Thread by @jonworth

So the meaning of life, the universe and... #tacticalvoting sites is:
6

possibly:
2

but probably:
1

... but perhaps a dozen or so "saved"

Read on to find out what I mean!

This evening I set out to assess how well my tactical recommendations had done in the 187 constituencies I had made recommendations for #GE2019

I got 9 wrong, but none that mattered - as I explain here

https://twitter.com/jonworth/status/1206653061852225538

But that's not the really interesting point

The interesting one is where did the sites differ in their recommendations, and did *that* matter

I added constituencies that I'd not assessed, coming to a total of 227 constituencies for this thread

We know from @chrishanretty's work - on the Lib Dems so far - that a recommendation assisted that party by 2-4%

https://twitter.com/chrishanretty/status/1206233331622457344

So are there cases, first of all, where the 5 tactical voting tools differed in their recommendations, *and* tactical voting could have worked?

YES

There are 6. Just about.

Broxtowe
@BestForBritain and @peoplesvote_uk recommended soubry
@remainutd @votetools @TacticalVote and I recommended Labour

Labour came second, Soubry 3rd

Results are here:
en.wikipedia.org/wiki/Broxtowe_…

Soubry + Green + Labour > Tory
But Tory 9.6% ahead of Labour

Chelsea & Fulham
All sites recommended Lib Dems except @votetools who recommended Labour

Lib Dems came second, Labour 3rd

Results here:
en.wikipedia.org/wiki/Chelsea_a…

But the Tory got 49.9% - you are never going to get to complete alignment unless candidates pull out

Cities of London and Westminster
@votetools and @remainutd went for Labour, everyone else for Lib Dems

Lib Dems second, Labour 3rd

Results here:
en.wikipedia.org/wiki/Cities_of…

Tory ahead of Lib Dem by 8.9%

Kensington
@remainutd & @peoplesvote_uk went Lib Dem
@BestForBritain @votetools @tacticalvote and I (eventually) Labour

Labour 2nd, Lib Dem 3rd

Results here:
en.wikipedia.org/wiki/Kensingto…

Tory ahead of Labour by 0.3%

Wimbledon
All sites recommended Lib Dem except @votetools that recommended Labour

Lib Dem 2nd, Labour 3rd

Results here:
en.wikipedia.org/wiki/Wimbledon…

Tory ahead of Lib Dem by 1.2%

Woking
All sites recommended Lib Dem except @votetools that recommended Labour

Lib Dem 2nd, Labour 3rd

Results here:
en.wikipedia.org/wiki/Woking_(U…

UKIP + Conservative = exactly 50%

That's right

6 constituencies of the 632 in England, Wales and Scotland where the tactical voting sites disagreed, and the Tories received less than 50% of the vote

But look at the details

In Chelsea and Fulham and Woking it would need such ridiculous coordination to work - we can rule those out

The Tory winning margin in Broxtowe and Cities of London and Westminster was more than double that of the effect discovered by Hanretty

That leaves 2

Wimbledon had only one site recommending something other than Lib Dem - @votetools - and that was one of the less well resourced ones. Would that have made the 1.2% of difference... possibly, but I doubt it

So actually it leaves just 1

Kensington

The most complex, difficult, horrible constituency to judge. One where I agonised over my own recommendation. And where, I think, had ALL the sites been united behind Labour from the start, Labour might well have held it

It maddens me that errors there meant @emmadentcoad lost out (and Emma, I might have contributed to that - and for that I am unreservedly sorry - although I recommended you in the end)

However all this the-sites-disagree-is-a-problem stuff before the election looks like a storm in a teacup overall

And then what about the other way? The "saves" that might have fallen had tactical voting not happened?

There are 7 Labour held seats won with a margin of less than 1%, and a further 11 won by a 1-2% margin - see: electionpolling.co.uk/battleground/t…
ALL of those had unanimous recommendations from the tactical voting sites

Even if the effect Hanretty has found for Lib Dems is less for Labour seats, I am pretty damned certain all this effort *saved* Labour from greater losses - perhaps of a dozen or so seats

The .xls sheet of all my workings for this is here:
techpolitics.eu/downloads/tact…

Comment and critique most welcome!

/ends