You might want to factor in karma density as well. Maybe something like:

    w = ln(k * m)

    w = weight of vote
    k = karma of voter
    m = median points per comment by voter

Instead of feeding into the post's score, it could also feed into the 'buoyancy' of a post (how long a fresh post gets to stay at the top of a page before beginning to sink to the bottom).