Does my new chart make sense?

[ciphergoth]

Imagine for a moment that it's 2:30 AM London time on November 5th and like me, you've stayed up late to watch the US elections. Six states have been called so far: Vermont and Virginia for Obama, and Indiana, Georgia, South Carolina and Kentucky for McCain. You go to my election page and see this diagram:

Does it make sense to you? How could I improve it?

thanks!

Update: wow, a flood of very helpful comments, thanks so much, keep them coming!

Comments

[clarisinda.livejournal.com]

I understand very little about the US election process, so I'm afraid that means very little to me :-/ Surely you would have an explanation as to what it represents?

[ciphergoth.livejournal.com]

You need to understand the electoral college to have a hope of making sense of this diagram, I'm afraid!

[atommickbrane.livejournal.com]

Um! Rotating 90 degrees? :)

*twists head*

[ciphergoth.livejournal.com]

Interesting idea! Any other votes for flipping the diagram to go the other way?

[silkyraven.livejournal.com]

It makes sense to me.

[robot-mel.livejournal.com]

Is the line in the middle how many votes for the college they need to win? Also it'd be good to see the difference between suspected and confirmed. (Or am I just missing that?)

[ciphergoth.livejournal.com]

States that haven't been called are in the middle; those that have are at the edges. Or do you mean something different?

[clarisinda.livejournal.com]

Yeah, I thought that too. Perhaps put the confirmed results in a brighter colour or something?

[hukuma.livejournal.com]

I would put a more clear separation between the called states and the uncalled ones. Also, I think you're using the 2004 results for colors, is that right? That took me a long time to figure out.

[ciphergoth.livejournal.com]

Yes, 2004 for colours, polling data for the stuff in the middle.

A more clear separation - maybe a background colour thing?

[andrewducker]

What it says to me is that two Blue states went Red. But then I looked at your explanation and now I don't understand why Virginia is red and all the way to the left in the middle of the blue states, when it hasn't been called yet.

I was assuming that the distance each one is to either side is the probability they have of falling either way - but that can't be the case (because of Virginia). I also don't understand why some are highlighted - I thought it might be "called" ones, but that also doesn't seem to be the case.

So some explanation would seem to be necessary.

[ciphergoth.livejournal.com]

In this scenario, Virginia has been called.

Vertically it falls into three sections - states called for Obama, states not yet called, states called for McCain. The height of the bar shows how much they favour one candidate over the other, either in final results or projected results, and the bars are arranged in height order. Probabilities don't come into it.

Iowa is red because it Bush won it in 2004. Colours alternate just to make the diagram easier to read.

[simont]

Without a key it doesn't make much sense to me.

The width of each bar appears to be based on that state's number of electoral college votes, which is also the number in parentheses after the state's name.

But then there are three separate things represented on the diagram which look as if they're related to which way the state votes or voted: the colour of the bar (blue or red), the height of the bar (and in particular whether that height is positive or negative), and the position of the state on the x-axis. Clearly all of these are largely correlated but not completely; so clearly they're all representing three correlated but distinct things. But which? Why do some bars have positive height but are red rather than blue? What's the significance of there being a couple of red bars on the mostly-blue side of the diagram?

[fizzyboot.livejournal.com]

An idea: have the graph and have a written explanation of it as well.

[martling.livejournal.com]

It would help to ensure that the labels didn't cross the boundaries. I looked at your list and thought "wait, it looks like Oklahoma has called too?" before following the lines.

And what's with the blue/purple colour alternation when there's no such thing for the red?

[ciphergoth.livejournal.com]

It would help to ensure that the labels didn't cross the boundaries. I looked at your list and thought "wait, it looks like Oklahoma has called too?" before following the lines.

Unfortunately that's only possible if there's room to fit the labels for the called states in the space available for them!

The red colours do also alternate, but maybe it's more subtle.

[elfy.livejournal.com]

except iowa i get it.

[ciphergoth.livejournal.com]

Iowa is red because Bush won it in 2004.

[boyofbadgers.livejournal.com]

This would make an awful lot more sense if the axes and lines were labelled. As it is, it took me a good couple of minutes to twig what was going on.

[hukuma.livejournal.com]

Man, this picture really drives home the point that someone else made, which is that to win, McCain needs to take a state where he's currently down by 8 points. Makes me feel good about the election (though after the last two, still very nervous.)

