>>2817197 (LB)

yes, this is 5GW. The (((Old Guard))) believed they had social media under control, but there were changes in demographics (image based memes among the younger set, networking among older voters who had participated in various 'Truther' movements, some boards like the chans that were not controlled opposition or censored).

They are doubling down on control and about to lose.

Polls are a special case – the practice appears to have been to have models based on demographics and 'past elections' – then, to 'update' them with new data at some weighting. Since the past history represents 'consensus history', and the weighting can be 'tuned' to make certain demographics count moar than others. In a real measurement, this bad weighting would be a source of error – but whoever sees the real world, at scale? It is, in effect, a way of manufacturing consensus.

This is similar to exponential smoothing, ARMA/ARIMA processes, and other time series – where the 'past memory' [of fraudulent results] is quite bogus and the weighting is controlled too. We learned a lot about how this worked during the Bush/Gore election.

At the time (2000), the MSM managed the results at a single media centre, a shared pool called NEP. Because of a leak of the exit poll data to a magazine called Scoop in NZ, we could see the aggregation process. In particular, the telephone polls had different lengths than the in person polls, so it was possible to look at the counts for the last few questions and deduce the ratio of telephone polling and in person polling, by region, and see several snapshots of how the weights evolved in different regions – North and West Coast being Blue, and South and Midwest (Redder).

The results, needless to say, were alarming.

>>2818235

I should add, there was a high-profile campaign using the renowned name of Paulos – the Nate Silver of his day – to try to convince people that the exit polls proved there was election fraud.

There was fraud, but not the sort that con possibly be deduced from exit polls. Because of the numbers involved, trying to prove fraud in an election using a sample the size of exit polls is a lot like trying to tell if an atomic clock is miscalibrated using a dime store Timex.

The 'mathematics' used by the Liberals was utterly bogus – the sort an MBA might think is vaguely true from a stats course he left through, if all the boffins are nodding the same tune. It amounted to believing that the only linear model (GLM) that is acceptable is a random effects model, and that fixed effects models are never appropriate, and there are no existing tests to choose which is appropriate when. Pure tosh and bilge, served up as prolefeed.

Cluestick: you don't draw red and blue states from an urn. States come with very obvious 'fixed effects'.

In any event, any lowly working scientist would try to do at least a chi-squared test with 51 cells, to test if the 'fit to the data is good'. I found anomalies yes – in DC, Tennessee, Texas, and the New England States. None of which tipped the election, which in any event was decided by the Supreme Court and not the polls.