Impression Baby

Project Description

This is a single-player game inspired by BS poker, where you play with Baby (a reinforcement machine learning algorithm) who learns your strategy over the course of the game.

Inspiration

We wanted to make the opponent "smart", forcing the player to change strategies over the rounds in this simple game.

Rules

• 2 players: you and Baby
• Each player gets 3 cards randomly from the whole deck of cards, you only know your own 3 cards.
• You enter a guess, which includes a quantity between 1 and 6 and a value between 1 and 13 or a suit (e.g. 3 9’s, 2 Spades)
• Baby enters true or false (BS)
• If Baby correctly judged that your guess is included/not included in the 6 cards, it gets 6-x points.
• If not, you get x points.

What's next

• More complex game rules (claim could include flush, full house, etc)
• Each round Baby gets to return a higher claim instead of simply saying True or False

Built With

• keras
• keras-rl
• openai-gym
• python
• pyxel
• tensorflow