It is not clear that points are so useful, as they creates the atmosphere of a social game with reputation, rather than a serious site for doing science. The game aspects can draw in many users quickly, that's for sure, but serious content draws in the users of the kind you want.
A simple idea is that the reputation for the answer is by the difficulty of the question, so that it is the product of the upvotes on the question, times the upvotes on the answer.
The idea is that there is a "difficulty" rating for the question--- generally, you upvote questions which you think are interesting, but perhapsnot difficult. A difficulty rating is useful too, so that people can upvote interesting but easy questions, so as to say that they are interesting pedagogically, without saying that they are difficult research-wise.
Then the up and downvote can be weighted by the difficulty. So that a +84 upvoted question (which is certainly difficult to answer) with an answer that is correct with (+40 upvotes) can give a reputation equal to the product of the difficulty times the accuracy. You can also increase the difficulty automatically depending on how long the question goes without an answer, by an upvoting bot which +1's the question every day, so long as no upvoted answer is present.
There should be a way of certifying an answer is correct and complete which is not dependent on the original person asking.