15

Stack Overflow is maturing very nicely. As many people are saying in a recent thread there are numerous good reasons for us all to be here and we're making the most of it.

A few moments ago there were 1013 pages of users, which means somewhere in excess of 28 thousand registered users (although over half haven't actively contributed and are still on just a single point).

The nature of the reputation points is seriously inflationary, assuming that you shy away from making blatantly inflammatory statements and voting down lots of posts, there is no way to both actively contribute and keep your points tally down. In recent podcasts have hinted at the bounty system coming along soon which confronts in part this problem.

And I do think it's a problem. There are now more than 20 users with over 10,000 points. To me that's no longer meaningful. Yes some of them are very good in their fields, but all it shows is that they spend more time here than most! And what is the difference between a user with 5, 8, 12 or 20 thousand points?

Over time, points that just stack up will cease to be useful. Would this be the eponymous Stack Overflow?

Is reputation at that level meaningful? Is the fact that some of these leviathons have 100 bronze badges at all interesting? - or is it actually more interesting? The marginal pleasure of getting those early badges and tripping the various benchmarks to unlock functionality is long gone.

Do these huge totals mean a mature system or that the Gods will leave to find another game? What drives the 10k+ clique to continue?

5
  • 3
    reaching 20K :) i dont think anyone is doing it for the points.. it's just an added bonus. the driving factor is mostly learning and teaching/helping others.
    – Gishu
    Dec 11, 2008 at 9:23
  • I'm just in it for the points :)
    – TT
    Dec 11, 2008 at 10:21
  • 2
    I'm in it for the lulz. Dec 11, 2008 at 13:55
  • 1
    There's nothing wrong with being in it for the rep, that's a perfectly valid answer!
    – Unsliced
    Dec 11, 2008 at 14:06
  • Didn't we expect it to be a long tail distribution, just like wikipedia? Even there only about 1% are contributors/editors. Oct 22, 2009 at 18:25

17 Answers 17

33

I've wondered about "reputation inflation" too. Perhaps Joel and Jeff have a West Wing-esque secret plan to fight inflation. It's hard to know exactly what to do about it though. We could all drop 1% of reputation each day - but that's really just feeding the MMORPG "must keep playing" game.

The key question is: what is the intended purpose of reputation?

If it's meant to show that the system trusts you, then that's fine - we get more and more users that the system trusts, because they've earned that trust from the community. No problem.

If it's meant to show that a user has helped a bunch of folks, then that's also still true of high-rep users, even if it's mostly due to them being here a long time.

If it's meant to show that their answers are always correct, then obviously it fails. If you guess at answers and are right (and upvoted appropriately) a quarter of the time - a pretty poor record, really - then you'll still gain reputation on average.

If it's really just meant to keep people coming back to the site, then that's a bit sad.

As for whether high-rep users will leave - I was mostly active on NNTP newsgroups before I came to StackOverflow. There was no reputation system there beyond the normal one of the community getting to know you and trust you. That didn't make me leave... (and I still post there now). Why do I do that? Because I like helping people. I won't pretend that's purely altruistic - it gives me a buzz to know that I've helped people, and obviously the recognition aspect is nice too. I suspect the same is probably true for most other high-rep users. If we didn't like helping people, we wouldn't be here. (I don't think you'll find many high-rep users with as many questions as answers.)

4
  • 3
    I just have to comment on NNTP newsgroups. Do you visit HTTP websites, or send SMTP email? I'm actually not being a punk here; I assume your point was to differentiate between Usenet and other forms of discussion groups. I just find it amusing that the protocol was chosen as the way to do it.
    – Mikeage
    Dec 11, 2008 at 10:09
  • 2
    I didn't want to use "Usenet" as the moniker, as that sometimes implies just the "big 7" or whatever it's called - whereas I'm usually on the microsoft.* groups. At the same time, I wanted to indicate that it's not just web forums, which might also be called "newsgroups" by some.
    – Jon Skeet
    Dec 11, 2008 at 10:15
  • I think the emphasis should be more on the quality for the answer as opposed to the number of answers.
    – Jacco
    Dec 11, 2008 at 13:21
  • 1
    i find it amusing that you find JonSkeet amusing. Dude...he's Jon Skeet!
    – Epaga
    Dec 11, 2008 at 13:22
15

I have this request in UserVoice since a lot of time. No one voted. That would be great to have more "action" to do. Moderation is hard sometime (lack of privileges) and we could do a lot more to help this website with more access to some features.

