Near Term Investment Decision Support for Currency Options

Abstract

Currency options on a future confer the right but not the obligation to trade the underlying currency future at a pre-agreed price (the strike). Most currency options trade over-the-counter (OTC). That means that they do not follow standard rules for option contracts. They can, in principle, have arbitrary maturities, strike prices, and volume. Valuing OTC options is complex and various models have been developed for that task. Here, we present a Financial Decision Support System that helps in gauging the probable price of a curreny option within the next thirty minutes. The forecast horizon is continuous and also an input to the model. We train an artificial neural network on past price data, using tick prices of option and underlying future of the past two hours. Fuzzy coding of time-points increases the robustness of our model. Error on the out-of-sample data set is small. Interestingly, we do not need additional data to obtain satisfactory results. The tick data time series of option and future prices contains enough information to lead to a good forecast. Especially, we do not use volatility or interest rate data. The neural network is, in most cases, able to correctly forecast the main troughs and peaks of the following thirty minutes. An ensemble of neural networks further smoothes the result. Our Financial Decision Support System is useful for both sellers and buyers of options. Sellers get a tool that is calibrated on the latest market data. Buyers can judge, whether they should hedge their risks now or rather wait a short amount of time.