This paper investigates the short-time exchange rate predictability in a developed and in an emerging market, and for this purpose we consider the Euro/United States Dollar (EUR/USD) and the United States Dollar/Turkish Lira (USD/TRY) exchange rates. We apply the benchmark test and compare the results of daily out-of-sample forecasting by Brownian Motion (BM), Geometric Brownian. Motion (GBM), Ornstein-Uhlenbeck Mean-reversion (OUM), Jump Diffusion (JD) stochastic processes, Vector Autoregressive (VAR), Autoregressive Integrated Moving Average (ARIMA) models and Uncovered Interest Rate Parity (UCIP) against the Random Walk (RW). We conclude that none of these models or stochastic processes displays superiority over the RW model in forecasting the USD/TRY exchange rate. However, GBM, BM and OUM processes beat the RW model in forecasting the EUR/USD, exchange rate. Furthermore, we show that these findings are robust and not time-specific. When we separately examine the pre-crisis and the post-crisis periods, results remain unchanged.