What drive me? I like to help of course and I like to try to keep getting the 200 limits every day, This kind of motivate me to learn thing that I do not know. Some day it takes 1 hours, some time it takes to whole day depend of the question and how many works I have. Since stackoverflow I am less tending to waste time on Digg or other website that doesn't give me something (knowledge).

But, Stackoverflow shouldn't penalize higher reputation with "tax" or something to slow down them. What is the point to penalize very active contributor? They are a lot of room for improvement to add functionality for them. They could be able to change to "Community Wiki" question that should be as it. They should be able to have a higher close power that only these range can re-open and not people too low (example someone over 10k close a thread, someone below 10k cannot reopen... this will reduce war of open-close). They should be able to restore comments if deleted. All these new privileges should be at 8k, 10k, 15k, 20k, etc


Update: If you check the top 20 of 2 months ago and the same top 20 now, you can see a lot of people that aren't any more. Most of top user are now at the second page or near of it... and pretty inactive when before they were like Justin Standard and many other. A lot of super-active user became not active after reaching all objectives. Theses 2 graphics below are a lot common:


(source: clip2net.com)


(source: clip2net.com)

I think it's no a coincidence that people stop being active, they do not care anymore to get some reputation or help if they do not get something.

3
  • Excellent piece of stats-work - it was exactly the kind of conclusion that I was assuming to be true, but didn't have the evidence so easily to hand.
    – Unsliced
    Dec 11, 2008 at 21:35
  • Oops I thought I could edit your entry but I can't :P Here's mine: img40.imageshack.us/img40/6694/imagen11d.png Same story.
    – OscarRyz
    Aug 25, 2009 at 2:14
  • But I see this as a relief. Is not SO doesn't attract me anymore, It is just I'm not obsessed with it as before the 10k. I said to my self I'll get 10k and then use it in a sane way.
    – OscarRyz
    Aug 25, 2009 at 2:16
10

I think that average upvotes per question and answer might be a better metric. That way we can really see if someone is consistantly posting good answers/questions or if they simply do spend more time on here and have a billion +1 upvote answers.

From looking at the 10k+ crowd right now though, most of them do seem to have a really high upvote average on both questions and answers.

6
  • And even then, the whole system tends to be subjective. Just like anything in real life ;-).
    – Gamecat
    Dec 11, 2008 at 9:21
  • average sounds like a worth metric at least in addition
    – bananakata
    Dec 11, 2008 at 9:26
  • Expected value would work better than average methinks.
    – Edouard A.
    Dec 11, 2008 at 10:54
  • 2
    Even then, popular questions get more views and votes -- so answers in a "what's your favorite quote" thread (or this one) will get more votes on average than some outstanding answer on some obscure, but possibly more important, topic. So maybe divide by the total votes for all answers per question? Dec 11, 2008 at 11:08
  • @ShreevatsaR: That might work. Look at my answers with the most votes - they're certainly not the ones which actually add the most value.
    – Jon Skeet
    Dec 11, 2008 at 13:40
  • 2
    well, low answer upvote may be a sign of question low popularity. There should be some function for this average votes metric, to take in accounting question views, question upvote, question age, answer age, and was answer accepted or not.
    – coldice
    Feb 24, 2009 at 12:53
6

Inflation sucks. Neither monetary nor fiscal policy will help so here are a few things that could:

  • Weighting There could be some kind of weighting to make the reputation model less linear. So for example if you received a "right answer" to a "trivial" question you should receive less reputation than for a "non-trivial" question. Perhaps reputation could be be a function of views AND up/down votes AND favourites for a question etc.

