538 maps Facebook likes for each candidate by county

[/]

If we try to find way to use this to predict voting patterns, we can use:
Iowa: Sanders 23%-Clinton 9%, +14, actual results Clinton 50-50
New Hampshire: Sanders 33%-Clinton 8%, +25, actual results Sanders 60-38

So a Sanders+14%like lead in a tie, while 25-14=11, multiplied by 2=22, Sanders margin in NH.

This means that Sanders predicted margin of victory = 2(%like lead-14)

Nevada: Sanders 23%-8%, +15, predicted margin: 2(15-14)=2, Sanders wins 51-49
South Carolina: Sanders 11%-4%, +7, predicted margin: 2(7-14)=-14, Clinton wins 57-43

math

[/]

Just for fun I ran the same equation for the Super Tuesday states:

Alabama: +5, -18, Clinton 59-41
Arkansas: +6, -16, Clinton 58-42
Colorado: +27, +26, Sanders 63-37
Georgia: +7, -14, Clinton 57-43
Massachusetts: +27, +26, Sanders 63-37
Minnesota: +22, +16, Sanders 58-42
Oklahoma: +9, -10, Clinton 55-45
Tennessee: +8, -12, Clinton 56-44
Texas: +8, -12, Clinton 56-44
Vermont: +73, NA, Sanderslide
Virginia: +11, -6, Clinton 53-47

You know, Castro, you might actually be on to something...

[/]

Here's a national county map using Castro's formula. So far, it correctly predicted Sanders winning every county in New Hampshire, although it was a bit generous to him with the margins, and (excluding ties both in the model and in reality) it got 60/90 counties in Iowa, predicting 23 Sanders counties would go to Clinton, and two Clinton counties for Sanders, for a generally Clinton-favorable prediction. But of the 100 non-tie (in model and in real life) counties to vote so far, it's correctly predicted 70, which is a pretty darn good record for just a formula and Facebook likes.

That must have taken a while . Great work!

But I'm having a little trouble believing that Sanders will win Chicago, LA, and Detroit.

Yay @ my county going for Sanders