In this thesis, we present one of the approaches to the problem of hedging in incomplete financial market models in discrete time; the focus is on minimizing the quadratic hedging error. In the first chapter, we introduce fundamental concepts used to define the financial market model in discrete time. In the second chapter, we present a local version of the minimization problem. We view the problem as a sequential regression procedure and we develop a process of recursive construction of a locally risk-minimizing strategy. The solution to the problem also involves an orthogonal decomposition of a given contingent claim. Furthermore, in the third chapter, we study the minimal martingale measures. An important characteristic of such equivalent martingale measures is that the value process generated by a locally risk-minimizing strategy can be described as the martingale of conditional expectations of the given contingent claim. We also introduce the concept of the remaining conditional risk. Finally, in the fourth chapter, we analyze the problem of variance-optimal hedging where we try to minimize the global quadratic hedging error. We show that the solutions to the local and global problem coincide if the underlying measure is itself a martingale measure.