The aimed at investigating the effectiveness of indirect coded CF and indirect non-coded CF on improving grammatical accuracy at the fourth semester students of English Literature Study Program Flores University in 2013/2014. This research was quasi-experiment with pre-test-post-test non-equivalent design. As the sample, 54 students in three intact classes with 18 students of each class were randomly assigned to three groups; indirect coded CF, indirect non-coded CF, and control groups. Free writing was the major source of data. The scoring scheme adopted based on the five categories; content, organization, vocabulary, language use, and mechanics. The instrument reliability was assessed by the inter-rater technique. ANCOVA was to analyze data continued by post-hoc Scheffe. The upshots denote that (1) indirect-coded CF and indirect non-coded CF have a significant difference of effectiveness on grammatical accuracy in the second draft with self editing strategy: (2) indirect coded CF is more effective than indirect non-coded CF, and that of control group in the second draft with self-editing strategy; (3) indirect coded CF and indirect non-coded CF have a significant different in a new draft; (4) indirect coded CF is more effective than indirect non-coded CF, and that of control group in a new draft