Lucki

| /$$ /$$ /$$ |
| | $$ | $$ |/ |
| | $$ /$$ /$$ /$$$$$$$| $$ /$$ /$$ |
| | $$ | $$ | $$ /$$/| $$ /$$/| $$ |
| | $$ | $$ | $$| $$ | $$$$$$/ | $$ |
| | $$ | $$ | $$| $$ | $$_ $$ | $$ |
| | $$$$$$$$| $$$$$$/| $$$$$$$| $$ \ $$| $$ |
| |/ _/ __/|/ _/|/ |
| |

| Written by Thomas Schaffer [email protected] in Java. |
| Licensed under the MIT license. |

| Description |

Lucki is a Poker AI agent that I developed for my MSc dissertation.

It is based on some research by the Poker research group at the University of Alberta, notably their bots Poki, Loki, etc. The global architecture of the agent (hand weight tables for the opponents, and an hand evaluator enumerating potential future draws and applying those weights) is inspired by their published research and completely reimplemented from scratch.

This project's contribution was to try and improve on the somewhat limited Betting Strategy component by experimenting with neural networks trained by a genetic algorithm.

This contains the agent, along with a complete framework for poker play and AI agents so that it should be easy to implement other strategies while benefiting from the existing hand evaluation / profiling components.