# Why is it not possible to offer a bounty for longer than 7 days?

I had offered a bounty for the following question that expires in 7 hours:

What did Gelfand mean by suggesting to study "Heredity Principle" structures instead of categories?

The results have been very disappointing. No answer was added, no comment, no reaction at all. Which leads me to the conclusion that the limitation of 7 days may not be suitable for a bounty to achieve its goal.

Especially when a question may require some longer thought or when people don't check this site regularly.

I imagine the limitation of 7 days comes from the Stackoverflow.com site, whose activity level and status of professional importance are completely different from Mathoverflow.com. While the 7 days might be appropriate for the former, it may not be so for the latter.

I hope for some consideration towards increasing this limit. With the current 7 days limit (which I regard as a too low for the system to work as intended in some cases), I will not offer any bounty again and will not recommend it to anyone.

UPDATE. When writing my response below to the answer by Joseph Van Name, I have looked at the 6 bounty questions and wrote this answer that the OP found helpful and awarded the bounty, even though my argument was incomplete to solve the problem posed.

Did the system work as intended in this case?

On the surface yes, since it was the bounty through which I have discovered the question.

However, it was a huge coincidence that I had found it, as the bounty was about to expire within a day! Had I looked just 1 day later, I would not have discovered and answered this question, with the bounty of 250 points would have possibly lost. In contrast, chanced were much higher if the bounty stayed for somewhat more reasonable period, especially given the coming holidays.

Now this leads to the question, how many similarly interesting questions are still there unanswered and undiscovered during their short bounty period, precisely because it was so short?

And finally, analysing a bit deeper the effect of the bounty on that question, it is clear that even there it did not quite work as might have been intended. As my answer did not completely solve the problem, it would have been again more beneficial to have the bounty for a longer period, that would attract further contributions to complete the solution.

• Somewhat related older post: Is bounties period sufficient? It is not clear whether bounty period can be changed individually for a specific site: Are bounty parameters network-wide or per-site settings? – Martin Sleziak Dec 19 '17 at 12:51
• I really think the reason is not lack of interest in your post, but just that no one has thought of anything great to say! I for one found your question very interesting -- I had intersected with Gelfand when I was a graduate student at Rutgers and he had tried to get a number of us Rutgers people interested in quasi-determinants, and I even thought at one point they might be useful for some things I was researching. But it could be no one at MO has an answer (now or in the foreseeable future); that's the way it goes sometimes. – Todd Trimble Dec 19 '17 at 14:32
• But are you in contact with Retakh for instance? He might have interesting thoughts on the matter. – Todd Trimble Dec 19 '17 at 14:33
• It acts precisely as intended, so I don't think you can call it a bug. I think feature-request would be a better tag than bug. – Gerry Myerson Dec 19 '17 at 16:38
• @GerryMyerson The point I am trying to make here, and in my answer below, is that the limit of 7 days is way too short for this site. As opposed to SO, where it does indeed work as intended. – Dmitri Zaitsev Dec 21 '17 at 3:06
• I think I understand the point you are making, Dmitri; I just don't think I'd call it a bug. – Gerry Myerson Dec 21 '17 at 4:35
• @GerryMyerson The definition of a bug I'm using here, is something not achieving the intended effect. – Dmitri Zaitsev Dec 21 '17 at 5:29
• There's a difference between not achieving your intended effect, and not achieving the intended effect. But I think we're going to have to agree to disagree. – Gerry Myerson Dec 21 '17 at 7:01
• @GerryMyerson I'd be open to any convincing evidence it is achieving the intended effect for anyone else. I have seen others pointing the same, so it does not seem like I am alone. – Dmitri Zaitsev Dec 21 '17 at 7:47
• @GerryMyerson I have removed the 'bug' word, as it perhaps sounds controversial. It might work in some cases, but not in all, so I'd still feel the need to change it. – Dmitri Zaitsev Dec 22 '17 at 2:21
• Perhaps quid's (=user9072) in these old posts are also worth reading in connection with this: Bounty policies and Are bounty reputation points assigned to unanswered question returned back? – Martin Sleziak Dec 24 '17 at 10:33
• To me it seems that both here and in the previous discussion, it would actually be more suitable for the OP to have listed the question somewhere permanently rather than longer time period for a bounty. I have recently made a proposal along these lines: What would be good place to list unresolved bounties? – Martin Sleziak Dec 24 '17 at 14:18

I consider the 7 day bounty period to be quite reasonable and effective for several reasons.

1. While bounties cost a minimum of 50 pts, in reality the cost for a bounty for one's own question is much lower. If you put a bounty for a question, then that question will receive more upvotes than if it did not have a bounty in the first place. One should consider the cost of a bounty to be more like 25 pts. A 7 day bounty period for 25 pts is therefore a very good deal since a person who answers will still be rewarded 50 pts for a correct answer within the timeframe.

