I haven't been here very long, but it seems to me that I frequently see people with lots of badges and hundreds, if not thousands, of reputation points asking pretty simple questions. I don't want to point any fingers at anyone, but I would like to know what the point of reputation is if you can't really trust it to evaluate the reliability of the poster's abilities.

That isn't to say that I haven't been learning things here or that I'm unhappy with the responses, just that my experience is that there doesn't seem to be a correlation between reputation and ability -- mostly just a correlation between reputation and longevity (on the site).

Your thoughts? Ideas on how to fix it (if it's broken)?


The general consensus seems to be that the SO reputation system is not broken, but that you need to understand that reputation on SO is not the same as programming reputation. Answers should be evaluated on content not on the basis of the answerer's reputation. Reputation on SO is an extremely effective way to increase SO participation but not necessarily an accurate predictor of ability.

migrated from stackoverflow.com Aug 5 '09 at 2:55

This question came from our site for professional and enthusiast programmers.

  • "[P]retty simply questions"? I get the point you're making but obvious grammatical gaffes of this sort detract from what you're saying a bit especially when you're commenting on the split between SO Rep and true competence. – Onorio Catenacci Oct 9 '08 at 12:26
  • It's a lot easier to catch grammatical errors in someone else's writing because your brain isn't filtering what you are seeing through what you are thinking. Fixed now. – tvanfosson Oct 9 '08 at 16:09
  • @tvanfosson: see--I was wrong. :-) A minor grammatical error didn't detract from the validity of your observation. Likewise some bad answers getting voted up doesn't detract from the overall utility of the system. I believe most of the answers that are voted up are sound. – Onorio Catenacci Oct 9 '08 at 16:37
  • Simple questions of competent programmers are not bad; if a good question is asked it is nice to have in the preexisting responses. Rep also encourages good answers, or answers that fill other areas. I suggest you down vote and comment on incorrect answers, up voting the correct one of course. – he_the_great Dec 4 '08 at 5:44
  • 5
    Downvoting and commenting seems to be a recipe for getting downvoted yourself -- perhaps even on an unrelated question. People don't seem to be able to deal with the downvote very well. I struggle with it myself at times -- too competitive, I guess. – tvanfosson Dec 9 '08 at 22:21
  • Old question recently linked to but worth commenting on: I've done a lot of downvoting and left comments and I can't think of one time that I've been downvoted that I didn't clearly deserve except as a result of non malicious confusion. – aaronasterling Nov 18 '10 at 6:46

25 Answers 25

up vote 85 down vote accepted

Reputation here does not indicate programming reputation. It indicates site participation reputation. Look at what increased rep gives you. It does not give you the ability to answer harder questions. It gives more control over the site.

  • I can accept that but the FAQ seems to indicate that it is both "how much you are trusted" and "how much you know." – tvanfosson Oct 9 '08 at 3:21
  • 2
    I have seen questions where the up voted answers were wrong. I don't me subjectively wrong, I mean code gives wrong answer. The two correct implementations had a total of one vote. Points is a way to keep people interested and reward participation. It has no relation to technical ability. – Peter Oct 9 '08 at 5:51
  • But then why is it that more editing power is given based upon reputation? High rep equals editing and closing rights. If rep is not equal to trust or correctness, then why give them more editing rights? – dacracot Oct 29 '08 at 15:11
  • 4
    Because editing does not require technical ability. I can look at a question without knowing anything about the subject matter and still make reasonable editing decisions. – EBGreen Oct 29 '08 at 15:53
  • I call BS... you think you can edit my question that you don't understand? I'm astonished. You should be the editor of a scientific or medical magazine! You got skills. – dacracot Oct 29 '08 at 22:05
  • 2
    If I see a teh I can easily decide it should be a the from context. No edit should ever change the intent of someones question or answer. – EBGreen Oct 30 '08 at 4:52
  • @tvanfosson: you're not reading between the lines. Reputation is not based on what you really know, nor does it matter if your answer is correct. It is based entirely on whether other people agree with your answer. – NotMe Dec 26 '08 at 20:08
  • 1
    @EBGreen: You got a lot of rep for this, you must be a great programmer! ;-) – Kredns Apr 20 '09 at 23:24

/me wins

Really it's just a cunning ploy to get people to remain active in the site, rather than growing bored of it after a few days.

I have to say, it works staggeringly well

  • 1
    I agree. I'm beginning to check my reputation before I eat breakfast in the morning... – JesperE Oct 9 '08 at 6:42
  • Notice the answer I chose to my similar question... stackoverflow.com/questions/234585/… – dacracot Oct 29 '08 at 15:13

Reputation is doing exactly what it's designed to do...it's an arbitrary number that has no meaning but it's that little bit of competition between users that make it so that we want to come back and see if we got more rep. And to gain rep we ask and answer questions thus making SO a site that people use.

Reputation is a measure of how well you help the community rather than a measure of how much you know. People who help, and generally don't post silly answers will at the worst not get down voted, while those who post reasonable answers, even if they aren't the best, may get a few upvotes here and there, and eventually get lots of reputation.

