stupid-horse

A bot that plays standard/960 chess on Lichess, all time controls. Still a work in progress!

How it works

• Set up a Lichess bot account
• Make an .env file with bot token and ID
• Run the bot (go run .), and it listens for incoming challenges and ongoing games.

To do

• Make the bot care about time controls. Right now, the bot thinks at a certain depth no matter the time left.
• (Possibly) create a web interface to look at bot evaluations in live-time

Known issues

• Bot does not put importance to sooner checkmate patterns. In this game, it played into a draw by repetition instead of a mate in 1.
• Bot cannot convert clearly winning positions (examples: one)
• Bot sometimes doesn't hear opponent move notifications causing it to wait for a move forever (needs to be restarted to fix)

My journey

Why was this bot created?

• I enjoy chess
• I have always wanted to create a chess bot
• I saw it as a daunting challenge as a child and now I am equipped with the skill and experience to avenge myself
• I saw this as a nice exercise with a lots of room to learn

My process

• Browsed through Go tutorials
• Created functions and structs to parse FEN, and set the logic involving board, pieces
• Set up move searching to get all valid moves for a given position (state of chess board in a game)
• Wrote tests and benchmark functions along the way to look for optimizations
• Implemented the minimax algorithm (with α-β) to choose best move after considering all possible outcomes at a certain depth (e.g. 3 moves ahead)
• Hooked up the Lichess API and made the bot accept challenges and play moves!
• To be continnued...

Things I learned

• Coding in Go! I found that I really enjoy Go due to its readability, strong typing, intuitive keywords, built-in test writing, etc...
• Writing modular components and tests & benchmarks, making it easy to add new features and optimize parts of the project
• Consuming an API in Go, especially handling streams in ndjson format
• The fastest stalemate from the standard starting position, researched while creating tests for game state
• Honestly a lot more that I'd love to chat about