bagheadinc wrote:
collegestudent22 wrote:If what you are measuring is analyzed in a way as to come up with a result that literally is insensible, you are doing something wrong.

His data pointed to people favoring Obama. They measured public opinion in a variety of polls and the results lead to the conclusion of Obama winning, makes sense to me.

Only 90% of the time, though. 10% of the time, Romney wins, right?

collegestudent22 wrote:This is a relevant point, and directly contradicts the claim it was in response to.

But it doesn't. Just because something changes, doesn't mean the past data is now invalid. It just means you have more data to evaluate.

It's not a matter of data changing. It is a matter of what is going on - abstracting from the candidates to try to claim your classifications work for prediction is no more sensible than abstracting away facts that are known about an individual with cancer to merely assign him the same chance of survival as everyone else with cancer.

Yes, an entire field of mathematics is nonsense because you fail to understand it.

This is part of your problem. It isn't that I "don't understand" the mathematics, nor the attempt to apply it. It is that is

*not a logically sensible statement*. It is on par with "2 + 2 = fish" - it really is. There are two kinds of probabilities and uncertainties. The frequency kind - "1 time out of 6 the dice will roll a 4", for instance, or "10% of salmon are 7 inches long". The other kind is that of complete uncertainty, dealing with individual cases. Here, a numerical statement

*can only be metaphor* for a subjective degree of certainty.

The mathematics of game theory tend to be either obfuscating (the analysis would be more sensible and obviously true without the mathematics, as in the prisoner's dilemma type analysis), or just - like mathematical economics - bordering on the absurd, and full of error legitimized as "scientific" by the mathematics. Especially the formulations that simply

*assume* human beings to either a) know about the absolute unknown - what the other individuals want and in what order or b) treat them as computer programs following inherent rules no matter how irrational or self-defeating. Nearly all forms of game theory analysis neglect the fact that human beings can

*learn*, and adjust their behavior accordingly.

Arres wrote:Yes, ever race is different. Yes the pieces that make up the whole are different. What is being said here is this:

"Regardless of the various inconsistencies in HOW we measure (polling), we find that when the RESULT of our measurement is 7, with a margin of error of 5, in 92 measurements out of 100, the candidate wins."

You ABSOLUTELY can make that statement. It is clear, logical, and makes sense.

It ignores various issues that are

*known causal factors*, preventing such a classification of races as a whole from being sensible. Historical trends matter, known information matters, as-yet-unknown information that will come to light matters, the actions of individuals running, the decisions of individuals who vote, and so forth. All of these things need to be factored in (as a matter of fact, Silver even does factor much of this in as part of his analysis), but a significant amount of it is either completely unknown or subjectively determined to be more or less important. Thus, any number, such as "92% certain this candidate will win" is merely metaphor for a subjective degree of certainty with an educated guess and estimation.

The fact that the "numbers hold up" is irrelevant. He's making a really decent guess, based on a lot of relevant information, and he's pretty good at determining just how relevant it all is. Frankly, I'd expect the numerical predictions of electoral votes, and the prediction of the winner to be correct more often then not. But it is not "90% certain" or any such thing. Such a statement, taken at its face value, is meaningless.

Of course, I don't expect any of you to actually listen to what I am saying to you. Clearly, it doesn't fit with your preconceptions of probability, and the inane idea that it can be applied to human decisions as if these are deterministic and statistical in nature. In fact, I think there is an 87.3% chance you just assume I'm wrong. (That being

*metaphor* once again.)