The idea behind reputation is to give people who can be trusted the ability to moderate the site. There's been talk on Uservoice of not showing reputation on Q/A's, for all sorts of reasons, this is probably one of them.

I know personally, I've posted my share of 'silly' questions. This doesn't make them 'bad' questions, just questions that are aimed at newer users. And I've learnt alot from them.

On the flip side, I've been able to answer questions and help out other people here and there too. I've managed to get a decent amount of reputation, but if you look at my Q/A count, I think it only averages out to about 2 upvotes per Q/A. I'm no elite coder, I'm only a student.

Reputation was never designed (and never could be) to indicate skill-level. It's simple a general indicater of how much useful input they have had around the site. Whether that's a little bit that helped alot, or a lot that helped a little shouldn't matter. It's only affects how much we can do on the site. It shouldn't be used as an indicator of how much someone knows. People with low reputation post great answers sometime, and people with high reputation make me smack my head into my desk. But we're all here to learn.

Peace, Stackers.

Well as someone who was here relatively early, I can definitely attest to the fact high rep can bring out a dictatorial streak in a wee small few. Arguments about semantics have played out while I have been on a thread; a post gets closed, the post gets opened again, yet another comes along and closes it again; meanwhile I and others sit back a watch the show.

Some of the questions during the StackOverflow beta would now be closed in minutes, and early adopters whom asked those particular questions accumulated alot of rep. With that being said, the rep gained was shared amongst a diverse range of individuals, all with differing, yet ultimately complementary demeanours.

Personally I would like to get to 3000 and post community wiki posts from that point onward; just so I can't be bullied, nor allow others to be. Yet, some seem to think the game is on, and someone has to win ...

A strange game perhaps. The only way to win is not to play along . . .

How about a nice Game of Chess ?

  • "The only way to win is not to play" Ahh, that brings back memories! – some Nov 15 '08 at 20:35

A high reputation in MySQL doesn't necessarily mean you're a C# expert.

I would imagine that questions that appear simple to one person look incredibly difficult to others.

Plus, since Jeff and Joel want this to become a repository for programming answers, perhaps those who are most well qualified to answer questions are asking them so that they can be found in the future.

  • I think what strikes me is not the inability to understand the nuances of particular platforms, languages, etc. but some wide discrepancies between reputation and understanding of fundamental principles -- bits and bytes level things -- not how do I do X in Y since I usually work in Z. – tvanfosson Oct 9 '08 at 3:17

I think there are some very compelling answers as to what reputation means. I wanted to also answer the first part of your question and say that when I'm looking at answers and deciding whether or not to up/down-vote them, I don't even look at a user's reputation. It's only after voting that I notice.

The sad thing is that if somebody answers immediately and it's an OK answer, it will probably get up-voted, possibly drowning out the really good answers that take more time to craft. My hope is that this will even out over time, when people come across questions on google and (hopefully) sign in to vote. This is why whenever I ask a question and new answers show up after some time, I really take the time to read them and see if they're better than previous answers.

Like @EBGreen said, reputation is just a measure of user participation. If user has high rep we can assume that he plays by rules and helping community. That's why high rep users have some control over the site.

Unfortunately voting\reputation system is abused by subjective, off-topic questions. IMO polls like "What is you fav X" should not give any reputation points to thread participants.

Also voting system is badly balanced - i.e. 1 up vote = 4 down votes. IMO it's wrong. Up/down votes should be treated equally.

I think that good idea is to measure user's rep per tag. For example if I answered (i.e. my answer was accepted) N times in X category (tag) than I can have "X specialist" badge. Another nice feature is to see level of participation of users, i.e. it could be a page with following charts:

  • N top users by # of accepted answers
  • N top users by # of asked questions
  • N top users by total number of posts
  • etc.
  • did you put this suggestion on uservoice ? – Ilya Oct 9 '08 at 6:45
  • someone already made a request on uservoice. too lazy to search for a link right now. – aku Oct 9 '08 at 11:32
  • IMHO giving something for the number of posts are wrong as it makes some people posting things with no value just to increase their post-count. – some Nov 15 '08 at 20:49

"The reward is reputation." --Marc Andreessen

The success of Stack Overflow demonstrates to me that Jeff Atwood made a great design decision to reward participation with reputation. I thought he and others might find the following quote from "The World is Flat" by Thomas L. Friedman interesting.

The Apache Web server had its roots in this form of open-sourcing. When I asked a friend of mine, Mike Arguello, an IT systems architect, to explain to me why people share knowledge or work in this way, he said, "IT people tend to be very bright people and they want everybody to know just how brilliant they are." Marc Andreessen, who invented the Mosaic Web browser, agreed: "Open Source is nothing more than peer-reviewed science. Sometimes people contribute to these things because they make science, and they discover things, and the reward is reputation. . . the peer review part is critical."

(Not that I'm particularly brilliant!)

More context for this quote is available from http://books.google.com if you search for 'Andreesen "reward is reputation"'.

