How the ranking works

One question, asked of cubers who own both cubes in a pair: if you could only keep one, which? There is always a third option — can't say, haven't used both enough — and it carries the same visual weight as the two cubes.

We do not ask anyone to rate a cube out of five. An absolute score is contaminated by what the rater paid: someone who spent ৳1,199 has a reason to defend it. A forced choice between two cubes the same person owns cancels most of that, because both cubes carry the same owner's bias.

Only cubers who own at least two cubes of the same puzzle type, and only about pairs drawn from what they own. If you own one 3x3, you are never asked about 3x3s.

Cubes bought from CubeNation are verified from delivered orders. Cubes you add to your collection yourself still count toward the ranking, at reduced weight — but they never earn loyalty points, so nobody can add the whole catalogue and farm rewards.

Judgments are fitted with a Bradley-Terry model — the same family of method used to rank chess players and, more recently, language models from head-to-head preferences. Each cube gets a strength estimate and a confidence interval.

Price, stock, margin, brand and whether we sell a cube are not inputs. They are not reachable from the scoring code at all. The board's order is produced from judgments and nothing else.

When two cubes' confidence intervals overlap, the data cannot tell them apart — so we show them as equals rather than inventing a first-and-second. Most cubes within a tier really are tied, and saying so is more useful than a ranking that pretends otherwise.

A cube also needs enough evidence before it can hold a position at all: a minimum number of comparisons and a minimum number of different cubers. Thirty judgments from one enthusiastic collector is one person's taste, not evidence. Cubes below that bar are listed as “not enough data yet” rather than given a position they have not earned.

Roughly one question in ten repeats a pair you have already judged. How often your answers agree with themselves becomes a weight — quietly, with no penalty and no accusation. Answers submitted faster than a human could read the two names count for nothing.

This measures perceived quality among owners, not measured performance. Nobody is timing solves here. Brand reputation certainly survives in the numbers to some degree, and we would rather say so than imply a precision the method does not have.

We also rank only cubes we sell. In this market that is very close to everything available, so it is not much of a gap — but it is a real one, and it is ours to declare rather than yours to discover.

CubeNation sells every cube on this board. That is a real conflict and it cannot be argued away — only made checkable. So the method is published here, positions carry their evidence counts in the open, and the board declines to rank what it cannot separate.

If a ৳499 cube outranks a ৳1,199 one, that is because the people who own both said so.