2) Championship Track: Here the Pokémon roster in randomly generated at the beginning of the competition and besides battling is the task of your AI Pokémon Trainer to choose the best team against the current game like Video Game Championships
Logistics
Competition Entry:
The submissions will be made using a specific google forms to be published like previous years in the competition site wiki and discord community. Last edition also started collecting opinions of the entries to improve the competition with feedback.
Winner Criteria and Publishing Results:
The winners will be announced both on the website plus the discord community as previous years. In the case of the Battle Track, the winner is determined automatically by performing a tree championship where only one prevails, allowing also to determine a 2nd place and also a 3rd place from a loser bracket.
The Teambuild track follows a similar structure, but instead of a tree-format championship, it ill run in rounds. In each round, player’s matchmaking of opponents is made and their ELO rating accumulated. The winner will be the one with the greatest rating.
Prizing:
We’re considering submitting again for the IEEE CIS Education Competition Fund, which VGC AI was accepted for prize funding last year of a total of 1000$. Vortex-CoLab will assist in funding or co-funding the prizes.
Barrier of Entry:
The VGC AI framework is written in the Python programming language, which we think will ease the barrier of entry for any level of computer or data science. The competition allows from beginners to masters, beginners can implement from basic rule-based or search algorithms with ease from masters’ tailored ML-based agents. There are many tutorials in the official wiki page, and the framework itself contains many examples as well.
Participants
The target audiences is first and foremost GameAI researchers, game developers/designers, last but not least computer science students and hobbyists. We will try to reach national Portuguese and international institutes via email. X/Twitter account is being considered.
Timeline
Competition will start at 1st of February 2025, where acceptance of submissions will start. Submission deadline: 30th June 2025. (may be rectified if it is too late for finalists to join) Result announcement: TBA (will be done during the conference). As per advice, we will pre-announce the finalists in the discord community if they wish to have the opportunity to join CoG25.
Organizers
Simão Reis, Research Scientist and Invited Professor Vortex-CoLab and LIACC/Faculty of Engineering of University of Porto
simao.reis@vortex-colab.com/simao.reis@outlook.pt
Obtained the Ph.D. in Automatic Game Balance with AI in 2024, is a research scientist in AI and Cybersecurity, gives computer science lectures in the University of Porto, has organized the previous two editions of the VGC AI Competition @CoG23/24, addressing automatic meta-game balance over Esports-like environments, in this case over a Pokémon-like.