[ciphergoth.livejournal.com]

Actually not quite - if he takes Virginia and all the states in which he's polling better than that, he wins, so it's more like six points than eight.

Thoughts

[wechsler.livejournal.com]

Up / Down for who's winning is difficult; I agree with those turning it 90° CCW.

I get the area/height thing, that's fine.

Having colour represent "last time" is difficult and counter-intuitive. Having label colour for this time and block colour for last time may better indicate the idea of a change.

The bit that gives me real trouble though is that you've got two qualities declared on the same (currently horizontal) continuum; vote proportion and called / uncalled. You've also got lines representing both centre point and status boundary.

There'd be an obvious solution if everyone had 3d monitors...

I also find the states moving *outwards* as they declare unintuitive. Pending states should be "in the wings"; declared should be "centre stage". I'd put a small (5px?) gap instead of the outer lines.

I'd be inclined to put uncalled states in outline only, too. That way states fill out and move centre stage as they declare.

(The "stage" metaphor actually works better if you *don't* turn 90°)

Re: Thoughts

[alextiefling.livejournal.com]

I second the outline idea; that would make things a lot clearer.

[alan1957.livejournal.com]

Apologies if everyone has already said this but I don't have time to read the many comments.
I find it unclear as to which states have declared and which are still to do so. The thin grey lines are not enough (for me). Is it possible to use differing shades of red and blue...darker for those already declared, pastel for those still to come in. Although I (and many/most?) of your 'readers' know that blue is used for the Democrats and Red for Republicans I would still make that clear, and also what the differing shades mean on the graph itself.

I hope that helps a little.

[wechsler.livejournal.com]

BTW, what are you using for chart generation? I was going to try and use SVG for my current project, having seen how well it works for CruiseControl / PHPUnderControl.

[ciphergoth.livejournal.com]

Python and Cairo...

[xquiq.livejournal.com]

Makes sense to me.

The only minor thing I wasn't clear about was e.g. whether the swing in Virginia was predicted by the polls (are the results going as expected), but to be honest, I'm not sure that type of information could be included without reducing the clarity of the chart.

Am I correct in thinking that the chart currently predicts an Obama win by 4 votes in the electoral college ?

[ciphergoth.livejournal.com]

I'd do a separate chart for how the polls did against reality; I have scattergram software for drawing that.

Forgot to answer your other question - no, this chart, in the (so far) invented scenario I describe, predicts an Obama win by 95 points. The colour of the states just indicates how they went in 2004; it's which side of the midline they are that indicates how they are predicted to go this time.

[weegoddess.livejournal.com]

I think, simply, that it's quite cool that you came up with this. And I love that you made an icon of it.

[ciphergoth.livejournal.com]

Actually I made an icon of it because I accidentally deleted my old election icon and needed a new one in a hurry :-)

I think it will be much improved by the suggestions I've got here!

Edge cases

[mooism.livejournal.com]

How will your chart display a state where the polling says it’s tied?

How will your chart display a state where the networks have called it as being too close to call?

Re: Edge cases

[ciphergoth.livejournal.com]

Currently tied states are labelled but otherwise pretty much invisible, which is bad if you get two in a row because you can't see the border between them, but if there's just one then the states to either side will mark its position.

A state will stay in the middle area until it is called one way or the other, even if as with Florida 2000 that takes a while.

Good questions - thanks!

[alextiefling.livejournal.com]

What polls are you using?

I'm slightly confused by the fact that the area of each bar has no particular meaning. I keep wanting it to be a histogram.

If I were doing this, I would make the height of the bar the number of college votes, and the width uniform, and either leave the polls out of it or use some kind of error bar to show the current status. I'm not saying you should do what I'd do, though.

[ciphergoth.livejournal.com]

Yes, I keep wanting the area to mean something. If EVs were proportional to population it would.

[friend-of-tofu.livejournal.com]

REVOLVING FOX HEADS ON POLES!!! (http://www.guardian.co.uk/commentisfree/2008/oct/20/uselections2008-barackobama)

Sorry, not helpful at all. But it is My Husband.

[shevek.livejournal.com]

Hang on, are you one of the fivethirtyeight people?

[ciphergoth.livejournal.com]

No, the name on it refers to the source of the data. I'm going to rejig that, because it is rather misleading!