Gaz is a plug and play system where you provide a brain for her to play tetris.She is sery selfish and has no remorse.

##About Tetris has always been an addiction of mine. I decided to eradicate this obsession by "solving" it so I'll never want to play it again. In case no one gets the reference.

###Prerequisites The only dependency for this repo is pygame and matplotlib if you want to see pretty graphs.

###Installing

If you have pip

pip install pygame

NOTE: matplotlib installations are weird... see this for help ---> link

or if you don't

sudo apt-get install python-pygame
sudo apt-get build-dep python-matplotlib

Currently there are 2 ways to use Tetris with Gaz.

1. Normal Tetris.

   By running python tetris.py you can play normal tetris as Alexey Pajitnov intended.

    Controls:
    LEFT ARROW - moves piece left
    RIGHT ARROW - moves piece right
    DOWN ARROW - moves piece down
    UP ARROW - rotates piece clockwise
    LEFT SHIFT - Gaz takes over
    ESC - ends the game
   

   If you want to record your games as human data for Gaz to look at or for personal use run python tetris.py -r [name or file] and it will save them to Gaz/gameplays.

2. Playing with Gaz

   By running python tetris.py -gaz -greedy Gaz will automatically starty playing using her simple greedy algorithm. You can always switch out of "Auto" mode by pressing LEFT SHIFT.

   By running python tetris.py -gaz -degreedy 0 3 Gaz with use a deep greedy algorithm. '0' is the depth, and '3' is the top_n skimmed moves off of each layer. Keep the depth at either 0,1,2 because the branching factor in tetris is to big and advanced pruning methods need to be used to reduce the number of moves to search through.

   A good deep greedy setting in python tetris.py -gaz -degreedy 2 2.

   The two other usages are python tetris.py -gaz -naive semihumanmodel and python tetris.py -gaz -knn which use naive bayes and K nearest neighbors respectively. They don't work too well but if you ca build a model that will make them work then please give me that model so I can analyze your gameplay to see what magic dust you have.

   Advanced Usage:

   By running python metric_tetris.py you can have Gaz play multiple games at once over many games, build summaries, and eventually will learn and get better at playing tetris.

   To build DataSets and Models edit build_model.py script to save and build models. Check Gaz/game_reader.py for the code that builds models.