Well, probability is always in play.

But if done well (which is no cheap or easy task), they can provide good estimates of momentary public opinion with a small probable error. The problem with most polls is that they are multiple choice, you don't get to say what's on your mind, you have to pick an approved response, the same sort of "management" that one will see tonight in the debate. Most polls find only what they look for, and tend to throw the results out if they don't get it, like tonight.

So, if one were to take a series of such "good" estimates, you would have a pretty good track of public opinion, and might be able to make some intelligent GUESSES about what motivated any changes. But you still have no prediction of future events without the hidden assumption that tomorrow will be like today, and anybody paying attention knows it might not, and the probability is far from miniscule, and all it can take is one to harm or help irretreivably, election wise.

It is worth remembering that one of functions of the MSM, the Mighty Wurlitzer, is to stall change, any fundamental change in who runs things here, they are 100% pro status quo, since they are privileged members of the elites themselves.

Nassim Nicholas Taleb, the Black Swan guy, would rip them to shreds, he understands the math, he understands how much they are assuming their own conclusions and building them into the questions and answers, and how false the premises of unitary population (i.e. it is sensible to talk about the pollees as though they were all part of the same population WRT the studied attributes), single peak distributions (they can have as many as you like), and follow something like the normal curve, all of which are ridiculous assumptions in the context of large human cultures and human mental states, which are anything but uniform as a rule.

But it gets too expensive without the simplifications, and anyway, they are not in the business of keeping us informed, exept incidentally. So with a few exceptions I consider them all complete crap, bascially propaganda tools.

They tell you nothing about future reactions to future events, and they assume the status quo as a beginning premise, that tomorrow will be like today, which is ridiculous in modern times, we are going through the period of most rapid change in human history, or for a long, long time in the past, we are talking rapid geological change now. Changes in sea level, changes in atmosphere, resource exhaustion, ...

But mostly, they are bought and paid for, with the usual consequences plain to see.

Edit: the few polls I give any credence to are one and all done by academics.