What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 133 Stack Exchange communities.

In order to avoid the system being overflown, a limit of one bounty per user at a time was imposed when the new bounty system was introduced, with the plan of looking at this again when some time has passed.

I haven't kept totally close tabs, but I've never seen the overall number of bounties rise above 170 in the past two months. The old average was about 100-150 bounty questions at a time.

In light of this, would relaxation of the rules, as promised, be conceivable? At least to two questions at a time, if not a bit more? If not right now, then at some point in September April?

share|improve this question
1  
Oh! I'm finally going to upvote this now that I changed my mind. Oops. :) –  Jon Seigel Jan 14 '11 at 2:18
    
Re your edit: if you see SEDE is loaded with March data and I haven't updated my graph, please comment ping me. –  Jon Seigel Mar 14 '11 at 22:26

2 Answers 2

I think we can consider running a controlled experiment to raise the limit to 2 simultaneously.

To bring some data to the discussion, I created a Data Explorer query to show the number of bounties set over time by month. I dumped the results into Excel, hit the graph button, and this is what I got:

yee-haw

Clearly, the number of bounties being set is generally increasing over time, but that is mostly due to increasing site activity (more people = more people using the bounty system). The January 2009 data is because the bounty system was only introduced to the site at that point; I'm not sure what happened in February/March 2010. It seems there's an upward trend after the new system was released, possibly with an "initial excitement" phase which will then level off a bit.

I don't think the system is escaping with nonlinear growth, so we can definitely consider increasing the maximum number of bounties to 2, even experimentally (say, 2-3 months). It would be interesting to see how many people actually take advantage of the feature.

At the same time, we have to realize that the number of bounties is increasing, which makes setting a bounty less "special" and attention-grabbing for those questions. That's why I think a controlled experiment may be in order. Or, perhaps Stack Overflow, with its "big-city" problems as Jeff calls them, may very well be the exception for increasing the limit -- it would certainly be safe to raise the limit on the other sites. With Stack Overflow, The Management may want to be more careful going forward.


EDIT: Added the total number of registered users over time. Here is the source query.
EDIT: Added the number of bounties set per 100,000 users. Sorry for the weird units, I wanted to get all the curves on the same graph. It's the trend that's important anyway. Also, I corrected the red line -- it should have been before the June 2010 data point since that point is the total over June. What I said above still holds, though, as the June 2010 data is about 2/3 old system, 1/3 new system.
EDIT: Updated with August 2010 data. It appears there is an upward trend in bounties being set, and it's slightly outpacing the rate of new users. We may need more than 1 additional data point to draw any kind of conclusion.
EDIT: Updated with September 2010 data. Rewrote the body of the post to give more current analysis.
EDIT: Updated with October 2010 data, including the trendline I added (but didn't post) last month. Revised the data analysis.
EDIT: Updated with Nov 2010, Dec 2010 and Jan 2011 data (man, I got lazy!). Astute users will notice the X-axis scale was corrected. The data points are now positioned such that they represent the previous month of data (i.e., the April 1, 2010 data point represents data between March 1-31). I think it makes more sense now.

share|improve this answer
    
@Pekka: Thanks! I added the total number of registered users to the graph for comparison. –  Jon Seigel Aug 22 '10 at 17:43
1  
Can you do bounties per user as well to see if that's increasing? –  ChrisF Aug 22 '10 at 17:52
    
@Chris: Okay. I need to correct the red line anyway -- I realized it's actually a month too far to the right. –  Jon Seigel Aug 22 '10 at 18:33
    
@Chris: Completed. I hope that's what you asked for. –  Jon Seigel Aug 22 '10 at 18:50
1  
Maybe the bounties/user rate isn't that informative and it should be bounties/activeUser instead (where active could e.g. mean at least one question/answer in that month)? I'd guess there are quite many "dead" accounts. –  Georg Fritzsche Aug 22 '10 at 19:36
    
@Georg: IIRC, there are mechanisms in place to purge old user accounts (with some set of criteria I can't remember right now) -- those accounts are deleted from the data automatically. So essentially, these query results take that into account already, according to the system's definition of "inactive user." –  Jon Seigel Aug 22 '10 at 19:47
    
@Jon It also stands to reason that the size of bounties plays a role... If you are willing to invest some more time in this :) –  Pëkka Aug 22 '10 at 19:48
    
@Pekka: I'd love to include something like the average bounty size, however, if a bounty expires with no award, the bounty size is not currently available in the data dump (ugh). I'm not sure grabbing those stats with the current data would be valid. (Also, in the new system, there is no extra +50 added to the bounty, which would skew the curve.) –  Jon Seigel Aug 22 '10 at 19:58
    
@Jon I see! That makes it difficult.... –  Pëkka Aug 22 '10 at 20:02
    
@Jon: Hm, but wasn't that something like >= 6 months? –  Georg Fritzsche Aug 22 '10 at 20:31
    
@Georg: Yes, something like that. Note that the query results for total users actually starts 6 months before the graph starts, so according to the site definition of an inactive user, the whole graph takes this into account. I also don't see the point of defining our own criteria for what qualifies as an active user, as it would always be under dispute -- might as well just use the same definition as the site itself. –  Jon Seigel Aug 22 '10 at 20:46
    
@Jon - yes that's exactly what I wanted. It shows that while the total number of bounties is increasing, it's not increasing as rapidly as the number of users. –  ChrisF Aug 22 '10 at 21:21
    
@ChrisF: you should probably count the number of users with more than 200 reputation who logged in the last month to get a more accurate picture... Many people register and never come back. –  Andreas Bonini Aug 23 '10 at 10:37
1  
@JonSe badge++; –  badp Sep 12 '10 at 18:41
    
@Jon great stuff, thank you! –  Pëkka Oct 16 '10 at 15:38

We now allow 3 bounties per user. This has cause a slight increase in the number of active bounties on Stack Overflow (which is pushing 300 at the moment)

share|improve this answer
2  
@Pekka: Can you accept this answer please, so that it rises to the top? Thanks. –  Robert Harvey Sep 21 '11 at 2:09

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .