This study aims to compare two MT outputs in terms of fidelity in translating compounds and collocations. The objectives of the study are to identify the common errors of GT(version 6.3.59158) and ST (version 8) outputs in the translation of collocations and compounds and to compare the outputs of GT and ST in terms of fidelity in translating collocations and compounds. Three source texts were selected and used in the study, the UN Position Paper on Education Post, 2015, a number of sentences with various expressions of compounds and collocations, and the story "The Open Window" written by the British writer Hector Hugh Munro. These source texts were selected for their different text-type, rich content of various structures of compound and collocations. The three texts were translated by GT and ST systems illustrated in tables, then the translated output was assessed by applying the criteria of Hutchins and Somers (1992) and Izwaini`s model(2006) to find the problems, lexically and syntactically, in translating compounds and collocations. Then, the Back-Translation technique was used for 70 samples of the translated output in order to examine the fidelity factor as proposed by Hutchins and Somers, then the data was classified into three categories, zero fidelity, minor fidelity, and major fidelity. The data was also measured for the fidelity rate in each MT output through the T Test to…… Besides that, the selected output was also tested by the investigator triangulation method to add more reliability and validity to the findings and to determine which MT output held a more fidelity rate. The findings indicated that the GT system output had fewer errors and problems in translating compounds and collocations than the ST system. The GT system also scored a higher rate of fidelity as it showed 77.1% of zero fidelity, whereas the ST system showed 14.3%. of zero fidelity, which determines that GT system output held a more fidelity rate.