Westminster Voting Intention: Poll of Polls
|22 Sep 2018||BMG Online||HuffPost||38||38||10||4||5||Tie|
|20 Sep 2018||Opinium Online||37||39||9||2||8||+2|
|19 Sep 2018||YouGov Online||Times||40||36||11||2||5||+4|
|18 Sep 2018||Ipsos MORI Phone||Evening Standard||39||37||13||5||2||+2|
|13 Sep 2018||YouGov Online||Times||40||36||11||3||4||+4|
|13 Sep 2018||Opinium Online||39||38||7||3||6||+1|
This is a rolling poll of polls, which is weighted so that a poll loses 1/12 (0.083) of weighting every day after the end date of the survey (which isn't always the date it is published). Therefore, a poll where the fieldwork finished today has a weighting of 1, yesterday 11/12, 2 days ago 10/12, and so on. When a poll is 12 days old it has a weighting of 0, which means it will drop out of the poll of polls altogether.
In many ways, a rolling average of polls is not a foolproof method of determining the current positions of the parties - averaging good polls with bad won't produce better results - but it does give a better idea of the current trend of the polls than just taking the latest poll, which in itself has differences in methodology which could affect the headline figures. As such, the polls which the rolling average has been drawn from are included above, which is likely to highlight any outliers (which satistically happen every 20 polls on average), and you can draw your own conclusion about them - having said that, the outling polls make a very small difference to the overall poll of polls.