The debut of the “rolling” technique
After a moment of stasis in the comparison between video games and enthusiasts, the big news arrived in 2020, when a player combined a multi-finger technique originally used in arcade video games with a finger positioned at the bottom of the controller to push it against another finger in the upper part. Called “rolling,” this much faster approach helped one player reach level 95 in 2022.
Then other obstacles arose. Because the developers of the original Tetris never intended for players to push the limits of the game so aggressively, oddities began to occur at higher levels. A particularly difficult problem arose with the game’s color palette, which traditionally runs through 10 easily distinguishable patterns. Starting at level 138, however, random color combinations began to appear, some of which made the blocks much more difficult to distinguish from the game’s black background.
StackRabbit to detect “anomalies”
Two particularly fiendish patterns – one combination of dark blue and green, later dubbed “Twilight,” and the other made up of black, gray and white blocks called “Coal” – proved particularly challenging for players. Add to that the strain of increasingly long matches, which could last 40 minutes or more, and progress slowed even further. It took a Tetris-playing AI program called StackRabbit to break the logjam, helping players pinpoint where they might encounter an anomaly that led to an elimination screen, and ultimately beat the game.
StackRabbit, which managed to get up to level 237 before crashing, used a modified version of Tetris, so its results aren’t strictly comparable to those of human players. Furthermore, his findings are not immediately applicable to the game played by humans. However, the results clearly demonstrated that game errors can be triggered by very specific events, such as blocks in play or the number of lines cleared at once.
In this way, human players were tasked with mapping out all the possible scenarios that could cause these crashes in the original game. These typically occurred when ten-year-old game code lost its position and began reading subsequent instructions from the wrong position, typically generating garbage input. StackRabbit’s experience led to the compilation of a large spreadsheet detailing the game levels and specific conditions most likely to lead to a crash.