I'm harvesting results from the BBC website, and comparing them with estimates from ACPO (http://apcoworldwide.com/docs/default-source/default-documen...).

You can see a feed of results here: https://gist.github.com/inglesp/0b0ba9efcefe54f965be9b042afd...

Caveat: I'm drinking increasingly heavily.

The New York Times has done live predictive models as results come in, which is the best election night tool I've ever seen. Here's the New York Democratic primary: http://www.nytimes.com/interactive/2016/04/19/upshot/live-mo...

All of the charts updated in real time based on how precincts actually voted compared to their predictions.

I agree completely...a shining example of how there's always more data to the data. I remember in 2014, they were tracking a Senate race in which the underdog, because of the order of precincts reporting, was seemingly on its way to a massive upset. To its credit, CNN made a few caveats while at the same time talking about how exciting things were as they did Jerry-Bruckheimer-zooms of their fancy electorate map...but all the while, the NYT kept calm and steady with its prediction and as the night went on, the vote results line gradually converged to nearly exactly what the NYT's analysis predicted.

It's definitely made election night a lot less dramatic.

I think that's a nice idea, but it's also exceptionally difficult to do in practice for something like this.

In the case of a general election, we have the previous election on which to base our expectations. The UK broadcasters use the "swingometer" (https://en.wikipedia.org/wiki/Swingometer) in these elections, which is a useful visualisation of the change of vote share on a per-consitutency basis, indicating the meaning of the result rather than just the number.

The EU referendum has no recent comparable campaign with which it can be compared, making it much harder to do this.

Just to piggy-back, these maps often don't show urban density very well. I'm guessing that Kettering (which is reporting "Leave") has a few less people than say, London.

Similarly, here is the 2012 election county map where "blue" had 5MM more votes than "red": http://politicalmaps.org/wp-content/uploads/2012/11/2012-usa...

That's a good point.

The Guardian do good election maps – this one has regional results scaled by population, so it's much clearer what's happening: http://www.theguardian.com/politics/ng-interactive/2016/jun/...

The Guardian is currently running an excellent live blog (http://www.theguardian.com/politics/live/2016/jun/23/eu-refe...) which includes comparisons between real and expected results, as well as projections by experts. I wish that link were at the top of HN instead of this submission; live results are completely useless for the average observer.

If you watch the BBC live stream they do in fact give information on how the results fair in terms of what was expected. This information appears to be missing from their website however.

I would like to second this -- they have a rather neat distribution that shows areas arranged in terms of leave-to-remain ratio. So as some results come in, they point out where that area was predicted to come, and whether or not the results are as predicted or surprising.

See my comment below. If you are interested in this and want to get some in depth analysis check out either BBC 1 or the BBC live blog. They cover these questions.

It's hard though, there isn't much to compare as there isn't an 'expected way' regions vote. There's a rough idea, but it's seeming like these are not what people expected. My money is on Scotland and Ireland being overwhelmingly remain, Wales swinging unexpectedly to leave (partly because they are disenfranchised with the government), while in England the non urban areas going leave and the urban ones going stay.

BBC Radio Four is doing a reasonable job of saying what's expected in each area.