In international economics and trading, the exchange rate is important. Forecasting theexchange rate helps in minimizing risks and maximizing profits. The study attempts to test threemodels to forecast EUR/USD exchange rate. Based on previous work by Meese & Rogoff(1983), we replicated the authors work of the Random Walk model of AR(1) on different periodand currency to check if the model was able to forecast the exchange rate. Then we ran ARDLand XGBoost models to find which of the two models performed better than the Random Walkmodel based on different measures. The measures are Root Mean Square Error (RMSE) andDirection Accuracy. First difference was taken to remove unit root from ARDL and XGBoostmodels. Adding lags to the independent variables in addition to the dependent, AR(1), generatedbetter outcome. This is an indication of how variables take different periods to affect thedependent variable. Moreover, choosing leading indicators like economic policy uncertaintyindex and Euro futures helped the model in forecasting. The result showed that ARDL andXGBoost models were able to forecast two periods prior the event took place, while the RandomWalk model was not able to forecast rather it replicated the previous period, i.e., it was lagging.