Skip to content

Latest commit

 

History

History
42 lines (32 loc) · 2.83 KB

README.md

File metadata and controls

42 lines (32 loc) · 2.83 KB

Chess Endgame Training

This app was created as part of the Tech with Tim August Timathon challenge. The goal of the project is to create a lightweight application that can generate chess positions for finding checkmate in a certain number of moves.

To view the fully deployed application on Heroku, you can connect to the link here: https://chess-endgame-trainer.herokuapp.com/


Data Creation

For more complete explanation on this, see the data folder. The majority of the work that went into this project was based on having to take chess games and parse out positions that were mate in a certain number of moves. Because of the large decision tree in chess, I could not find a fast way to generate this data ad hoc. Instead I used the open database from lichess.org and used games from actual players as the basis for the database.

To save on disk space, only 10k positions for each move option was kept. This results in a database of 100k checkmating patterns, which should be sufficient for most people to not see a duplicate position for a long time.


How the App Works

Since the majority of the work was pre-processing the chess positions, the generation portion is a simple query to an sqlite database:

con = sqlite3.connect("data/db.sqlite3")
# gets a random chess position from the database
sql = """
	SELECT *
	FROM data
	WHERE moves {}
	ORDER BY RANDOM()
	LIMIT 1""".format("> 0" if mate_in_x == 0 else f"= {mate_in_x}")

# pandas is just so convenient for these queries, even if it is overkill
df = pd.read_sql_query(sql, con)

The visualization portion of the application uses the Dash library from Plotly. I like this because it allows me to keep the entire application (even the components based on React.JS) in python, where I'm most comfortable. The application is broken down into three pieces:

  1. The FEN (chess position) text - this component is hidden, but used as a reference by the next two components.
  2. The selection of number of white moves to checkmate - this component is a dropdown option for any number of different positions.
  3. The resulting image, based on the selection, with randomness.

Ideas for improvement

The first thought that comes to mind is simply expanding to a larger database of positions. Currently, there are only 100k possible positions to generate, and there should be many, many (incalculably many) more.

A second idea for improvement is to actually allow the user to play the position. This would require a much greater understanding of JavaScript than I have and goes beyond the scope of this project anyways.

Lastly, and somewhat tied to the second improvement, is the ability to give a hint for the first move to make to follow through on the checkmating pattern.