The maximum value of this is 1 and the smallest value is 0. As written, if every person upvoted a post, it would be 1, but not controversial, so that wouldn't work. Say you slightly modified it and used downvotes / total votes, at first glance it seems sensible because then if everyone upvoted, you get 0, indicating that it's not at all controversial, a sensible result. But if everyone downvotes it, you get 1, indicating maximum controversy, but there was no controversy. Everyone agreed it was bad.

With variance, if something is unanimously liked or disliked the variance is 0. If it's split 50/50, the variance is greatest. You want something to adjust for the sample size, so you could use variance * total votes. It is similar, if not identical, to a binomial distribution.