I'm guessing I'm falling into an edge category here but I wanted to point out a slight issue in the auto-awarding of bounties.
I'm using an example from English Language Learners but this can happen on any site:
This question has several answers. The top rated answer (mine) has 12 upvotes and one downvote. Of these votes, 7 up and 1 down (I think) were awarded after the bounty started.
The OP edited the question severely, asking more and more questions until the entire question was reverted to the original but I edited my answer, giving the information that was requested.
Even if you only include votes after my edit on 7/20 (the day the bounty was placed), I have 4 upvotes and one down, which is more votes than the other answers (the highest other answers have two total three up and one down or two up and no down).
I'm not saying that the OP won't award me the bounty, but is it really acceptable for a lower-rated answer to get the bounty simply because it was answered after the bounty was started? Particularly in the case that the answer has been edited to address the bounty reason.
If I had known this, I could just as simply have deleted my answer and reposted it, just to be eligible for the bounty... which is certainly not the behavior we want.
If we're auto awarding bounties based on what communities think is the best answer (through votes), should we not consider all answers?
When a bounty is set on a question, the system remembers the beginning vote count on the existing answers.
When the bounty does not awarded by the bounty poster, the system compares the vote status of all answers.
- Answers submitted after the bounty was placed use their ending value.
- Answers submitted before the bounty was placed use their ending value - beginning value.
If we want a secondary restriction, we can require that the OP of the answer must edit it to be eligible (though this could lead to people editing simply for eligibility reasons and not to make substantial changes, so I'm not sure how I feel about it).
Alternately, we could have a window of time prior to the bounty being placed that makes the answer still eligible, like a week or a month. This would prevent very out-of-date answers from being auto-awarded (as they're likely not correct any longer) but allow the award to go to a more recent, correct answer that just happened to have been posted shortly before the bounty.
A little epilogue... the OP of the question mentioned above did not award the bounty, so an answer with only +3 (vs my +13) has received the half-value bounty.