Second draft of chart

[ciphergoth]

I've incorporated some but not all of your suggestions - I don't quite have time to describe why I made the choices I did, I'll do that in a later post.

Remember, this is a made-up scenario to test how this might look during election night

Explanation

To make sense of any of this you must understand the Electoral College

Red states are states that went for Bush in 2004
Blue states are states that went for Kerry in 2004

States against a blue background at the top have been called for Obama
States against a red background at the bottom have been called for McCain
States inbetween have not been called.

States to the left of the line have gone to Obama or are projected to go for Obama
States to the right of the line have gone to McCain or are projected to go for McCain
The further left or right they extend, the further the (projected) margin of victory

The wider a state is, the more electoral votes it has.
The numbers inbetween states near the finish line mark how far that boundary is from the finish line.
The projected winner is the person who takes the state on the finish line.
Once the area for states called for one person or another crosses the finish line, the election as a whole can be called for them.

Each state also carries a code such as "CA (55) 18%", which means that CA (California) has 55 electoral votes and is projected to go for Obama by a margin of 18%.

I know the red/blue state thing caused a lot of confusion but it's dead important to the pol junkies, so I've just put it right at the top of the explanation! It also makes the chart prettier - more colourful :-)

Comments

[andrewducker]

Could we have a white line between Projections and Called states?

At the moment it reads "Projections Called for McCain" at the bottom, which isn't right :->

Other than that it's all good.

[andrewducker]

Looking at your explanation you say that the background is red/blue for the called states. On my monitor the difference from black is so close to not visible that I had to stare at it from about 3mm away to be sure it was there at all. Lightening it it somewhat might help.

[palmer1984.livejournal.com]

Much much clearer :)

I should probably know this, but by "called" you mean - votes have already been cast, right?

[lovingboth]

"The result has been announced".

[palmer1984.livejournal.com]

Ah, thanks!

[mooism.livejournal.com]

But initially, “a projection of the result has been announced, based on exit polls”.

[mooism.livejournal.com]

“Called for” initially means the exit polls indicate that that candidate will win.

They don’t use exit polling from areas where the polls haven’t closed yet. For some states split across two time zones, where the polls therefore close at different times within the state, they will use exit polling from the areas that have closed, even when votes are still being cast further east in the state. Read notes above/below this table.

As the votes are counted, and individual precincts release their local tallies, these are incorporated into the predictions for a state.

Exit polls can initially be inaccurate. e.g. Florida in 2000 was initially called for Gore before being reclassified as being too close to call, and later on that night called for Bush.

[mooism.livejournal.com]

s/further east in the state/further west in the state/. Ahem.

[ciphergoth.livejournal.com]

It's basically eg CNN saying "we are now confident based on exit polls etc that such-and-such a candidate will take such-and-such a state".

[lovingboth]

Is it possible to have a visible indication of how the states' actual results differed from the poll projections? Some vertical line so that the nearer it is to the actual result, the more accurate the polling?

Apart from that, fab! This format works better than the other one for me.

The thing that gets me about US politics is that the right wing uses red :)

[softfruit.livejournal.com]

The thing that gets me about US politics is that the right wing uses red

I coped with it last time around by imagining the two sides as being Man City and Man Utd ;)

[mooism.livejournal.com]

I think there needs to be more emphasis of what the overall prediction is. I’d go for subtly different background colours where there is black now, but a labelled predicted overall result line between Missouri and North Dakota could work too.

Perhaps change the “projections” labels to “projected for Obama” and “projected for McCain”.

[simont]

Provided there actually are any states in each category, of course. Near the end of the process, it wouldn't be at all surprising to find that the last few uncalled states were all on the same side of the line.

[mooism.livejournal.com]

Yes, and presumably the “called for” labels are not present when no states have been called yet, and the labels are thus uncalled for.

[lizw.livejournal.com]

I find this much clearer and more intuitive - thanks! Now that it's clearer which states are called for Obama, I think the red for Virginia is also less troubling, although of course it's difficult to judge whether that's just because I read the comments on the previous post. Just one small FYI: in Google Chrome, the red background for McCain's called states is almost indistinguishable from the black background above it, at least for me. It may be my eyes doing something weird that other people's don't do, but at first glance I read this as no states being called for McCain. When I realised that couldn't be right and looked again, I saw it correctly, but when I glanced back at the chart while typing this comment, it started looking as if every state from Texas down was called for McCain - possibly some sort of optical illusion linked to the size of the red area for Texas? Obviously you can't change that, but there may be a better shade of red, or even pink, that would avoid the problem, assuming it's Not Just Me. The blue for Obama's called states is lovely and clear :-)

[simont]

Whereas I find the red background clearly visible but the blue one a bit dim (just to be contrary). Perhaps just reinstating the dividing line in each case would help show up the contrast?

I like all the changes, though. Vertical format definitely an improvement, backgrounds for the called states a good idea regardless of quibbles over the precise RGB values used, finish line an excellent addition, labels at the various dividing lines very helpful.

[yoyoangel.livejournal.com]

Much clearer to me (in that I understand it, this time round, and hadn't a clue with the previous version: maybe I have a horizontal brain and needed a vertical one, or vice versa).

This is gorgeously geeky!

[purplerabbits.livejournal.com]

The finish line was slightly confusing me before, but if I read this version right, Obama only needs down to new Hampshire to win, but is actually projected to get the states between newHampshire and Montana as well?

