The aimed to look into the most common subject-verb agreement errors in nonfiction and fiction works written by Indonesian secondary 2 students for the Cambridge Checkpoint writing exam. A corpus-based research design is used in the study, which allows for large sample sizes and authentic language analysis. The sample consisted of 20 students (10 from each category) chosen at random using stratified sampling. The study employs manual analysis to identify subject-verb agreement errors in writing samples, followed by a coding scheme to classify the errors. The findings indicate that subject-verb agreement errors are more common in nonfiction works than in fiction works. In both categories, the most common type of error is the use of a singular subject with a plural verb. Other common errors include using a plural subject with a singular verb, missing auxiliary verbs, and incorrect agreement when compound subjects are present. Subject-verb agreement errors were also found to be more common in Paper 1 (essay writing) than in Paper 2 (directed writing). These findings have significant implications for language teaching and evaluation. Language teachers can use the findings to develop targeted pedagogical approaches to help students improve their writing skills, while examiners can use the findings to improve assessment criteria and identify areas that require additional attention. The study also emphasizes the utility of corpus-based research designs for analyzing language errors in a real-world context.