This report describes the development of a superhuman sports multiplayer game that makes use of augmented reality through the HoloLens. The game is a reenvisioning of the original League of Lasers, a pong/football-esque game for Android phones using camera tracking and special headgear for player positioning. The game was specifically redesigned to use neither a camera or a phone, but instead use the HoloLens' spatial localisation features. The goal of the project is to provide a superhuman sports multiplayer experience on the HoloLens. To be able to provide this, a system for managing spatial anchors and a custom networking architecture were developed, making use of a game server, game clients, a master game client and a web server for sharing anchors. Along with this, large-scale playtests were performed, where data was collected. The corresponding game design changes from the original game are described and implemented to ensure an optimal experience. As the game was developed in Unity with the Mixed Reality Toolkit, software quality was kept at adequate levels throughout the project through a custom CI setup, code written with testability and maintainability in mind through the humble object pattern, and rigorous PR reviewing guidelines. For the development process itself, Scrum was used with weekly meetings, a Waffle project board and daily status reports. Numerous problems with the combination of Unity, the Mixed Reality Toolkit and the HoloLens made the project a laborious endeavour, but the final result and design of the game exceeded our initial expectations.

Normalizing flows have demonstrated their ability to learn complex and high-dimensional distributions. However, the behavior of normalizing flow likelihoods are not yet fully understood, particularly when exposed to outlier data, where it has been observed that large likelihoods are often assigned to inputs that are substantially different from the training set. To better understand the likelihood behavior and outlier detection capabilities of normalizing flows, we analyze a more restricted version of the model using synthetic test data from parametric distributions, allowing access to the density of the underlying distribution.