How is this speedrun possible? Super Mario Bros. World Record Explained
UPDATE: This run was beaten on May 25th, 2018 by somewes with a 4:56.245: \r
\r
Watch the unedited run here: \r
- My Twitch: \r
- My Twitter: \r
- My Discord \r
\r
Kosmic (4:56.462 WR)\r
- Twitch: \r
- YouTube: \r
- Twitter: \r
- Warpless WR: \r
- Glitchless WR: \r
- Minus World Ending WR: \r
\r
Notes:\r
At 0:47, I refer to the console Kosmic is using as the Nintendo Famicom Disk System. He wasnt ually playing on the Disk System, just the Famicom. The Famicom is essentially the Japanese equivalent of the NES, and the Disk System is an add-on to it that reads special floppy disks.\r
\r
At 4:06, I mention that Kosmic holds the Glitchless WR with a 5:06.709. However, there is a video by Scott Kesler of a 5:05. This is because of timing differences. Glitchless is how Twin Galaxies (a deprecated high score website) view Super Mario Bros. speedruns, and its timing rules are different. They time from the start press on the title screen, which, because of framerules, adds a variation of up to 21 frames to the time of the run based on which frame you pressed start on. This timing difference adds about 2.5 seconds, so Scott Keslers 5:05 was ually a high 5:07 by TG rules, and Kosmics 5:06.709 is a low 5:04 with regular timing.\r
\r
At 13:15, I say that when the white pixel is missing, the judges are in a bad mood. For the sake of simplicity, I omitted that its ually a 50/50 of good/bad judges if you dont get the pixel. However, this is frame dependent and not entirely random, and Kosmic rarely gets good judges with the bad pixel, so I took that shortcut.\r
\r
At 26:35, I claim that the theoretical human limit is 4:55.496. The ual time is 4:55.49668, and it should have been rounded up to 4:55.497.\r
\r
There is another framerule that was not included because while it is considered too hard for humans to ever get in a run. You can save it by doing a frame perfect fast acceleration at the start of 8-1 (after several seconds of black screen, with inhuman reion time) to save 2 frames, and then doing a TAS flagpole glitch to save 19 frames instead of 15. TAS flagpole glitch never slows down to set it up and is based around doing a left-right wiggle at some point during the level and praying that it set your position up perfectly. The best consistency at this trick is just about 1%, compared to 40% for regular FPG. On March 4th (2 days after the release of this video), StuckInAPlate managed to save the 8-1 framerule from a savestate immediately before gaining control, which eliminates the need to time your frame perfect inputs after a very long wait. \r
\r
Super Mario Bros. Leaderboards: \r
Frame Count spreadsheet (also by me): \r
\r
Other speedrunners featured in this video:\r
\r
darbian (4:56.528)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
somewes (4:56.894)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
StuckInAPlate (4:56.978)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
EddieCatGaming (4:57.543, tied 8-4 IL WR)\r
- Twitch: \r
- YouTube: \r
\r
andrewg (4:57.693)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
HappyLee (Author of the human theory TAS)\r
- TAS (4:54.03): \r
- YouTube: \r
\r
\r
Special Thanks, for helping in some manner:\r
Kosmic\r
somewes\r
HappyLee\r
SummoningSalt\r
EZScape\r
darbian\r
andrewg\r
StuckInAPlate\r
EddieCatGaming\r
Znernicus\r
theballaam96\r
1ted59
\r
Watch the unedited run here: \r
- My Twitch: \r
- My Twitter: \r
- My Discord \r
\r
Kosmic (4:56.462 WR)\r
- Twitch: \r
- YouTube: \r
- Twitter: \r
- Warpless WR: \r
- Glitchless WR: \r
- Minus World Ending WR: \r
\r
Notes:\r
At 0:47, I refer to the console Kosmic is using as the Nintendo Famicom Disk System. He wasnt ually playing on the Disk System, just the Famicom. The Famicom is essentially the Japanese equivalent of the NES, and the Disk System is an add-on to it that reads special floppy disks.\r
\r
At 4:06, I mention that Kosmic holds the Glitchless WR with a 5:06.709. However, there is a video by Scott Kesler of a 5:05. This is because of timing differences. Glitchless is how Twin Galaxies (a deprecated high score website) view Super Mario Bros. speedruns, and its timing rules are different. They time from the start press on the title screen, which, because of framerules, adds a variation of up to 21 frames to the time of the run based on which frame you pressed start on. This timing difference adds about 2.5 seconds, so Scott Keslers 5:05 was ually a high 5:07 by TG rules, and Kosmics 5:06.709 is a low 5:04 with regular timing.\r
\r
At 13:15, I say that when the white pixel is missing, the judges are in a bad mood. For the sake of simplicity, I omitted that its ually a 50/50 of good/bad judges if you dont get the pixel. However, this is frame dependent and not entirely random, and Kosmic rarely gets good judges with the bad pixel, so I took that shortcut.\r
\r
At 26:35, I claim that the theoretical human limit is 4:55.496. The ual time is 4:55.49668, and it should have been rounded up to 4:55.497.\r
\r
There is another framerule that was not included because while it is considered too hard for humans to ever get in a run. You can save it by doing a frame perfect fast acceleration at the start of 8-1 (after several seconds of black screen, with inhuman reion time) to save 2 frames, and then doing a TAS flagpole glitch to save 19 frames instead of 15. TAS flagpole glitch never slows down to set it up and is based around doing a left-right wiggle at some point during the level and praying that it set your position up perfectly. The best consistency at this trick is just about 1%, compared to 40% for regular FPG. On March 4th (2 days after the release of this video), StuckInAPlate managed to save the 8-1 framerule from a savestate immediately before gaining control, which eliminates the need to time your frame perfect inputs after a very long wait. \r
\r
Super Mario Bros. Leaderboards: \r
Frame Count spreadsheet (also by me): \r
\r
Other speedrunners featured in this video:\r
\r
darbian (4:56.528)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
somewes (4:56.894)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
StuckInAPlate (4:56.978)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
EddieCatGaming (4:57.543, tied 8-4 IL WR)\r
- Twitch: \r
- YouTube: \r
\r
andrewg (4:57.693)\r
- PB: \r
- Twitch: \r
- YouTube: \r
- Twitter: \r
\r
HappyLee (Author of the human theory TAS)\r
- TAS (4:54.03): \r
- YouTube: \r
\r
\r
Special Thanks, for helping in some manner:\r
Kosmic\r
somewes\r
HappyLee\r
SummoningSalt\r
EZScape\r
darbian\r
andrewg\r
StuckInAPlate\r
EddieCatGaming\r
Znernicus\r
theballaam96\r
1ted59