The main objective of this Thesis was to develop a real-time optimization algorithm for tournament poker. The main difference between cash game and tournament poker is that in tournament poker the amount of chips only represents the expectation of the monetary value. The subject of the optimization was a special kind of tournament poker called Super Turbo Sit’n Go. The complexity of this game is the least compared to other poker variants. The appropriate assumptions make real time optimization feasible. The optimization was achieved with the help of fictitious play algorithm which is an iterative learning process for solving games. I have implemented a test environment which makes it possible to test the performance of the algorithm versus static and dynamic strategies via Monte Carlo simulations. During the analysis my aim was to reach the worst case performance hence selecting only the prime opponents for further simulations. With this method the return on investment of the algorithm in worst case was approximately 11%. I have also implemented a decision support system for Full Tilt poker client program. This decision support system is able to extract all the relevant information from the client program and make the optimal suggestion real time. During the development of the support system I used a stealth method to avoid the possibility of detection. I have tested the system on play money tables. The return on investment was 54% on a non-representative amount of samples.