Abstract:
We consider the position management problem for an agent trading a mean- reverting asset. This problem arises in many statistical and fundamental arbitrage trading situations when the short-term returns on an asset are predictable but limited risk-bearing capacity does not allow to fully exploit this predictability. The model is rather simple; it does not require any inputs apart from the parameters of the price process and agent's relative risk aversion. However, the model reproduces some realistic patterns of traders' behaviour. We use the Ornstein-Uhlenbeck process to model the price process and consider a finite horizon power utility agent. The dynamic programming approach yields a non-linear PDE. It is solved explicitly, and simple formulas for the value function and the optimal trading strategy are obtained. We use Monte-Carlo simulation to check for the effects of parameter misspecification.