Design and Evaluation of Competitive Programming Platforms for Computer Science Education

Competitive programming is the art of solving well-known algorithmic problems in a limited amount of time. It is a fun and challenging activity that requires and enhances problem-solving skills, code efficiency, and algorithmic thinking. While it is commonly practiced in the form of contests, it can also be used as a tool to teach and learn computer science. However, the main contest management systems used to host competitive programming contests are not designed to be used in an educational context and need features that would be useful for teachers and students. This thesis presents two new contest management systems designed with education in mind: Turing Arena light and Code Colosseum. Turing Arena light is a lightweight system that focuses on interactivity, feedback, and ease of use, ideal for exercises and exams. Code Colosseum is a platform that hosts challenges between student-written programs, fostering engagement and learning through game-like competition. Both systems are evaluated through a real-world user experience and participant feedback. For Turing Arena light, the user experience comes from its use in the competitive programming course at the University of Verona for both exercises and exams. For Code Colosseum, the user experience comes from its use in a tournament hosted for high-school and university students. In both cases, feedback from the participants is collected through a survey. This thesis also presents other works concerning competitive programming contests, including a report on technical challenges faced during the 2020 Italian Olympiad in Informatics (OII), analytics of learning progress on the OII's online platform (CMSocial), and statistics from the SouthWestern European Regional Contest (SWERC) editions organized in 2022 and 2023. Furthermore, this thesis showcases two additional research projects unrelated to the thesis's main topic. The first is searching for an AI capable of playing the vertical scrolling shooter game Touhou only by looking at the pixels on the screen. The second one is the encoding of the Disjunctive Temporal Problem (DTP) into Satisfiability Modulo Theories (SMT), Mixed Integer Linear Programming (MILP), and Satisfiability (SAT) models and the evaluation of the performance of the resulting encodings.