From a purely scientific perspective...
...this statement is the oddest thing I have ever read on #TacticalVoting.
If you cannot see why, you shouldn’t be anywhere near data-based decision-making.
But allow me to explain anyway.[thread]
...this statement is the oddest thing I have ever read on #TacticalVoting.
If you cannot see why, you shouldn’t be anywhere near data-based decision-making.
But allow me to explain anyway.[thread]
Firstly, tactical voting is about voting for whomever has the best chance.
To do that, *accuracy* of data on voter intentions is the key thing you need.
To say “We don’t dispute the accuracy”
...means you acknowledge it is accurate.
And have nothing better.
To do that, *accuracy* of data on voter intentions is the key thing you need.
To say “We don’t dispute the accuracy”
...means you acknowledge it is accurate.
And have nothing better.
So if you don’t dispute the accuracy, what’s the “but”?
The quote continues that “using a number of different pieces of data... is a better way”
This shows deep ignorance of how @BestForBritain derived their results... as they use tens of thousands of rows of data.
The quote continues that “using a number of different pieces of data... is a better way”
This shows deep ignorance of how @BestForBritain derived their results... as they use tens of thousands of rows of data.
So yes they use “previous election results and local factors”. That’s the absolute modelling minimum.
They combine this with multilevel regression & post-stratification (MRP) polling over repeated time samples, incumbency factor, demographics and a ton of other considerations.
They combine this with multilevel regression & post-stratification (MRP) polling over repeated time samples, incumbency factor, demographics and a ton of other considerations.
The question then becomes, for a scientist: Does the model have proven predictive powers?
Yes. As B4B have been doing this for a while, they have track record on this, including predicting the hung parl and surprise wins of Labour MPs @emmadentcoad & @RosieDuffield1 in 2017.
Yes. As B4B have been doing this for a while, they have track record on this, including predicting the hung parl and surprise wins of Labour MPs @emmadentcoad & @RosieDuffield1 in 2017.
So in summary, to say that you don’t dispute the accuracy of a highly-accurate model...
...but think it should really include stuff that it already includes - is fairly daft.
What is the purpose of making a critical statement if you have no evidenced criticisms?
...but think it should really include stuff that it already includes - is fairly daft.
What is the purpose of making a critical statement if you have no evidenced criticisms?
Is the B4B model 100% perfect? Nope. Nothing can be.
But with a 46,000 voters sample size, it’s looking the most data-rich one.
Clearly, there will be outliers to iron out (predicting >500 seats is no breeze)... and sampling will need to be updated as this GE develops.
But with a 46,000 voters sample size, it’s looking the most data-rich one.
Clearly, there will be outliers to iron out (predicting >500 seats is no breeze)... and sampling will need to be updated as this GE develops.
However, we really need to strip away the politicising of the tactical voting resources which will, combined, provide a powerful tool for those who *do* want to vote tactically.
Ultimately, doing it helps *both* Lab & Lib Dems gain seats.
It’s a smart vote swap, it’s a win-win.
Ultimately, doing it helps *both* Lab & Lib Dems gain seats.
It’s a smart vote swap, it’s a win-win.
If you want a video overview on how to approach and understand the tactical voting to keep out Tories/BXP... here’s one I made earlier:
Watch on #Periscope: Tactical voting: Chillax. Here’s what you need to know.
pscp.tv/w/cI7rYjFtTUVQ…
Watch on #Periscope: Tactical voting: Chillax. Here’s what you need to know.
pscp.tv/w/cI7rYjFtTUVQ…
