Reinforcement learning
How AlphaVerse learned to play.
During my university work, I wanted to see how a learning agent’s decisions would change after repeated simulated games. I created AlphaVerse to explore that question and shared the work under my name MrBtcgGambler.
Why use a game?
A simulated blackjack hand gives an agent a limited set of choices and a clear outcome. I could run the environment repeatedly, watch the decisions it made and compare what happened as it gained experience.
This is reinforcement learning. The agent acts, receives feedback and adjusts its policy for the next decision. A simulation made that cycle visible to people who were new to the subject.
Opening up the experiment
I developed AlphaVerse beyond the university project and published code and demonstrations. People following the project could inspect the work, compare strategies and discuss how the agent behaved.
MrBtcgGambler is the name I use for these independent experiments. The public repositories and demonstrations give readers a way to explore the work for themselves.