The Senate snapshot is in the right sidebar. It probably won’t change much by Tuesday. Here’s the full documentation…
States calculated: Alaska, Georgia, Kentucky, Minnesota, Mississippi-B, and North Carolina. Democrats have 54 safe seats, Republicans have 38 safe seats, and the two Independents (Sanders and Lieberman) are assumed to caucus with the Democrats.
Win probability: This is determined in a similar way as the Presidential race: Take the two-candidate margin in the last 3 polls and the last seven days of polls. The median margin divided by the estimated SEM defines a z-score, which is turned into a win probability using the t-distribution. (The Meta-Analysis uses a normal distribution, which needs to be fixed for next time.)
Georgia: This is a special case. If nobody reaches 50 percent, a runoff is forced. Although this is an unlikely event, it is currently where most of Martin (D)’s chances are. The probability of a runoff is estimated as (# surveys nobody reaches 50% of decided voters)/(# all surveys). In the runoff, Libertarians are assumed to split anywhere from 30:70 to 70:30 with a uniform distribution.
Finally, the entire exact probability distribution is calculated as a multinomial.