Super Mario in the Pernicious Kingdoms: Classifying glitches in old games

TL;DR

This study investigates glitches in four classic Super Mario games to understand underlying software weaknesses exploited by speedrunners. By mapping $237$ glitches to CWE codes using the Seven Pernicious Kingdoms taxonomy, the authors identify seven game-specific weaknesses absent from CWE and map them to $32$ codes across the games. The methodology combines data from the Super Mario Wiki, independent coding, and consensus-building, with threats to validity considered. The findings show that while many glitches align with conventional software weaknesses, several game-specific categories illuminate unique interactions between game state and memory, informing debates about expanding CWE or preserving game jank.

Abstract

In a case study spanning four classic Super Mario games and the analysis of 237 known glitches within them, we classify a variety of weaknesses that are exploited by speedrunners to enable them to beat games quickly and in surprising ways. Using the Seven Pernicious Kingdoms software defect taxonomy and the Common Weakness Enumeration, we categorize the glitches by the weaknesses that enable them. We identify 7 new weaknesses that appear specific to games and which are not covered by current software weakness taxonomies.

Figures (1)

• Figure 1: The Mario games have featured many different characters and items which may or may not be familiar. We show a selection from various games to jog the reader's memory.