Classifying games as test environments for RL/psychometrics? Figuring out DOM/event handlers from video data of video games?

Oh A LOT of resources, may have to break this down. I’m most interested in using video games for quantified self, though a lot of gaming APIs come initially from RL AIs on them

x.com for a lumosity game or game not easily in their database?

https://openai.com/research/neural-mmo

https://openai.com/research/gym-retro

https://openai.com/research/openai-baselines-ppo

Joseph Suarez has some packages + reading group

https://openai.com/research/openai-gym-beta

https://openai.com/research/scaling-laws-for-reward-model-overoptimization

x.com (he will update his tutorial to use codebases way more elegant than tensorflow)

Arthur W. Juliani, PhD (awjuliani.github.io)

https://supermemo.guru/wiki/SuperMemo_Guru

===

PIECES PER SECOND (PPS) in tetris - I know someone who tracks them in response to sleep quality and modafinil

Strooper | Andy Kong => HE is totally the essence of the quantified self

1 Like

You really need to write out in full sentences what you want to say and what is on the other end of those links.

That’s a really interesting mix of resources — definitely a lot to unpack there. Using video games as a tool for quantified self is a fascinating idea, especially since they naturally generate rich behavioral data. Tracking metrics like PPS in Tetris or reaction times in cognitive games can give some surprisingly good insights into attention, fatigue, and even sleep patterns.

The connection between reinforcement learning environments (like Gym Retro or Neural MMO) and human performance tracking is a cool overlap too. The same frameworks used for AI agents can be repurposed to measure and model human learning curves.

If you’re diving into this area, it might help to narrow your focus first — maybe pick one domain (e.g., cognitive performance, motor response, or decision-making under stress) and build from there. Quantified gaming can get overwhelming quickly, but the potential for self-analysis and experimentation is huge.