  • Capping Limit the amount of reputation you can gain from a single question / answer. Because ONE super question / answer doesn't make you a guru (well not in my book).

  • Time linked deflation / depreciation Another slightly bolder idea might be to do some time related deflation to decrease "reputation" acquired from old questions as time progresses. It would also encourage users to be more active, and not rest on the back of their previous work :D and also has some real world significance which I will try to explain at a later date :D

At the moment it is something like this:

Users gain reputation when:

  • Your Question or Answer is voted Helpful: +10 reputation
  • Your Answer is marked Accepted: +15 reputation
  • You Marked an answer to your question as Accepted: +2 reputation
  • A down vote on your Question or Answers is removed: +2 reputation
  • You remove vote from a Question or Answer you have votes as Not Helpful: +1 reputation

Users lose reputation when:

  • Your Question or Answers is voted Not Helpful: -2 reputation
  • You voted a Question or Answer as Not Helpful: -1 reputation
  • You wrote the latest revision to a post flagged Offensive 5 times: -100 reputation
  • A user removes a vote on your Question or Answer that they previously voted as Helpful: -10 reputation
  • You have multiple user accounts merged. Reputation is recalculated during the merge so you lose reputation gained from questions that have since been deleted.

Additionally:

  • All users start with 1 reputation.
  • A user's reputation may not drop below 1.
  • A maximum of +200 reputation may be gained per day from voting. (A new day starts 0:00 UTC). Users who have reached the 200 rep limit in a certain day may still earn rep from having one of his or her answers accepted, canceling previous downvotes, and accepting other peoples answers to their questions.

source: https://stackoverflow.com/questions/303954/how-to-get-over-200-reputations-every-day

9
  • I like the 'weighted' suggestion, especiallys since you could go to any high-rep user's page and find at least a few questions or answers that garnered high rep because they were exceptionally 'soft' questions. Dec 11, 2008 at 13:24
  • I agree with the weighting suggestion in principle. However, who decides what questions are trivial? Take a look at the top questions by vote to see why you can't base it on views, votes, or favorites. Also, there's already a per-answer cap on reputation (I think it's 300). Dec 11, 2008 at 13:35
  • Bill: Another voting section next to the question. Vote on how 'technical' the question is. Dec 11, 2008 at 13:37
  • @Gortok;@Bill I was thinking about something like a technical vote too.
    – Ian G
    Dec 11, 2008 at 13:39
  • @Gortok: Interesting idea. I was thinking tags could be used somehow. Those can be too easily changed by one user, though, while votes on "technicality" would be decided by the community. Dec 11, 2008 at 13:54
  • @Bill: Are you sure there's a per-answer limit? This is the first I've heard of it. With the per-day cap as well, it can be hard to tell to be honest. Boo hiss for the time-based entirely-artificial cap :(
    – Jon Skeet
    Dec 11, 2008 at 14:33
  • @Jon: I think none of these should apply to you. Your reputation isn't high enough as far as I'm concerned :D
    – Ian G
    Dec 11, 2008 at 19:43
  • Perhaps the easiest way to "weight" answers is to reduce the reputation earned for non-technical questions (e.g. your favorite cartoon, or the one about "Jon Skeet Facts" - no offense Jon!).
    – Mark Brittingham
    Dec 13, 2008 at 3:27
  • That's a good idea. Who do we tell?
    – Ian G
    Dec 13, 2008 at 10:11
5

If you contribute only for the reputation, you will be bored if you are number one with a great margin (more than 2 k).

But if you are here to help building a great site. It is just a number. (And if it is 3k+, you can do anything that is needed).

5

The same as money - you just want to have more? :)

1
  • I want more money so I can buy a synthesizer.
    – Iraimbilanja
    Mar 19, 2009 at 19:24
4

What drives the 10k+ clique to continue?

Prestige

1
  • professional interest in the subjects they choose to be active within I guess
    – noesgard
    Dec 11, 2008 at 9:22
4

And what is the difference between a user with 5, 8, 12 or 20 thousand points?

