Sometimes show highest answer before accepted answer?

Question

For some questions the accepted answer is not that useful but the OP likes the answer for whatever reason. Later possibly, someone else comes along and provides a better answer (which ends up being rated much higher). For examples you can just look at the populist badge.

I wondered - if an answer has more than 20 votes and more than 2x the votes of the accepted answer should it show first in the list?

Justification

A great deal of the popular questions are hits directly from google. The user is going to read the answer, try it and move on to searching google further. Occasionally you will have users that want to read further into the other answers but that isn't always the case. This is still probably important because I'm guessing a good percentage of stack users operate this way. Just theorizing a bit.

One More Thought

The OP is may not be an expert (tons of questions are asked by first time users). So we are putting the first answer that appears in the hands of someone who probably doesn't care all too much. I agree their vote needs to count more than others (Since they know what they needed), but to always show it first may give them too much power.

Answer 1

The OP is may not be an expert (tons of questions are asked by first time users). So we are putting the first answer that appears in the hands of someone who probably doesn't care all too much.

That's a very compelling argument, in the way that you put it. Questions are often asked from limited perspectives, and sometimes a user doesn't even know just how useful the feedback received is. and Pekka makes a good point when he said:

There's nothing more embarrassing for the site than an incorrect accepted answer. There's plenty of those, and the general consensus is the check mark isn't that important. Why should it make a difference in the sort order then?

Also, Bill-the-Lizard points to a useful query here showing many questions that are accepted but have a negative score

Also, On very popular questions, currently it's sometimes a bit of hassle to scroll down to the find the best answer if it isn't the chosen answer.

Answer 2

What should be the focus of the sorting order on a question

Ultimately the sort order should be for one thing, and one thing only: putting things the person viewing the page in that moment is most likely to find useful.

Once you've got that straight it is possible to see past the arguments that "the OP has a right to accept whatever answer they feel best answers their question". Of course they do, but that is entirely separate to everyone else's sort order, which should be for their benefit, not the question OPs.

What data do we have available?

Given that we cannot know what will actually be most useful to the OP we have to look at the data points we have available. These are as follows

• What the viewer is sorting by; votes, activity or newest
• Which answer was accepted
• What the score on the answers are.

What should be the sort order?

The viewer is sorting by activity or newest

Taking the accepted answer into account here at all seems unhelpful. The person clearly is looking for something that happened recently, not what answer is the best. More times than I care to remember I have seen activity on a question, sorted by "activity" to see which one was most recently edited and starting to try to figure out how the top answer (the accepted answer) was edited. Only to realise it isn't actually that answer that was edited - this is irritating. I would suggest in this case ignore the accepted answer altogether for sorting.

The viewer is sorting by votes

In this sort the viewer is clearly looking for some question of quality, so the accepted-answer-tickmark has some value as a data-point in this case. But it certainly doesn't have more value than a potentially infinite number of votes; if one answer has an accepted-tickmark and no votes and the other has 1063 votes I would be pretty confident on which one was more useful. Worse still if the accepted answer is downvoted.

Deciding how much an accepted answer is worth compared to an upvote is probably challenging. My money would be no more than 5. But that will create a somewhat confusing sort order where the accepted answer appears seemingly randomly in the list therefore we would need some form of rule that didn't lead to this confusion.

A lightweight solution may be to relegate downvoted accepted answers which certainly has merit and would not be too complicated a rule but would not deal with accepted answers which are "fine but not great".

A somewhat more extreme solution may be to relegate the accepted answer to a tie-breaker; if two answers have the same score the accepted answer goes first. This has a key advantage of giving the viewer what they actually asked for; "sorting by votes" while still using the Accepted-Answer data point. This I feel under values the accepted-answer-tickmark but is probably the best solution that doesn't "look strange".

Answer 3

Your analysis is fundamentally flawed because it is completely based on observation bias. You see questions where the highest-scoring answer is better than the accepted answer and conclude that this should be done all the time.

If the accepted answer wasn't shown at the top, there would be a feature request for it to be shown at the top, to allow the asker to convey that no matter how much the community loved one answer, this other answer is the one that actually works.

Supposing that there is one answer which is objectively best, there are several categories of threads:

1. The best answer the highest-scoring one and is accepted.
2. The best answer is neither the highest-scoring one nor the accepted one.
3. The best answer is the highest-scoring one but not accepted.
4. The best answer is accepted but is not the highest-scoring one.

In case (1), there is nothing to change. In case (2), there is nothing to do unless you have a different way of recognizing the best answer. So the question is which of (3) and (4) should be given preponderance.

To decide which one of (3) and (4) to favor, one factor is which one happens the most often. Yours isn't the first request to favor (3), but so far I haven't seen anyone put forward evidence that (3) happens more often by (4).

Unless the frequency argument strongly favors (3), there is another reason to prefer (4). The highest-scoring answer is guaranteed to come out second at worst. On the other hand, if the asker determines that a different answer is better, then the only way they have to make it be more visible is to accept it: otherwise this answer may be buried anywhere in the list.

This is especially useful when a new answer comes along a long time after the question was posted. By accepting the new answer, the asker can convey the message that this new answer is better, more up-to-date, etc. — this is a lot more useful than a comment below the highest-scoring answer, especially if that answer already has a lot of other comments.

Yes, sure, I've seen plenty of questions where the asker has accepted an early bad answer and a better answer outscores it. But askers who do that tend to ask mediocre questions that aren't very useful anyway. There is a strong corellation between askers who maintain their questions — including accepting new answers when they're better than what was there before — and askers who ask widely useful questions.

Thus putting the accepted answer on top has many advantages:

• It provides an opportunity that is completely missing from the opposite choice.
• It has a better chance of being the right thing on useful questions.
• It doesn't do much harm: at most the highest-scoring answer will come second; and if the accepted answer has a negative score, this is a sign to readers to take it with caution.

For these reasons, I am strongly against moving the accepted answer away from the top¹ (no matter what its score is), unless it is demonstrated that there is a significant majority of cases where this is the superior choice.

I'd be more in favor of allowing some way to have a third party accept answers on abandoned questions, which should handle most cases of accept-the-first-answer-and-move-on. I don't know what form this would take. In any case, if a thread is really bad, there's always the option of closing it as a duplicate of a thread with better answer.

¹ _{Excepting self-accepted answers, where there is no independent review at all, unlike non-self-accepted answers which involves two distinct users.}