On 7 February 2026, at a competition called GLS Big Cubes Gdańsk 2026, a Polish solver named Teodor Zajder did something no human being had ever done. He solved a standard 3×3 Rubik's Cube, in official competition, in 2.76 seconds. The previous world record, 3.05 seconds by China's Xuanyi Geng, had stood since the spring before. Records at this level usually fall by hundredths of a second. Zajder took more than a quarter of a second off, all at once.
And then the scoring rules of his own round set the solve aside. It did not count toward his result that day. This article explains how that can happen, why the sport designed things this way on purpose, and why this publication is named after the rule that did it.
Start with what a competition round actually is. A competitor in a 3×3 round solves the cube five times. Before each attempt, officials mix the cube using a computer-generated sequence of turns called a scramble, so that every competitor in the round faces the same starting positions and none of them can be predicted. A judge watches; the competitor starts and stops the timer themselves. Five attempts, five times on the scorecard.
Those five times are then turned into a single result using the sport's standard measure, the average of five, which has governed 3×3 rounds since 2003. The rule: your fastest solve is removed, your slowest solve is removed, and your result is the average of the three in the middle. This publication is called The Middle Three, after those three counting solves.
Now apply the rule to what happened in Gdańsk. Zajder's five solves were 4.99, 5.36, 5.76, 2.76 and 5.34 seconds. The slowest, the 5.76, was removed, as it is for everyone. The fastest was removed too, and his fastest happened to be the fastest solve in recorded history. His result for the round came from the three in the middle: 4.99, 5.36 and 5.34, which average out at 5.23 seconds.
The three that counted, and the result they produced. The world record sat outside the average, in the discard column, next to the worst solve of the round.
To be clear about the scale of that discarded number: only 26 people in the history of the sport have ever recorded an official average faster than the 5.23 that Zajder's round scored, out of more than 275,000 who hold an average ranking. It was one of the great rounds ever solved. The record books took the 2.76 as the fastest single of all time. The round itself was scored as if the 2.76 had never happened.
Why would any sport throw away its best performances? Because a single solve can mislead, in both directions. Scrambles are random, and random is not the same as equal: occasionally a scramble leaves the cube part-way helpful, so that a solver stumbles into a shortcut they could never plan for, and posts a time that flatters them. In the other direction, disaster can strike a solver at full speed. Cubes are physical objects; a piece can pop out mid-solve, a wrist can slip, a moment's misreading can cost seconds. If results were judged on single solves, luck, good and bad, would decide titles. Removing the best and the worst of five strips the luck off both ends. What is left, the middle three, measures the thing the sport actually wants to measure: how fast you are when nothing unusual happens.
The side effect is what happened to Zajder, and it is not rare. It is close to universal.
world-record 3×3 singles were set inside a completed average of five. By rule, not one of them counted toward its own round's result.
The five exceptions all come from the sport's early years, before the average of five became the standard, when rounds used a format called best of three: three solves, with the fastest one simply taken as the result and nothing discarded. They are a short and historic list: Minh Thai's 22.95 from the first World Championship in 1982, Dan Knights's 16.71 and Jess Bonde's 16.53 from the 2003 revival, and two records by Shotaro Makisumi in 2004. Every world-record single since, without exception, has been set inside an average of five and trimmed out of the round that produced it.
Other sports also refuse to count some of their best performances, and the comparison is worth a moment because the reasoning is so different. In athletics, a sprint record does not stand if it was run with a tailwind stronger than two metres per second (World Athletics rules). The reason is fairness: a strong wind at a sprinter's back physically pushes the runner along, so part of the speed belongs to the weather rather than the athlete, and the time is treated as assisted. Cubing's rule removes performances for almost the opposite reason. Nobody thinks Zajder's 2.76 was assisted. There was no wind, no shortcut anyone has found, no asterisk. The sport discards it anyway, not because the solve is suspect but because the question a result answers is about repeatable speed, and one extraordinary solve, however genuine, is not evidence of what a solver can do again on demand.
That is the whole idea this site is built on, so it is worth saying directly. We are named for the three solves that count. The archive underneath keeps all five of every round back to 2003, the discarded ones included. A results page will tell you that Teodor Zajder averaged 5.23 in Gdańsk that morning. Only the discarded column can tell you that, for 2.76 seconds, he was the fastest human being ever to pick up the puzzle.
Record-round, format and ranking data from the WCA results export; recomputed nightly.