They've likely helped more people?

3

At the moment, answering popular non-tech questions with a thoughtful answer seems to be a way to gain a lot of rep. That's cool, but it doesn't mean that the person actually has expertise in any field.

It would be good to see some kind of "expertise" reputation as well. I'm thinking that questions and answers could also be up or down voted in terms of technical difficulty or knowledge. That could be used to build a technical rep.

2
  • Answering popular questions is only a good way of getting upvote-rep, which is capped at 200 per day. After that, you need rep from accepted answers - which are much easier to get with technical in-depth questions about obscure topics (so no-one else answers!)
    – Jon Skeet
    Dec 11, 2008 at 17:26
  • The system has a lack about non-tech question and a lack of tool to close or switch to wiki these post. So, as Jon said, the way to be over these people who just get rep by non-tech is to get accepted answer. Dec 11, 2008 at 22:01
2

Tagging on to Simucal's answer, maybe two metrics. I know on eBay, raw reputation tells me someone has been around, which is a good thing to know. But the percentage is all I really look at before a sale, and having a lot of high upvote answers could show something significant. Better than a user who never answers anything anymore to avoid blemishing their perfect average.

1
  • Good point, I do the same when using ebay
    – mmcdole
    Dec 11, 2008 at 9:22
2

I certainly can't speak for the rest of the 10k club, but as for myself, I keep going in the spirit of teaching and learning, and trying to add new information to Stack Overflow.

I do keep an eye on my reputation, but only as a gauge of how my recent questions and answers are being received. It wouldn't really matter if it were 10,000 or 2,000 since it's really only the last few digits that tell me what I need to know.

2
  • 1
    The difference being that at 10K you don't get to see the last few digits :(
    – Jon Skeet
    Dec 11, 2008 at 13:38
  • I'm hoping that if I hit my user page 1,000 times per day someone will notice and give us back those digits. :) I really didn't think it was a big deal when they first changed that, but now I really want it changed back. Dec 11, 2008 at 13:50
2

What is reputation for? That seems to be a key question. A common complaint is that too much comes from too easy questions. Several people have suggested expertise or tag based reputation to highlight technical scores, that was certainly the solution I was toying with as I wrote the original question.

I've made a uservoice suggestion along these lines.

1

I think having a 'K' somewhere in the points means that either you are an expert in one or more disciplines (likely) or that you always have one window open and reserved for SO. Either way I do not think that it diminishes the idea behind this forum. And remember there's always 'M' and 'G' after 'K':)

1

It's too high when you're Donald Knuth.

Until then, keep at it.

1

As Jon, Bill, etc say - there is a lot of benefit etc from mutual learning. I first started contributing on usenet forums, and I learnt huge amounts from reading the replies on things new to me, and simply trying things... "how do I do x?" - dunno, but I can do y... is that any similar? etc.

It is also a good way (alongside blogs, etc) to keep up to date on things... if something new comes out, somebody will be asking about it.

1

I just wrote an answer to one of Jeff's questions that he deleted whilst I was writing the answer, but this is a pretty similar question.

I think the comparison to MMORPGs is appropriate. There needs to be a good leveling system. Rep could be like experience, and you need more and more of it to get to higher and higher levels.

Or you could have different stats, like helpfulness and accuracy. That way you tell the difference between someone with a low score who always gets the answer right, and someone with a high score who only gets it right 25% of the time.

1

For arguments sake: What would the negative impact be if reputation was hidden after you get to the ability-to-close-question threshold? It could be visible to that user, but not displayed else where.

Why? (Again, remember this is just for arguments sake) The system is misleading.. Given two answers, one from someone with x-thousand rep, one from someone with 2 rep - both answers can be equally good or bad, the rep-points of that user should be utterly irrelevant. In reality, people are more likely to accept or up-vote the answer from the higher-rep'd person, when really the goal should be to encourage new(er) users.

Not-displaying a users rep over a certain limit would, in theory, work just as well - the best answers get upvoted, the bad ones don't.

You must log in to answer this question.

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