In this directory I develop and implement the card game Skyjo with python to play with a computer.
The aim of the game is to finish with the least number of points. When there are three equal values in a column, they all vanish. When one player finishes opening all cards, the others may take one more card. Then the numerical values of the cards are added together. Caution, when the first player to finish does not have the smallest number of points his points are then doubled. This is one round. Several rounds are played until one player has at least 100 points. The player with the smallest number of points wins. The detailed actions are explained in the GUI. For playing Skyjo against a computer it is necessary to download the following files and to place them in one directory: skyjo_human2_games.py xgb_model1_column2.txt xgb_model2.json
The computer needs to have the standard python libraries (check what is imported in the top of skyjo_human2_games.py) and also xgboost and simpleguitk.
Then just run python skyjo_human2_games.py and the rest are mouse clicks in the GUI.
Besides this prefreed version to play skyjo, single rounds can also be played in skyjo_human_developing1.py In skyjo_human2.py the newest levels are added, but it is not improvement, see below. It is needs also xgb_model3.json
Here I mean the implementing and the functionality of creating the game. This includes applying machine learning results. It is developed, tested and partly also run for output skyjo_exploring.ipynb, skyjo_developing.ipynb and skyjo_developing.ipynb in that order, the last in parallel with skyjo_human_developing1.py because the GUI works better outside a notebook The skyjo functions are also in skyjo_functions1.py, skyjo_functions2.py, skyjo_functions3.py, skyjo_functions4.py The most recent are in the latter which is used most often.
The first explorative machine learning is in machine_learning_skyjo1.ipynb. The first part where you can see the machine is actually learning is in machine_learning_skyjo2.ipynb. In order to achieve this result, functions are developed which are stored in ml_functions1.py and ml_functions2.py (the newer) version. For the next machine learning step more data is produced with produce_data_for_machine_learning.py It is used in machine_learning_skyjo3.ipynb. The output of it is analysed in machine_learning4.ipynb. It does not improve the performance both against the computer nor against human, the reason is likely that Ito mainly learns to predict the choice of the algorithm used by level 1 and not which option is actually better. Future games of human versus the computer are analyzed in analyzing_human1.ipynb. None of them lead to improvements compared to the first improved. I am now working on improving the computer against other computers using new methods in optimization_developement.ipynb