Shrewdies i wouldv'e thought that was the obvious conclusion anything around 40+..

I don't think there's anyone who really thinks it will be less than 30 minimum,i think it couldv'e been just so many open goals missed by labour everyday there's been an opportunity to slam may,the chance has gone..

I am hearing that the Survation poll tonight has the lead down to 1%.

Survation....

Con 41.5 Labour 40.4 LibDem 6 UKIP 3.

There is a 50% chance that the polls showing leads of 1% are nonsense. There is also a 50% chance that the polls showing leads of 11% are nonsense.

The polling companies that have 11 and 12% leads are Comres and ICM. Their methodology greatly favours the Tories. Their true unweighted polling is nearer to 3%.
Survation have used the same model that they used in 2015. On election day they had a poll (which they never released) which was spot on. This year they had a Con lead of 18.1% on 15th May, 9% on 22nd May, 5.8% on 30th May and 1.1% on 5th June.

My reading of that Survation poll is that it is completely and utterly meaningless.

From what I can make out the unweighted sample has 563 remain voters and 363 leave voters.

They then try and weight it.

Why would you start a poll off that is so unrepresentative of the population and then try and weight it?

IT and his usual tory bias , same old song

The weighted sample on which the poll is based is 494 leave, 458 remain. To suggest that polling companies only poll Labour types is hilarious and smacks of desperation.  The polling companies poll by surveying an accurate as possible representation of the electorate, to do otherwise would mean going bust.

You are effectively saying if you received a good education and went to university you vote Labour. If you didn't you vote Conservative?

Not all at. It is an example of underlying biases.

https://yougov.co.uk/news/2015/06/08/general-election-2015-how-britain-really-voted/

The yougov poll has GCSE or lower 38 con/ 30 lab. Whereas for uni its was 35 con/ 34 lab.

The point is if there is an underlying bias in the sample the poll cannot be accurately extrapolated.

....

For an example lets say you want to get an answer how many people in Europe wear coats.

Let say you sample 2000 people. 400 were eating ice-cream and 1600 were drinking hot-choc. of those eating ice-cream 350 were wearing a coat. Of those drinking hot-choc 1500 were wearing a coat.

You want to know how many people wear coats in Europe.

The raw number is 1850 out of 2000 wear a coat. 92.5 of people wear coats.

You know on average 50% of people have ice-cream and 50% have hot-choc in Europe.

So you adjust the sample to 400 ice-cream and 400 hot-choc. This gives you 725 people out of 800 wear a coat. So 91% of people in Europe wear coats.

I then tell you the sample was taken in Sweden in winter. Do you really think your 50/50 sample of ice-cream eaters and hot-choc drinkers is representative of the whole of Europe as to if people wear coats or not?

When there is a clear bias in the raw data it cannot be trusted.

There is no bias whatsoever in the weighted total. It is weighted to give a correct representative sample of the electorate.

The only person suggesting it is you.

Meadow I am sure you are right..

Do you know how people were actually sampled originally?

I am new to this kind of 'science' and trying to get my head round it.

For male female there are 609/494 raw adjusted to 538/565. Both have 1103 but clearly needs to be at least 1174 in total or they are weighted to be more females then actually as in the raw data?
For leave remain there are 377/575 raw adjusted to 494/458. Both have 952. So where are the others for the leave/remain? 151 less.

I am struggling to get my head round it and how the adjustments are made to make sure the leave remain split is accurate in the weighted total.

Good to see you here again CJ, sorry you are not feeling so grand today. The Scottish Labour party are making up ground quickly. For that reason yesterday I placed small amounts on Labour in all the Glasgow seats and some in the central belt.  Strictly I don't expect them to win any of them but at prices up to 20s I am ok with that.  East Lothian is Labours main target but the betting value has all but gone (it's 3/1, was much higher). I am fairly sure Labour will retain Edinburgh South.
The Scottish Labour surge threatens to derail Conservative chances in some of their target seats.  Effectively, any seat they are after where there is a significant Labour vote now looks a far less likely gain. East Renfrewshire, Aberdeen South and Stirling come into this category.
Jo Swinson would make a great Scottish MP in parliament but she has a tough task. The Conservatives voted tactically in numbers for her in 2015 and she still fell short. You have to feel many of those folks will vote Conservative this time.
As you will see from my original post on this thread I am in heavy on the LibDems having a bad night. Less than 10% and less than 10 seats would be the icing on the cake.

1/4that is one huge price for an overall majority thought it wouldv'e been smashed into 1/8 by now already got 1/8 for most seats ..these crazy polls are making some cracking bets..

Meadow X1,

Genuine question.

Do you know how the sampling adjustment works in the poll discussed above when the different attribute groups do not all add up to the 1103 sample size?

At the last election it was to close to call and the majority on here were backing the torys at huge prices to win with a majority just hours before the first result isn't it the exact same scenario again...

I also remember that poster earlier cj saying several days in advance of last election that the labour vote was getting decimated by ukip from some canvassing that he had recieved,he's put up the 1.1 and 1.4 today,so i presume he's hearing the same again and he will be putting up bets against labour again.He was right as well as the polls didn't reflect anything that he knew from the doorstep..

Anyone that bets based on their political allegiance is a walking money machine for the bookies.