In a previous discussion, the general consensus was that whether or not to upvote is, for most people and most questions, a function of the number of existing votes. To put it another way, the number of votes represents a sort of 'post quality'.
This, in conjunction with the up/down-vote asymmetry, leads to an interesting possibility.
Lets say an answer is 'worth about a 4' to most people, and it gets downvoted. It is subsequently more likely to be upvoted. In fact, the chance of being upvoted becomes higher the more downvotes it gets, as the pendulum swings away from equilibrium.
This damping loop is obviously by design. Answers will tend to converge on their 'true value', as determined by the audience. But the reputation adjustment is skewed in favour of upvotes. It takes 5 downvotes to match one upvote.
So, if you had an endless stream of views, and an answer that:
- Was downvoted every few votes
- Was perceived to have a positive value to most people
Then you would have an infinite reputation generator.
There are some special situations that can reach this state. Here is an example of an answer I had like this: 3 downvotes, 5 upvotes, interleaved in quick sucession. It's not a great answer; taken in conjunction with the others, it's worth probably 1 vote. But the key feature is that the answer wasn't stable until after the explanatory comment and my subsequent edit.
The reason was that the original answer was ambiguous; it was technically correct, but not considered good form. The first two downvoters left no comment, so this was not made explicit. It was clearly 'worth something' to a casual observer, but offended those with relevant experience (the other answerers, for example). The original question was a rep generator.
The point is, an answer that was worth 10-20 reputation actually generated me 44 - almost as much as the top answer.
Another, more familiar and extreme example is with polarised opinions, but usually these are nullified by the community wiki attribute. Facetious or moderately funny (but not too funny!) answers that appeal to some but not others are a third common example.
There are some relatively subtle consequences:
Downvoting rival answers (that are actually reasonable) on a question you are answering will backfire in the medium term. So not only is that an ugly gaming strategy, it doesn't even work.
Downvoting for any reason unrelated to answer quality will backfire, long term. This implies that behaviour such as 'revenge downvoting' is at best ineffective, and will probably cause a net reputation gain over time for the targeted user.
Answers that are good enough to get a vote, but controversial, misguided or imprecise enough to annoy knowledgeable users will reap much more reputation than they are worth.