WAT.ai Tetris

Tetris is a classic video game having first been released for the Electronika 60 in 1984. Your aim is to build a Python bot to maximise your score in Tetris by clearing lines of Tetrominoes. There is a wide range of interesting strategies you can employ – how well can you do? 👀

Getting started

To get started, clone and follow the instructions in our getting started repository:

git clone https://github.com/DoxaAI/tetris-getting-started

Alternatively, you can download the getting started materials directly as a zip.

Fun fact: a move clearing four lines using an I-shaped piece is known as a Tetris, which is where the name of the game comes from.

Evaluation

Your agent has a time limit of 4 minutes to achieve the highest score in Tetris possible. There is also a block limit of 1000 pieces per game – can you reach it? 👀

Clearing a single line, two lines, three lines and four lines score 100, 250, 750 and 3000 points, respectively.

We will run your agent three times with different preset seeds and average the results to compute your final score. You can view the games your agent played by clicking on their respective results on the scoreboard.

Gameplay

This version of Tetris implements the Standard Rotation System for the game. As such, pieces can rotate provided there is space to do so, and if this would cause pieces to rotate off the edge of the board, they will get pushed out to stay on the board, provided there is no piece in the way.

This version also does not allow you to look ahead at what pieces are upcoming, so you can only make decisions based on the current falling piece.

Pieces fall after each action returned by your agent is performed. This means you have one last chance to move or rotate a piece just before it lands.

If an action is invalid (such as trying to move left while the current piece is up against the left wall of the Tetris board), it will be treated as a NOOP, and the piece will just fall one position.