Quick update.
After playing Tetris, I realised that the way people play Tetris usually involves making lines whenever possible. This means that the agent would benefit from knowing if a gap that can be filled is the last gap in the line. This would require another variable stating whether the row is nearly full.
I feel that the strategy would work whether or not this variable was there, but the variable would make the agent better. Something for investigation.
Edit: The agent would also benefit from putting pieces next to walls when in the first stages of the game (or any stage with a flattish surface).
Edit: Playing the game you are hoping to create something for makes it so much more involving! Placing pieces next to walls and to complete lines should have higher artificial rewards. By artificial, I mean these rewards would temporarily influence the agent to put the piece there, depending on the playing field at the time.