Top worldwide with social-engineering and a cheat that's still undetected (www.ud2.rip)

A group of developers created a cheat for vertical scrolling rhythm games (VSRGs) like osu!mania and Stepmania using mathematical principles and reinforcement learning (RL). Their cheat utilizes a trained model to accurately predict scrolling notes, allowing it to produce perfectly timed keystrokes without any game-specific coding. This approach led to a top 20 global ranking on osu!, but the developers were banned due to suspicions of multi-accounting rather than concrete proof of cheating, revealing that fooling the community can be more challenging than evading anti-cheat software. This project highlights significant implications for the AI/ML community, combining computational techniques with sociological insights. By analyzing play data of top players, the developers modeled human-like timing variances and error patterns, which are crucial for avoiding detection. They developed a system that adjusts keystrokes to mimic human behavior, utilizing a feedforward neural network to refine their results. As a result, their work not only explores the technical limits of game cheating but also provokes a discussion on the ethics of machine learning applications in competitive gaming and the broader community's scrutiny in maintaining integrity.