I want to start to code Deep Reinforcement Learning (DRL) models with Tensorflow and I need a fun task to begin with.

98 Cards is an Android game that needs strategy.

Your deck contains 98 cards ranging from 2 to 99. Distribute all of these cards on four different piles - one after another. The two piles on top need to be in ascending order, the two piles at the bottom need to be in descending order. To make it a little easier: if the difference between a card and a pile is exactly 10 you can put that card on that pile, the order doesn't matter. Use this rule to "shrink" your piles! Now try to get rid of as many cards as possible!

So the goal of this game is to collect as much score as possible. When everytime a player piles a card, a score will be given. But I fail to figure how how they assign the score. Therefore, I just assgin one point to each successful placement and the goal is to pile as many cards as possible.

• env.py: the game simulator.