2. The purpose of a bounty is not simply to give people extra points for a correct answer if one gives the answer. If the purpose of a bounty were to simply reward a good answer, then a longer bounty period would be more understandable. However, the purpose of a bounty is in part to give the question more attention and to signal its importance. When I have given bounties in the past, I have considered the 50 pt investment into the bounty to be a very good investment even if the question has not been answered simply because the bounty gives the question more attention.

3. If you would like to offer an entity points when it gives a correct answer, then you may freely give your points for that answer as a bounty after the question has been answered.

4. A longer bounty period will simply mean that every individual bounty question will be given less attention for the time that it is a featured question since there will be many questions with bounties to choose from. This sort of bounty inflation will defeat the purpose of having a longer bounty period.

5. A longer bounty period will not guarantee that anyone will answer the question. If a question has been unanswered for 7 days, then it is unlikely that such a question will be answered within a month. Some questions are not answered because they are difficult.

6. In the case that a question goes unanswered after the 7 day bounty period, one may put up another bounty to that question. The 7 day limit is only for those who do not have enough points to pay for multiple bounties. If multiple people consider a question to be important, then those multiple people can each offer a bounty on that question.

Since the Hereditary Principle question has 38 upvotes and over 1000 views, it seems like people are in fact interested in such a question. They simply have not come up with good answers.

• I think it is optimistic to think that the expected change in the number of upvotes after a bounty is "+5". Quite a few of bounty questions tend to be nasty and technical IMHO; if they also happen to be on a not-so-well-represented topic (on MO, I mean), not a lot of people are really going to upvote them. Of course, maybe you had a different experience with them; or maybe you have some statistics behind you. – user137767 Apr 11 at 19:08
• @StepanBanach Here is a SEDE query with bounty statistics which I took from this answer. (But run on MO rather than on Mathematics.) If I read the results correctly, average score of a question is 3.57, for question with a bounty it increases to 5.99. – Martin Sleziak Apr 12 at 7:50

I would like to address the points in the extensive answer by Joseph Van Name (EDIT. unfortunately deleted by the author), that would be too lengthy to put in comments:

1. In case of my question, all the up-votes occurred before the bounty was assigned. Then, not a single up-vote was added as the result of the bounty. This disproves the suggested effect that the up-votes attracted would compensate for the bounty. What this also likely means, the bounty did not attract any further attention. In my interpretation, the reason is that active users have already seen the questions, whereas a bounty of 7 days does little to attract attention of less regular users, who could provide additional valuable insight.

2. I would assume the ultimate purpose of the bounty should be to encourage quality answers. If "attracting attention" only refers to further up-votes and views without any tangible activity, I would question its usefulness, and would not feel compelled to ever offer a bounty again. In my view, the prime goal should be valuable activity, whereas attracting attention is merely a byproduct effect. It is the former that we (I) want to reward, not the latter. Especially if it will take time for the person to write a good answer, I would like to reward it.

3. I did not know that it was possible to assign reward after the answer. However, this way I would still lose valuable inputs by busy people, who might have been more compelled to take the time if the reward was declared beforehand. That way my bounty still fails to achieve the positive effect I am aiming at.

4. At the time of writing, I can see 6(!) featured questions in total on the Mathoverflow site! For the whole mathematics! With none of them even close to my area, just to put it in a context. I would hardly call it "too many questions with bounties to choose from". In comparison, the Stackoverflow site has 366 bounty questions! That I can refine by selecting the database tag, to get only 2, which explains how the 7 days over there makes perfect sense. Whereas the 6 questions in total on Mathoverflow, also explains how the 7 days there is way too short. In fact, working with the ratios, we arrive at $$7 \ {\rm days} \times 366/6 = 427 \ {\rm days},$$ to reach the proportion of the SO.

5. A longer bounty period will very likely increase the chance of an answer, because the question will get more likely noticed by more busy people who only come here sporadically. In contrast, the 7 days period will be of little effect for anyone in this category.

6. Being myself not a regular user, I probably fall in the category of the people who "do not have enough points for multiple bounties". However, more fundamentally, when I see how little effect the bounty achieves with the current system, I don't feel encouraged to keep offering bounties, even if I had more than enough points. Concerning bounties by other people, none had been offered that far for this question, despite the obvious interest judged by the number of votes. Which might yet be another sign that the current system does not really work (i.e. is a bug).

• In fact I am rarely looking at the featured tab -- there are just too few bounty questions to make it likely to find one close to my area. – Stefan Kohl Dec 20 '17 at 21:45
• And I have deleted my answer. – Joseph Van Name Dec 24 '17 at 4:29
• @JosephVanName Why? – Dmitri Zaitsev Dec 24 '17 at 10:45
• I do not want to talk about why I deleted my answer. – Joseph Van Name Dec 24 '17 at 13:51
• I have undeleted my answer. The smallest minority on Earth is the individual. – Joseph Van Name Apr 6 at 0:01
• I know this is a useless comment but I had similar experience with bounties in the past. Either my questions were boring, or the topic is not interesting to anyone. – user137767 Apr 11 at 19:12