After more than four decades of practice, a Super Mario Bros. perfect run is approaching a remarkable possibility: a human player matching a time calculated through frame-by-frame tool assistance. There are now just two frames separating the fastest human completion from that benchmark. The remaining gap is so small that an ordinary stopwatch cannot meaningfully capture it.
On October 31, 2026, speedrunner Niftski completed Nintendo’s 1985 NES classic in 4:54.299, setting a new any% world record during a Halloween livestream while dressed as Mario. Yet the costume is only a footnote to the bigger story. The real question is whether one of gaming’s most thoroughly optimized challenges can finally be completed without losing another frame.
Why the Super Mario Bros. perfect run has a 4:54.265 target
To understand the achievement, it helps to separate a world record from a theoretical limit. Niftski’s previous best was 4:54.332, recorded in August 2026. His Halloween performance reduced that time by 33 milliseconds, equivalent to two frames. That may sound negligible, but every improvement is increasingly difficult when a game has been studied this closely.
The target is 4:54.265, established by a tool-assisted speedrun, commonly shortened to TAS. Rather than relying on live human reactions, a TAS uses tools to arrange inputs frame by frame. It can demonstrate what a precisely executed sequence might achieve. Crucially, this particular benchmark follows the same input restrictions as human competitors.
Another TAS finishes in 4:54.032, but that result uses simultaneous left and right inputs. Such a combination cannot be entered on a standard NES controller and is prohibited in human competition. Consequently, 4:54.265 is the relevant theoretical target, not the superficially faster number. Niftski acknowledged the remaining two-frame gap in a social media post about his Halloween record.
The strange 21-frame rule that makes tiny mistakes enormous
Super Mario Bros. runs at approximately 60 frames per second. Each frame therefore lasts around 16.7 milliseconds. However, most level transitions do not reward every tiny improvement directly. Instead, the game checks when to move forward in cycles of 21 frames, often called frame rules.
Imagine reaching a level’s endpoint several frames earlier than usual. If that improvement does not cross the next transition threshold, the following level still begins on the same cycle. In other words, a faster-looking section may save no time on the final clock. This unusual timing system explains why elite runners concentrate on particular thresholds rather than simply rushing through every screen.
World 8-4 changes the equation. The final stage does not provide the same frame-rule cushion at the finish. Movement, positioning and inputs can directly influence the completed run’s time. As a result, the closing sequence becomes a decisive test of precision. Matching a TAS requires an entire route to align, not merely a spectacular final jump.
How a 2021 breakthrough became a rivalry measured in frames
Niftski has been chasing this boundary for years. In 2021, he became the first runner to complete Super Mario Bros. in under four minutes and 55 seconds. That achievement was compared within the community to athletics’ historic four-minute mile: a psychological barrier that once seemed extraordinarily difficult to overcome.
Progress did not stop there. In June 2026, Brazilian speedrunner LeKukie recorded 4:54.365, taking the world record and bringing the category within six frames of its theoretical limit. Niftski reclaimed first place in August with 4:54.332 before improving his own mark on Halloween.
This rivalry matters because the leading runners are no longer separated by major route discoveries or entire seconds. Instead, they compete through repeatability, technical execution and the ability to assemble an almost flawless sequence under live conditions. Both continue streaming attempts, keeping the possibility of a TAS tie alive.
There is also a broader reason the achievement resonates. Unlike a modern competitive game, the original Super Mario Bros. does not receive balance patches that rewrite its movement or level design. Its underlying challenge remains familiar. What changes is the community’s understanding of how efficiently it can be played.
Could the next world record be the last possible improvement?
Any% allows players to reach the ending through the fastest permitted route, including warp pipes that skip entire worlds. That makes it a specialized category rather than a complete tour of every stage. Nevertheless, mastering the route still demands remarkable consistency. A promising attempt can collapse because of one badly timed movement or missed setup.
For a player trying to match 4:54.265, the challenge is different from setting an ordinary personal best. There is no meaningful reserve of easy time left to recover. Every important section must reach its required timing, while the final moments leave almost no room for correction. Even knowing the necessary inputs does not guarantee that a human can execute them all in one attempt.
If someone eventually matches the rules-compliant TAS, it would represent a remarkable milestone: human execution reaching the established theoretical benchmark. It would not mean every conceivable interpretation of the game had been exhausted. Different categories, techniques or future discoveries could still produce new challenges.
For now, the Super Mario Bros. perfect run remains unfinished business. Niftski has pushed a game released in 1985 to within 33 milliseconds of its accepted limit, while LeKukie continues the pursuit. After 41 years, could the next historic moment be a world record that nobody can improve under the current rules?