Reputation counter encourages participation, also, but not so much, encourages quality. For a newcomer, reputation may be a motivator to get hyperactive on stackoverflow. But for quality we have the peer recommendation which has contribution to the reputation as well.

reputation is like the points in an arcade game. badges are like the power-ups. both provide a little reward for participation and keep you coming back for more

they have nothing to do - necessarily - with programming skill. Especially considering the incredible popularity and number of fuzzy favorite-candy-bar-for-crunch-time questions ;-)

since reputation gain is capped at 200 points per day, the rep score does place a lower limit on how long a poster has been a member of SO. For instance, at present, anyone with a rep score over 5000 must have been a private beta tester.

I think that the intent behind reputation has more to do with how helpful the member is to the community, whether it be by providing concise, timely, and correct answers or by asking questions well that a lot of people have found helpful.

Unfortunately, in practice it means how popular that person's questions and answers have been in aggregate, including fluffy subjective questions with no technical merit as well as solid technical and well-informed answers.

I think there should be a distinction between 10 answers with one vote, and 1 answer with 10 votes...

in my mind, the 10 vote one is more likely to be a "good" post.

but then again, it also could just mean its a popular thread.

stuff like this is hard to normalize against.

Perhaps it would be nice to have the ability to see reputation per tag?

Reputation (and badges) are due only to other users that have approved of your questions and answers. This does not really relate to the difficulty of the answers you give or questions you ask. So yes, longevity will always help, but the reputation is really a gauge of how involved you are in the site.

Anyways there is no such thing as a dumb question some might say.

I have noticed that poor and negative answers are getting voted down a lot.

Early good answers seem to get voted up.

I haven't noticed any other tends yet, but if jeff could do some statistical analysis of the answers asking things like

  1. how long was the answer?
  2. how quickly was it answered?
  3. does the answer contain code
  4. what language is the code
  5. does it contain a formula

I suspect the results would be very interesting. It would reveal how people vote and therefore what type of answer people want most.

Reputation is also an indication of how cleverly some people manage to ask open questions, that seem to get major voting, like this one;-) (Perhaps I'm jealous cos I do not have time to think up such questions.)

Perhaps this phenomenon will die down after the beta period, once there are sufficient "closed" technical questions. I believe that there has already been some discussion in this area (although I can't find them with the search tool).

  • To be fair, this question was asked when I was fairly new to the site. If I had understood how things worked, I would have originally opened this as community wiki so I wouldn't have garnered any rep from it. – tvanfosson Dec 15 '08 at 17:22

Don't use someones reputation as a guide as to whether the answer is right or wrong. Use the communities up / down count as a guide. That's the whole point of the site. Good answers will float to the top.

If you're looking at reputation it means you're doing something wrong. The reputation is for more advanced areas of the site. Which logically you would want to protect from someone who has only just signed up (e.g. closing threads).

I find it depends on the subject tags. For c++, I find that reputation has so far correlated to quality of answer, at least in my unscientific observations so far :) Your mileage may vary with other subjects, not sure why that is and don't care to speculate :)

Could a more effective algorithm be developed to calculate a user's reputation? Right now, it is easy to correlate longevity with reputation but it's hard to correlate quality answers with reputation. I wonder if there would be a more effective calculation therefore...

Currently each up-vote per answer equates to 10 points...so 10 questions, 1 vote each carries the same reputation as 1 question with 10 up-votes, even though probabilistically (is that a word) speaking, the latter would carry more weight.

Maybe if votes carried a sliding scale for example: The first vote on a question is worth 1 point, the second is worth 2 points, third worth three etc, then someone having 20 votes on 1 answer is likely to have a far greater reputation than someone having 20 questions each garnering a single vote... 210 versus 20.

I think this would probably make people think a lot harder about writing quality answers too.

This wouldn't completely combat the longevity vs. quality issue, but it would certainly make the gap between the two camps considerably more difficult to cross.

Just a thought...

I think, the more the ability of a poster, the less he or she feel uncomfortable with asking simple questions. Programming nowadays includes constant learning, there's nothing wrong in learning simple things even if you're already good at something extremely complicated.

One way to fix the issue (if there is one) would be to add more factors into the calculation.

Time would be a key one. Reputation in real life diminishes over time.

Negative reputation should have more influence than positive.

Also trading of reputation points. If someone is a master you might give them 10 points, its a way of recognising someone as being talented.

  • One of the unique things about how reputation is calculated in this system is that upvotes are +10 and downvotes are -2. The system, by design, weighs a positive contribution more heavily than a negative contribution. – Kyle Cronin Oct 9 '08 at 12:45
  • What about connecting with Ohloh.net and grabbing some rep from there? @nobody, don't like how low that subtraction is though, I get docked half of that. – he_the_great Dec 4 '08 at 5:48

Reputation reminds me of an incentive, which is considered not so good by some ;) http://www.inc.com/magazine/20081001/how-hard-could-it-be-sins-of-commissions.html

Reputation tells me that someone has posted a lot of helpful answers ... the best indicator is to see if someone is highly active in your tag AND has a high reputation.

Reputation, alone is just a neat way for SO to say thanks to people who participate. Badges are another matter, looking how active someone is in a given tag brings it all together.

I'm highly active in Linux .. so if I answer a .NET question, someone with a reputation of 1 may (and probably does) know MUCH better than I do.

You must log in to answer this question.

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