By the end of 2026, will AI have empowered me to build a bot that can reach the top 500 in the current gen of OU?
Mini
4
Ṁ51
2027
38%
chance

In July of 2021, Youtuber The Third Build released a video detailing how he built an AI that could play competitive Pokemon. To my knowledge, his was the most successful Pokemon AI to date. His AI maxed out at an Elo score of 1630, whereas an Elo score of 1701 was required to reach top 500 at that time.

Since watching that video, I have often thought that it would be a fun project to try making a Pokemon AI of my own. Unfortunately, replicating The Third Build's project is beyond my current programming abilities.

When GPT4 came out, I tried to use it to guide me through such a project. I couldn't replicate The Third Build's project, but I was able to have some success using the poke-env interface to build some very simple bots. The bots were very bad.

As more and more capable LLMs become available for public use, I will revisit this project.

Resolves as YES if, before the EOY 2026, I have developed (without paying for human assistance) an AI that manages to place within the Pokemon Showdown top 500 leaderboard for the then-current generation of OU, with at least 20 games played.

Resolves as N/A if, before resolving YES, some change is implemented to Pokemon Showdown which prevents such a project from succeeding (e.g. the site shuts down, there is no leaderboard, bots are quickly detected and banned, etc.) and the change is not reverted before the resolution date.

Resolves as NO otherwise.

Get
Ṁ1,000
and
S1.00