Your scepticism is well-placed, even if there might be some information value in technical signals! Believing they represent an El-Dorado in the absence of understanding, let alone hard work and inevitable losses along the way, will always be the greater sin, irrespective of any potential value (or not) in the technicals!!!
So the "logic" behind technicals basically argues that in the short-term, price movements don't always reflect changes in market participants' expectations. Rather participants have already positioned in expectation before the news; so the reaction to the news has as much to do with this positioning, as with the surprise the news might represent. So, in English, the consensus for X might be 1.5% versus 1.4% last time and it realises at 1.6% but the market might sell off on this news, because everyone was already betting on the rise. There are no new buyers (all who wanted to buy have already bought); and plenty of sellers ready to take profits. So the market sells off (more sellers than buyers) despite the "good" news.
If you accept that there might be two levered buyers and two short-sellers for every cash-funded holder of X, then the thesis above need not sound so crazy.
Suffice to say that in a past life, I was well-paid as a sellside market strategist for 20 years to watch this happen in real-time; try to make sense of it; and explain it. Market behaviour is not, and was not, consistent with the randomness that the economics textbooks calls for. There was, and is, "something else" out there. This does not, of course, mean that technical analysis or "positioning" can explain, and fill this void...
So Technicals basically argue that the effect of positioning on prices in response to news leads to a tendency towards particular patterns of price action. Given these patterns, it should then (basic Bayesian inference) be possible to infer positioning. And thus a likelihood of a bias in price response to new news. It needn't be a big bias, let alone a certainty of direction of response. If you had a 45:55 every week, you could retire rich without having to get a job...
This hypothesis is almost impossible to prove, even to reject. Which is why Technicals remain so popular. The alchemist's financial "philosopher's stone" remains extant.
The problem with rejecting this all out of hand is equally simple. One can create very simple Hidden Markov Models of markets that differentiate between a "stable" state and a "chaotic one", where returns in the latter are lower for higher volatility. With p=values good enough to write a statistical paper, better than much of the medical evidence that informs policy decisions about Covid-19.
But the HMMs are "technical" in nature, no...? How does dressing this up in right-sounding modern statistical modelling 101 make it any more legitimate than talking about "head and shoulders" patterns? That is the real problem ;-)
To believe that the HMMs are spurious, you have to believe that there is NO effect on the distribution of the mean and volatility of returns in low-vs-high volatility regimes. Good luck arguing that one! Else you have to concede there are regimes where "greed" and "fear" (for want of better explanatory variables) do influence the distribution of returns. So sentiment matters. And if sentiment matters, there is indeed some technical dimension to asset returns. QED.
Where the limits of this process start and finish is where investing paupers vs poor retirees vs content retirees (who cruise a lot) vs millionaires/billionaires limits end up getting drawn, IMO... ;-)