A game-theoretic based dynamic stock market modeling & solution

In this paper, we examine the applicability of genetic algorithm (GA) for solving a non-cooperative stock-market game. We formulate a repeated game where some agents compete with one another to maximize their expected profit/wealth. Each agent estimates the future prices of the trading stocks in order to maximize its expected profit function. Each agent has two actions (selling its own stocks or buying other agents´ stocks) for maximizing its expected profit function over time. Our stock-price model is linear, stochastic, and controlled by the agents´ actions. We use the GA algorithm for each agent to optimally select its actions (selling or buying stock) in the market. To test our algorithm, we simulate a 4-player stock-market game for one hundred rounds.