Re: This is gorgeously geeky!

[ciphergoth.livejournal.com]

Yes, exactly (not including Montana itself),

[mooism.livejournal.com]

It strikes me that at 2:30am I might care more about when the polls in a state close (and thus when I can expect the state to be called; this would make to-close-to-call states more obvious) than which way the state voted last time. (And I count myself as a pol junky.)

[drdoug.livejournal.com]

This is way cool, and a definite improvement. Proper keys FTW!

For my money, though, and going against the consensus, it'd be better back to horizontal. (With the labels and additions you've made.) With an information-dense graphic like that, I prefer to be able to see as much of it as I can at once, and having it vertically means I can use half of my monitor area to look at it. I can see the left-leaning/right-leaning state thing, but I prefer the conception of the two opponents on either side of the screen, and advancing towards the final balance point. And basically the winner is the one who takes more of the centre (or rather, center) ground, which is another of those psephological truisms.

(I remember back in the day when there were still OHPs, and someone convinced me that landscape was the One True Format for displays by arguing - not entirely spuriously - that a) our eyes track side-to-side more easily than up-and-down, and b) ten billion TV sets can't be wrong.)

The colours and contrasts are Ok but a bit iffy for me too - I fear you're straying too far in to the terrible world of monitor gamma diversity.

[ciphergoth.livejournal.com]

All of these are pretty good reasons to go horizontal, but at this point I don't think I'm going to go back to it. Fitting into a normal browser width without scrolling is a big benefit, and being able to put labels horizontally is a big boon. Plus it's quite a lot of coding effort to switch between them :) Mostly it's the area around the finish line that really matters, so it's not so bad that you can't get it all on screen at once.

It sounds like most people would do better with a brighter background on thee called states. How come we still have monitor gamma diversity in these LCD days?

[alextiefling.livejournal.com]

I'd recommend sticking with a websafe palette. You should have fewer gripes of this sort.

[drdoug.livejournal.com]

How come we still have monitor gamma diversity in these LCD days?

Four main things I can think of: a) people set their own contrast and brightness differently, and that makes a huge difference; b) the Mac/PC differences in both encoding and decoding gamma are still there; c) LCDs vary even more widely in producing nonlinear signal:response terms than CRTs, and not all calibration is done perfectly; d) people's eyes/perceptions/tastes are different. And then there's a small bonus number of people who go mucking about in their monitor's gamma settings by hand, but that's probably a smaller effect.

[stgpcm.livejournal.com]

perhaps we could have a less black uncalled statue - a grey area, so to speak?

[stgpcm.livejournal.com]

perhaps we could have a less black background for the uncalled states - a grey area, so to speak?

[alienspacebat.livejournal.com]

only thing I'd add is when a state is called, I'd also like to be able to see the predicted result and how much the change was. perhaps a small line (a bar top effectively) could persist or if this is web only it could be on mouseover

[ciphergoth.livejournal.com]

Yep, several people have asked for this - I'll do it with a separate diagram using a scattergram.

Another edge case

[mooism.livejournal.com]

Maine and Nebraska allocate their electoral votes differently from the other states. Instead of winner-takes all, they elect one elector based on the vote each individual congressional district and the last two electors based on the total state-wide vote.

Normally Maine is uniformly enough (D) and Nebraska uniformly enough (R) that this makes no difference in practice. But this time Obama has a realistic chance of winning an elector in Nebraska. However, your graph doesn’t seem to allow for this situation.

Re: Another edge case

[ciphergoth.livejournal.com]

Yeah, I know about that one - if it happens, I'll split the state up into separate bars. It sounds like Obama probably won't win any electors in Nebraska so it probably won't arise.

[mooism.livejournal.com]

“States to the left of the line have gone to Obama or are projected to go for Obama
States to the right of the line have gone to McCain or are projected to go for McCain
The further left or right they extend, the further the (projected) margin of victory”

Will the projections be static (based on non-exit polls and not updated) or dynamic (initially based on non-exit polls and then updated as exit polls and real votes come in, possibly inclusive of demographic correlations between states)? I was assuming static until I read that paragraph.

[ciphergoth.livejournal.com]

Depends on what fivethirtyeight.com do - I use his projections...

[softfruit.livejournal.com]

Adding my voice to the "yes, that works better for me" chorus. Thankyou! :)

[alextiefling.livejournal.com]

That's definitely better for me. Can we avoid red text on a red background and blue on blue likewise, though, please? Thanks.

[henry_the_cow]

I agree this is a lot better, but like some other people I have problems with the background colours. In my case, it's hard to distinguish the blue from the black. Would it help to change the black to grey, or even a third colour such as pale yellow or green?

I also agree with the suggestion of a white line between the projected and called states. And the labels do seem to read "Projections Called for McCain" - perhaps they should be in the middle of each section rather than adjacent to the borders?

Someone else commented about how it will display when any of the four areas are empty (i.e. right at the start or the end). I hadn't thought of that but I assume you have some test data sets for those cases.

In the explanation of the state labels, you say a figure of 18% means that the state is projected to go for Obama by a margin of 18%. Does this mean that states projected to go for McCain will have a negative percentage (which could be perceived as bias in the chart), or would it be better to say that the state is projected to go for "the indicated candidate" by that margin?

[paulthegeek.livejournal.com]

Very nice. I like it a lot.