This paper presents the NICT’s machine translation system combining neural and statistical machine translation for the WAT2018 Myanmar–English translation task. For both translation directions, we built state-of-the-art statistical (SMT) and neural (NMT) machine translation systems and combined them to improve translation quality. Our NMT systems were trained with the Transformer architecture using the provided parallel data. Our systems combining SMT and NMT are ranked first for this task according to BLEU. This paper also describes the impact of using a small quantity of back-translated monolingual data.