8
$\begingroup$

This question is driven by sheer curiosity - maybe even vanity?

My reputation curve acquired curious shape lately:

enter image description here

I can even guess reasons (mostly personal) why the pace is so slow ever after June 2019.

I thought - how typical might such dynamics be? What of it is really personal and what reflects some general patterns? Maybe many users experience this sort of slowdown after a period of high activity here? May I expect return of an epoch of more energetic jumps?

Of course I might simply ask those of you who would like to do it to post your own curves, but this is obviously too much to ask anybody.

Is there a way to somehow generate the "typical reputation curve"?

$\endgroup$
15
  • 2
    $\begingroup$ It would be helpful to be able to generate such a curve in higher resolution and for a longer period, but unfortunately I don't know how to do it. In my case the curve I can see in the same period is quite close to being affine. $\endgroup$
    – YCor
    May 22, 2020 at 11:57
  • $\begingroup$ @YCor There's supposed to be the (mostly hidden) https://stackexchangewebsite.com/reputation page, with a dated list of one's own reputation changes, but I'm not sure how to visualize it the most easily. $\endgroup$ May 22, 2020 at 12:39
  • 8
    $\begingroup$ since old questions and answers can still acquire reputation, presumably at a constant rate, I would argue that the "typical reputation curve" for an inactive user is a linear increase, and activity is signalled by a superlinear increase. $\endgroup$ May 22, 2020 at 13:53
  • 7
    $\begingroup$ @YCor You can see similar reputation curves for an arbitrary user by clicking on Network profile on the user page and then selecting the reputation tab. For instance yours is on stackexchange.com/users/1637952/ycor?tab=reputation . You can click on the name of a site in the legend to remove it from the plot (and get a better-fitting scale). $\endgroup$ May 24, 2020 at 17:36
  • $\begingroup$ @FedericoPoloni Wow! $\endgroup$ May 24, 2020 at 18:19
  • $\begingroup$ @mypronounismonicareinstate, that link stackexchangewebsite.com/reputation doesn't seem to work, at least for me …. $\endgroup$
    – LSpice
    May 24, 2020 at 18:53
  • 3
    $\begingroup$ @LSpice: replace "stackexchangewebsite.com" with "mathoverflow.net" (or any other SE site). :-) $\endgroup$ May 25, 2020 at 0:34
  • 21
    $\begingroup$ You're just doing your part to flatten the curve. $\endgroup$ May 25, 2020 at 13:04
  • 2
    $\begingroup$ Stackexchange makes reputation, user, post, and vote data available for download. If I am reading their data model correctly, you could build a reasonable reconstruction of each user's reputation curve, then take the average. data.stackexchange.com/mathoverflow/queries $\endgroup$
    – Neal
    May 27, 2020 at 12:30
  • 2
    $\begingroup$ @SamHopkins Judging by upvotes, obviously yours is a clever remark, but I am too stupid for it. Could you please explain to the poor guy here what do you mean? $\endgroup$ May 29, 2020 at 8:02
  • 1
    $\begingroup$ @Neal Going along with this, you would want to consider whether to include/exclude inactive users (those who haven't done anything to gain/lose reputation for n years). If you include those, I suspect the curve will be mostly flat due to the large number of zero/one time question askers/answerers on the site. For example, there are ~110,000 users on the site, but the all time reputation league only tracks ~7500. $\endgroup$
    – Tyberius
    May 29, 2020 at 21:21
  • 2
    $\begingroup$ @მამუკაჯიბლაძე: "flattening the curve" is an expression that's been used lately to describe strategies (social distancing, etc.) to combat COVID-19. $\endgroup$ May 30, 2020 at 15:39
  • $\begingroup$ @SamHopkins Oh my. I see, thanks :) $\endgroup$ May 30, 2020 at 17:10
  • $\begingroup$ The study of the average reputation curve could be split out into exogenous and endogenous factors. Exogenous factors might include answers per day, and what fields a user usually answers in. For endogenous factors, one might try to predict the number of upvotes each answer receives in terms of: how many other answers on the question, number of upvotes on the question, number of views on the question, the question's tags, the user's reputation, the question asker's question... Put these together and you have an idea how a user's reputation will grow given their interests and habits. $\endgroup$
    – Neal
    Jun 2, 2020 at 14:30
  • 5
    $\begingroup$ @YCor like this? data.stackexchange.com/mathoverflow/query/1170724/… It's not 100% accurate since reputation lost by downvotes given to answer isn't accounted for, but it comes reasonably close (it does account for the daily reputation cap of 200, for instance). $\endgroup$
    – Glorfindel
    Jun 3, 2020 at 20:48

0

You must log in to answer this question.

Browse other questions tagged .