We developed a method for the early on-site detection of strawberry anthracnose using a portable Raman system with multivariate statistical analysis algorithms. By using molecular markers based on Raman spectra, the proposed method can detect anthracnose in strawberry stems 3 days after exposure to Colletotrichum gloeosporioides. A fiber-optic probe was applied for the portable Raman system, and the acquisition time was 10 s. We found that the molecular markers were closely related to the following subjects: i) an increase in amide III and fatty acids of C. gloeosporioides invading strawberry stems (Raman bands at 1180-1310 cm^-1) and ii) a decrease in metabolites in strawberry plants, such as phenolic compounds and terpenoids (Raman bands at 760, 800, and 1523 cm^-1). We also found that the increased fluorescence background caused by various chromophores within the invading C. gloeosporioides could serve as a marker. A two-dimensional cluster plot obtained by principal component analysis (PCA) showed that the three groups (control, fungal infection, and pathogen) were distinguishable. The linear discriminant analysis (LDA)-based prediction algorithm could identify C. gloeosporioides infection with a posterior probability of over 40%, even when no symptoms were visible on the inoculated strawberry plants.