West Virginia Prediction

For WV I'm predicting Sanders 55.3 %, Clinton 44.7% (percent of the shared Sanders + Clinton vote).

I am also predicting a higher than average error due to fluctuations in the google search trends. The ratio of Sanders:Clinton searches has ranged from 2:1 to 1:1 - and in my model this is a difference of 5.7%. I'm using the last 7 days (as of now), so Sanders does well due to winning Indiana and getting lots of press coverage from that (and I think he had a mass rally or two during this time). By contrast, the past several days, including the weekend when you would expect the Sanders volunteer phone banks to have an impact (and also any advertising campaign) - the ratio was closer to 1:1.

If the 1:1 ratio is more accurate, then my model would give Clinton a tiny edge (50.4%).

There has also been a lack of polls. I'm using the 538 polling average which is based on two polls both of which are over a week old. Up until recently Sanders has been gaining fairly steadily in the national polls, but recently the trend had reversed and he is losing ground as his chance of winning the nomination is below 10% (and based more on scandals than electoral victory).

Lots of Trump supporters

Lots of Trump supporters (and "Independents") voted in the Democratic primary and they voted for Sanders (and probably the third and fourth place candidates) - making for weird results.

Sanders 51.4 / Clinton 35.8 / Other 12.8

As a share of the Sanders + Clinton vote (87.2%)
Sanders 58.9 / Clinton 41.1

So my prediction was off by 3.6%. By contrast BenchMark Politics did the best with a 57/43 call. My prediction did beat 538, the polling average, and Tyler Pedigo - but that could easily be luck.