I would like to know if members believe I am on the right lines when I say:
It is best not to follow just one poll, but every poll published when calculating the true ratings of both presidential candidates and only to start tracking one week after the last convention ends using a seven day rolling average and ignore all polls that have an "Undecided" response
If I am correct in this assumption (and please comment on that first) which pollsters should I track and do they publish their polls at the same time (i.e do you have to have a paid subscription to get the poll at the same time that the press corps do?)
Every pollster has a different methodology and can change it at will to fit an agenda.
One literal way to ignore the undecided is to figure that the undecided will not vote, so if one has a 50-40 split then divide 50 by .90 and 40 by .90 and you get a 55.6-44.4% split. Or you can divide the undecided by half and add 5% to both.
You can ignore pollsters connected with special interest (trade associations, lobbies, unions, ethnic associations) and those that someone says uses questionable methods. Push polls in which the one statement is an ideological promise and the one involving the person asked about smear ("Do you approve of interracial sex?" followed by "Will you vote for John McCain who has a BLACK BABY?" -- never mind that the child is adopted from Calcutta) are to be avoided. Sometimes people catch onto a questionable methodology and if there is some apparent statistical fraud that suggests that the pollster is fabricating results you can and must ignore the poll.