Informative texts, including those in sport sciences, demand precise factual communication yet pose translation challenges due to specialised jargon, culturally bound idioms, and context-dependent nuances. This study aimed to identify the most recurrent errors in human translations of thesis abstracts in sport sciences using the Multidimensional Quality Metrics (MQM) framework, focusing on a period prior to the emergence of Neural Machine Translation (NMT) and artificial intelligence (AI) (the 1380s Iranian calendar years, approximately 2001–2011). A sample of 15 English abstracts was randomly selected from Persian-language sport sciences theses from that period. Errors were classified according to MQM dimensions, and two independent raters confirmed the identification and documentation of errors. The analysis revealed a fluctuating error pattern between 2005 and 2010, characterised by an initial decline followed by a subsequent rise, indicating temporary variability in translation quality. These findings provide empirical evidence of recurrent translation challenges in pre-NMT and pre-AI academic sport sciences texts, with implications for translation scholars, academic editors, and thesis supervisors.