A data-driven approach for identifying winning strategies in trading environments
This project is a collection of ideas on how we can identify behavioral patterns for reaching our goals in a specific context. Since this is a very broad description, the starting point for this project is a game scenario. For getting started quickly and to be more hands-on, the first idea we are want to focus on is a concrete example.
Imagine the following scenario. We are invited to a party where a game is played we never played before. Since we are open-minded, we are not hesitating to join. The other folks are used to the game and give you basic information about it. For instance the initial state of the game, the actions and the final goal. Since we are not used to play a lot of games and probably the fact that we already had a couple of drinks, we only remember the following.
- The final goal is to maximize the number of chips at the end of the game
- The game ends after 20 rounds or when we ran out of chips
- We are starting the game with 100 chips
- In each round we are able to buy and sell assets
- Each asset has an individual price, which is about to change in each round. However, we are not aware of the dynamics of the prices
- We only know the current price of the assets
- There is a fee for buying and selling assets
- We can buy and sell multiple assets in each round
- At the end of the game, all assets are sold and the resulting number of chips decides about our performance