if we had decent poling with big sample sizes, we could make some better projections, but that rarely happens in senate races. when low quality polls with small sample sizes, bad methodologies and high margins appear to produce surprising outcomes, that may say more about the low quality of the polling than the likeliness of the result being projected.
if i did a poll of 500 people online that told me that piss tasted better than apple juice, i might question the accuracy of the poll before i changed my grocery list.
if these polls are completely wrong and trump wins kansas by 10 points, that should not surprise anybody. these polls suck.