They don't answer the question "whatever, just tell us who's gonna win?". But they are good from a qualitative standpoint if not a quantitative one.
For example, if I told you guys there were say, 10 swing states in the US and that 7 or 8 of them showed that they were heavily breaking blue, would could you deduce from that? Not who's gonna win necessarily, but you'd know that things were moving in our favor...and perhaps in a big way. Similarly, if we know that many/most of the swing counties or districts that were close in the previous election were suddenly showing up as heavily in Democrats' favor, you'd know that the Republicans were likely in some serious shit if that trend continued through the election.
Below is a famous image that showed up not long after the 2018 midterms. It showed those House districts that flipped from one party to the other. It takes all of about 2 seconds to recognize that the blue side probably did a helluva lot better than the red side and that the former likely picked up a shitload of seats. Which is exactly what happened. It would likely look very similar if we were to graphically display the findings of polls like the one in the OP.
Try not to read too much into any one poll or focus too much on the actual numbers themselves...sometimes just knowing which direction they are moving is enough to give one some hope about the next election...or dash them against the rocks. Based on what I'm seeing these days, I sure as hell wouldn't want to be in a Republican boat right about now. Especially if Captain Chaos was commanding it.