# Correlation between upvotes and effort

While I do not even have proper data for myself, my subjective impression is that the correlation is, while probably still positive, not strong at all. So, I sometimes get quite a few upvotes on answers that I had considered posting as comments, and sometimes see little to no reaction to answers I had put quite some effort into.

Am I alone with this? Are there confounders to this relationship?

• If true, probably because more users understand the simpler answers than the elaborated ones. But independent of votes, the latter are more valuable, in my opinion. – Andre Silva Jan 6 '16 at 12:03
• I think the correlation is negative. We live in the tl;dr era. – Cliff AB Jan 7 '16 at 20:26
• I believe you might find it illuminating to read all the comments to both (current) answers at stats.stackexchange.com/questions/35387. It is one of many instances I can recall in which I worked hard on an answer--and in this case it was scarcely upvoted; it received a downvote; and it was repeatedly attacked by a very reputable statistician (a Fellow of the ASA, no less). I will leave it to you to decide whether that answer is correct or sufficiently well justified, but it stands as a stark reminder that there is no guarantee that effort leads to upvotes. – whuber Jan 8 '16 at 16:03
• That seems very intereting. I shall try to have a closer look (but do not expect to be able to weigh in between the two of you anyhow as far as the concrete question is concerned). What the episode suggests to me that other users seem to have substantially more challenging criteria to what makes an answer worth upvoting than I do. Essentially, I (at least try to, and do my best to adhere to my goal to) upvote whenever I have the impression that the answer is likely to leave the OP better informed than he was prior to the answer. – Christoph Hanck Jan 8 '16 at 16:12
• @whuber I have read both answers in that thread, and you and Michael clearly have interpreted the question differently. I find it unfortunate that this led to some angry words in the comments, because the question is very ambiguous; I re-read it carefully but still can't say whose interpretation (if any!) is correct. The fact that your answer got accepted does indicate that you might have understood the Q correctly, but the wording of the Q itself is quite confusing. – amoeba Jan 8 '16 at 16:27
• @whuber I have edited the Q and left a comment there. Please do let me know if I misunderstood anything, as I would not like to bring additional confusion to that old thread. – amoeba Jan 8 '16 at 16:39
• @Amoeba I think you have identified an important criterion: ambiguous questions (and it's not always obvious at the outset that they are ambiguous!) often do not get much attention or love, nor do any of their answers. This is one more reason why it's so helpful to read questions critically when they appear, to point out possible ambiguities or alternative interpretations, and to ask for clarification. Some of my higher-voted posts in fact did all of this by themselves: they pointed out two or more interpretations and discussed them before offering any answer. – whuber Jan 8 '16 at 17:43
• @whuber You have too much respect for people flaunting their titles. – Did Jan 12 '16 at 7:41
• @Did I have respect for people, individually, and reasoned argument generally. – whuber Jan 12 '16 at 14:42
• @whuber Is this what I am talking about? In the present case (which I happen to know a little bit), to mention the ASA fellowship might be slightly misleading. – Did Jan 12 '16 at 14:45

I think there are several (confounding) factors here. The biggest, by far, is the number of views a thread gets. If no one views a given thread, there are obviously no opportunities for people to vote. Moreover, every view isn't necessarily another actual opportunity for people to vote anyway; many of the views may come from the same people (who may have already voted or decided not to).

The number of views a thread gets is largely a function of the topic. It is quite clear that some tags get consistently more views than others. The title will also make a big difference in whether someone navigates to the tread. The length / complexity of the question can make a difference as well: when someone lands on the thread and sees that it is going to take some work just to get through the question, they may decide it isn't worth their time (cf., this meta thread: Does length of a question influence probability of receiving an answer?).

The correlation between quality and votes isn't zero, however. There have been a number of answers that I put some effort into, only to receive few votes. But I have found that those are the answers that continue to accrue upvotes (albeit at a slow rate) over time. Whereas other answers may have gotten a few upvotes immediately, but then stay at that level forever.

Whether your posts are getting many upvotes or not, be assured that they contribute to the site and that we appreciate your efforts here.

I'd agree that there is at best a weak positive correlation.

But what is effort? It's not just the effort that went into an individual answer. The effort of following the forum over a few years needs to enter the accounting. So effort is obvious short-term effort $+$ whatever long-term effort is pertinent.

Over time on the forum, you learn at least subconsciously more about what the forum likes, as a collective set of preferences, and so learn a little about what works better. So, you get a little better at planning your effort.

Here is a very partial list, in no particular order, of what helps to get upvotes. (It should be no surprise in a statistical forum that the response of upvotes depends on several covariates.)

• Writing clearly and fluently, with good command of English and strong self-editing, but also fairly concisely.

• Graphs! For example, @Glen_b designs very clear, tasteful, customized graphs for many of his posts.

• Some rigour and professionalism in analysis, wide sense. Even at a low level small details such as careful use of terminology and notation and mathematical typesetting help greatly. At a higher level, outstanding answers may use whatever mathematics it takes to get to the root of things. Here @whuber is among the masters.

• Analyses of real data. Or simulations. In general, really good examples.

• Some minor wackiness. Anything that evokes a wry smile.

• Good lists of references.

• Already having a fair reputation. It shouldn't be the case, as every answer should be up-voted only on its own merits, but having a good reputation for high-quality answers and forum maintenance attracts attention to your answers.

• Multiple edits. Again, that in itself shouldn't be a reason for upvoting, but edits bump a thread to the top of the active list and draw attention thereby.

Here are some of the things that don't help much:

• Rants.

• Rather obvious riding of hobby-horses.

• Marginal or limited relevance to the question. This is difficult to manage. Sometimes a poor question can be steered gently but firmly in a different good direction that finds the pearl amid the grit, and that's all positive (except possibly for the poor OP). But anything looking like "That reminds me obliquely of this different thing" usually gets few upvotes (and occasionally gets flak), however good it might be as an answer to a different question.

• Being sloppy.

• Being wrong!

• Plenty of great points, too! – Christoph Hanck Jan 7 '16 at 20:18
• I agree with the edits point. I noticed getting a few points when fixing grammar of my most voted old answers. – Pere Sep 19 '16 at 15:56
• And I would add another point: The most people can understand and answer (and question), the most points it gets. Sometimes that favours answers to easy questions (for example, questions on basic probability) and sometimes it favours layman stile answers in front of rigorous journal of mathematics style very elaborate answers, although the later needs more effort. – Pere Sep 19 '16 at 15:58

I think another important factor to take into account is preferential attachment -- if something has been voted on a thousand times, omg, I have to see what that question is too! Meaning, especially if you want to get particular about "correlation," that there are important non-linear effects in popularity that do mess with a straightforward exchange of votes for effort.