We've experimented with vote weighting over the years but the results have been disappointing. It's not easy to come up with criteria for users whose votes are more reliable.
So perhaps recruit some small group of math minded HNers and see what insights and ideas they wring from access to the fully voting history?