90% of the polls they post are fake and manufactured to sway public opinion.
Better to say 'most' rather than picking some random number like 90%.
But you are essentially correct. News polls are not polls at all, just numbers printed to sway public opinion. Other polls are just random numbers of type randU, since you don't know that questions that were asked, who was asked, or even the time the so-called questions were asked. Further, the math to show the statistical summary is never shown, nor the dataset.
So there is no alternative but to consider them random numbers of type randU. Claiming any such numbers as a prediction is just a math error. Claiming any such numbers as 'the truth' is just an argument from randU fallacy.
People spew numbers all the time: JPP members, the government, some university, some so-called 'expert'. Most of it is just trash.
Before I accept numbers as data, I must first know:
* How that data was collected. If a poll, that includes publishing the questions and the number and type of people asked the question. If not a poll, the instrumentation used (if any), how it was calibrated, and when it was last calibrated; and the measurement conditions in detail.
* When the data was collected and by whom and for what purpose.
* The dataset iitself must be publicly available. No paywall.
* If a statistical summary is quoted, the math work must be shown, including:
* How the variance was declared and justified.
* Calculation of the margin of error value.
* The selection set used (the set selected by randN).
* Calculation of the mean, including normalization by paired randR.
* BOTH numbers published...not just one. Publishing only number without the other renders the quotation a randR in and of itself.
These are simple requirements for anyone that has real data and performing a valid statistical summary.
Note that a statistical summary is incapable of prediction normally inherent in mathematics, due to the use of random numbers.
Indeed, one statistical summary on the SAME dataset may radically differ from a statistical summary done again, all due to the randN nature of building the selection set.
Short definitions of random types:
randR: like rolling a single die. A repeatable random number.
randN: like picking a card from a deck. Once picked, it cannot be picked again until some reset event (aka a shuffle).
randU: a number thought up in someone's head, or an algorithm simulating it. Such numbers are often use to embellish some claim as a 'proof'. This is the psuedo-random number. It is not a true random number.
Random number mathematics is it's own branch of mathematics, and is even in a different Domain of mathematics (one axiom changes to create this Domain). It IS possible, however, to import numbers from one Domain into another. This importation removes the capability of prediction normally inherent in mathematics.
Thus, probability mathematics, and statistical mathematics are incapable of prediction. They are useful summary tools, and only if they a re properly used, but they cannot predict anything, and they are not a proof of any kind.
