A historical analogue is a past stretch of chart whose shape resembles the current one. The point of the search is simple: find places where price moved similarly and look at what followed each of them.
What exactly gets compared
Not prices, but shape. That matters: Bitcoin at twenty thousand and Bitcoin at ninety will never coincide in levels, yet the pattern of movement can be the same.
So before comparing, each stretch is converted to a comparable form — usually to percentage change from its own first candle. After that, "plus five percent over ten candles" looks the same at any price level and on any instrument.
What is compared, then, is trajectories: where the rise was, where the pullback, how sharp, in what order.
How it differs from patterns
A pattern is a named figure: head and shoulders, a flag, a double top. There is a finite number of them, and each is described in words.
An analogue is neither named nor described. It is simply a stretch that came out similar by a number, with no name and no interpretation. Because of that there are far more analogues than patterns, and among them are shapes nobody ever named.
The flip side: a pattern comes with a tradition of reading, an analogue does not. What to do with a matched stretch is up to you.
What an analogue shows
Exactly two things: that a similarly shaped situation existed in history, and that price afterwards ended up where it ended up.
Everything else is interpretation. An analogue does not explain why the move looked that way, takes no account of circumstances, and does not know whether they will recur.
What it does not show
Not probabilities. The share of cases that rose describes the sample you found, not a chance of what comes next; turning one into the other requires an assumption that the market repeats, and that is not established.
Not causes. A similar shape can arise for entirely different reasons, and the chart says nothing about them.
Where people get confused
Treating the number of matches as a mark of reliability. Usually the reverse: many matches mean the similarity criterion was too loose.
Looking only at the summary. It is more useful to open three or four matched stretches: often the "similar" cases resolved in different directions, which is itself an answer.
Forgetting they defined the sample. The similarity threshold, the sample length, the instrument — each choice changes both the number of analogues and the split of outcomes.