A recent study by Lee and Shin (2022) suggests that changes in the swap basis, defined as the difference between the cross-currency and interest rate swap rates, can predict the one-week ahead changes in the Korean Won-United States Dollar exchange rate. In this study, we propose using the cross-currency and interest rate swap rates as separate predictors, which corresponds to the unrestricted version of the swap basis model. The predictive power of the swap basis may not be stable depending on foreign exchange market and economic conditions, in which case the unrestricted model can better predict exchange rate changes. The unrestricted model shows superior performance in both in-sample and out-of-sample tests, and this result is robust when controlling for potential contemporaneous effects of the swap basis and other instruments, as the predicted variable instead of the original FX return. Our results are also consistent when we use daily and monthly data. In a monthly horizon, the cross-currency swap rate loses its predictive power and the interest rate swap rate tends to be a dominant predictor, which again makes the unrestricted model a better predictive model.