Diablo speedrun sleuths proved its 15-year world record was fake after reverse-engineering the game and failing to replicate the run in 2.2 billion possible randomized dungeons

First reported by Ars Technica, a team of Diablo speedrun enthusiasts have decisively proven that a three minute any% world record that has stood since 2009 was faked. The effort required reverse-engineering Diablo's map generation system, which has 2.2 billion possible legitimate outputs.

The offending speedrun was uploaded to Speed Demos Archive by Maciej "groobo" Maselewski back in March 2009, and featured an unbelievably "lucky" sorcerer run with a finishing time of 3:12. The run was segmented any%, meaning groobo could use certain allowed exploits, as well as save the game and redo a portion to get a better time, adding those best-of segments together for the final count. The Diablo mapgen team alleges that groopo's manipulation of the game went far beyond what was disclosed or allowed.

"It seemed very unusual that we would have so many levels with the upstairs and the downstairs right next to each other," said speedrun expert and TASbot keeper Allan "dwangoAC" Cecil about groobo's apparent luck. "We wanted to find some way of replicating this."

The Diablo mapgen tool, whose development is primarily credited to Matthew Petroff on Github, required reverse-engineering Diablo. It quickly lays out the entire dungeon below Tristram, including item, exit, and quest distribution, in a given "seed" of the game. Those seeds are the possible procedurally generated layouts of a given game of Diablo, with a playthrough's seed set when a game is started. There are 2.2 billion possible legitimate variations, each assigned to one second on the system clock between January 1, 1970, and December 31, 2038, which might have been a sort of mini-Y2K for Diablo if modders hadn't figured this out.

Though groboo claimed to have performed his winning run through luck and skill alone, the mapgen tool proved you could only find certain highly advantageous conditions he ran into on maps outside Diablo's valid date range⁠—you can hack the game to make it work outside the legitimate seeds, though it's unclear how groobo produced his exact run back in 2009. Groobo defended his run by saying "it was a segmented/spliced run. It always has been and was never passed off as anything else, nor was it part of any competition or leaderboards. The SDA page states this outright."

But that doesn't account for the sheer impossibility of the dungeon layout he encountered or other irregularities that have since been uncovered in groobo's run, like the ease and speed with which he defeated Diablo or mismatched details that indicate he spliced the footage from different versions of the game. You can read a full analysis debunking groobo's run from TASbot. Groobo's 3:12 has been stricken from Speed Demos Archive, with its footage remaining up on YouTube for posterity. Meanwhile, it's still listed as the "fastest completion of an RPG videogame" on the Guinness Book of World Records, which has not responded to the Diablo mapgen team or Ars Technica.

"It did harm," argued dwangoAC. "Groobo's alleged cheating in 2009 completely stopped interest in speedrunning this category. No one tried, no one could." Indeed, groobo's run might have sucked all the air out of the room right before speedrunning truly exploded in popularity with the advent of streaming, kneecapping a potential Diablo speedrunning scene centered around any% at a crucial juncture. The good news is that the Diablo mapgen tool is inspiring a new category of "seeded" runs where the player purposefully selects an auspicious layout from the 2.2 billion legitimate seeds. Here's a recent 5:13 set by runner diablo_sb.