Assuming the set has an order

The materials a rank gets built from

1. Position on a page, which is usually arbitrary. See reading order is not a ranking.
2. A speed difference on one or two attempts, which belongs to the path rather than to the address.
3. Which one the reader happens to have used most, which is habit picking for you.
4. A label added by some page, examined in a list that ranks.

Why the belief forms so easily

Unordered sets are uncomfortable. People order things automatically, and a reader who has to choose between three identical looking options will construct a reason and then remember the reason as if it had been given to them.

Nothing in the environment corrects it. There is no feedback that would ever tell a reader their ranking is invented.

The first consequence: false alarms

Once one route is the good one, anything it does becomes meaningful. It is slow, so something is wrong. It is quiet, so the market is in trouble.

The same events on one of the others would have been shrugged at. The rank has made a neutral observation into a signal, which is how a quiet evening turns into a search.

The second consequence: the set shrinks

The two lower ranked entries become spares, spares become background knowledge, and background knowledge is not something you can paste. The reader now holds one address in any practical sense.

That is the mechanism by which almost every reader who loses the benefit of plurality loses it, and it starts with an invented ordering.

What would be needed to rank them honestly

Repeated measurements from many places over time, of properties belonging to the addresses rather than to paths. No reader has that and no page has it either, which is why there is no correct choice.

Absence of a ranking here is not a gap in what this site knows. It is a fact about what can be known.

How to notice you have done it

Try to say why one of the three is better. If the reason is that it usually works, ask how many attempts that is based on and what the others were doing at the same moments.

The answer is nearly always a handful of attempts, remembered selectively, with no comparison.

The correction

Not to argue yourself out of a preference, which is difficult and unnecessary. Just keep all three where they can be pasted, and use a different one occasionally, which is rotating on purpose.

A preference with all three in reach is harmless. A preference with two of them out of reach is the whole problem.

Why this site refuses to help

It would be easy to say try this one first, and it would be a claim with nothing behind it. The three strings here carry no labels, no numbers and no marks, on every page, and that is the only arrangement consistent with the rest of the site.

Why this is the hardest one to stop

It is not a decision, it is a perception. Nobody chooses to believe one of the three is better, and there is no moment at which the belief is adopted that a reader could interrupt.

That is why the remedy in this study is physical rather than mental. Keeping all three within reach and using a different one occasionally does not require the reader to stop believing anything. It simply removes the consequence of the belief, which is the part that costs them something.

What this study is not. It is not a claim that the three behave identically for any reader, only that no observation available supports ranking them.