An analogue answers "what followed a similar stretch", not "what will follow this one". The difference is not in wording but in the assumption you would have to add to get from the first to the second.
What that assumption is
For a share of outcomes to become a probability, you must accept that the market repeats, and that stretches alike in shape produce alike continuations.
That is a strong claim, and it is not established. It also runs against the obvious: an identically shaped stretch can occur in completely different circumstances, and nothing in the shape of a curve says anything about those circumstances.
Until the assumption is accepted, "price rose in N of M cases" remains a description of the past and nothing more.
How this differs from a probability
A probability needs either a model that generates outcomes or a long run of uniform, independent trials. A coin and a roulette wheel supply both.
Historical analogues supply neither. The episodes are not independent: they cluster in time, come from the same market periods and carry largely the same information. The sample is small, and it was not drawn at random — it was selected by your own settings.
You define the sample
This is the thing to remember. The similarity threshold, the sample length, the instrument, the timeframe — every one of these choices changes both the number of cases found and the split of outcomes.
Hence a simple corollary: turn the dials until the picture becomes convincing and what you have made convincing is the picture. Checking yourself is easy — fix the settings before you look at the result.
What analogues are for, then
To see how a situation can develop, not to learn how it will.
The most useful part is not the summary share but the episodes themselves: three or four charts show the spread of scenarios. Often the "similar" cases resolved in wildly different ways — which is itself an answer: the situation is uninformative.
A common mistake
Reading the share as a chance: "7 of 10 rose, so the probability is 70%." Even if the repetition assumption held, ten observations do not support precision in percent: the difference between "7 of 10" and "5 of 10" is two cases.