In Taiwan elementary schools, Scratch programming has been taught for more than four years. Previous studies have shown that personal annotations is a useful learning method that improve learning performance. An annotation-based Scratch programming (ASP) system provides for the creation, share, and review of annotations and homework solutions in the interface of Scratch programming. In addition, we combine the ASP system with the problem solving-based teaching approach in Scratch programming pedagogy, which boosts cognition development and enhances learning achievements. This study is aimed at exploring the effects of annotations and homework on learning achievement. A quasi-experimental method was used with elementary school students in a Scratch programming course over a 4-month period. The experimental results revealed that students' thoughts and solutions in solving homework assignments have a significant influence on learning achievement. We further investigated that only making annotations in solving homework activities, among all other variables (the quantity of annotations, the quantity of one's own annotations reviewed, the quantity of peers' annotations reviewed, the quantity of one's own homework solutions reviewed, and the quantity of peers' homework solutions reviewed), can significantly predict learning achievements.