This study develops an empirical model to forecast daily exchange rate dynamics by separating market participants into two groups: short-term speculators and longer-term investors. These two types of participants have different expectations and impacts on daily exchange rate movements. The proposed model is based on a chartist-fundamentalist approach, i.e., the expected daily exchange rate movement is influenced by its past movement as well as the extent to which the market rate deviates from its fundamental value. Additionally, data for some global risk factors, FX market intervention and Quantitative Easing (QE) are also incorporated into the model as control variables and to consider policy implications. The model is specified as a two-state Markov switching model due to its empirical support for in-sample prediction. This study examines the daily exchange rate movements of the five most traded currency pairs from 1999 until the first half of 2013. The proposed model logically and significantly explains daily exchange rate dynamics and outperforms a random walk model for both in-sample and out-of-sample periods. The model implies that monetary authorities can implement policies to potentially stabilize short-term exchange rate movements by influencing the expectations of short-term speculators. In contrast, the expectations of longer-term investors are quite stable and tend to converge to the trend and long-run equilibrium in the foreign exchange market without the need for short-run policy interventions.