In search of the perfect post, or the riddle of Habr

Before I got to Habr, my acquaintances told me that this is a resource for which it is very difficult to get an invitation, where serious people gathered, where they write serious things, and the community’s reaction is completely unpredictable: sometimes you can try for a long time, write more interesting and useful article, and it zaminusyut, and sometimes, will publish some nonsense, and it suddenly take "with a bang."

Now, when I’ve been at Habré for some time, I read quite a lot of articles, and even wrote some of my own, I want to share my impression of the mysterious and unpredictable Habré, how it looks from the outside, and about a question that many authors care about: “why posts are minus ?

It is clear that mostly habrayuzer positively relate to the articles, but nevertheless there are always disadvantages, although they are much smaller than the advantages. And this is understandable: it is impossible to write something that absolutely everyone will like, or rather, that there is not one who would NOT like it. And then I had a thought: were there any posts that do not have a single minus? That is, say, 200 pluses, and not a single minus?

I was suddenly introduced to this perfect post: something incredibly beautiful, causing a bright feeling, inexplicable, like the silent beauty of stars. How to find it, a post without a single minus?
')
Of those articles that I read, I could not recall such ones, at least 1-2 minuses, but they were surely. There was a logical idea that if there were such, then they should be looked for somewhere at the end of the list of the best posts for the month, because it was there, in theory, that there should have been posts where, with the greatest number of pluses, there were fewer minuses. After reviewing the last few pages (97-100), I found one such post.

Unfortunately the miracle did not happen, a good post, but nothing special. Then I thought that it is necessary to collect and analyze all the statistics. Perhaps among the posts with the least number of minuses there will be some regularities, for example a maximum, which will show me the cherished perfect post among the whole set written over the years of the Habr.

I downloaded all 100 pages of the “best of all time” posts, wrote a simple program for analysis and allocated for each post the number of “pluses” and “minuses”. Here's what happened:


These are graphically displayed along the X axis the thousand best posts, sorted by final grade (rating), shown in green, and in red - the number of minuses. It can be seen that the cons are quite nibble even the best articles. Perhaps another image will seem more vivid to someone, where the 500 best posts are shown larger, sorted by the number of pluses, so the minuses turned out to be “pressed into” the graphics:


Then I sorted the posts by the number of minuses. The picture turned out unexpected:

At first I did not understand what it was. Here, each prong is posts with the same number of minuses (on the left 1, 2, etc.), sorted inside by the number of pluses. By the way, in the end, there were only 2 posts without a single minus. The picture itself is interesting, but here I noticed that: although in theory the height of the teeth should have a normal or random distribution, it is clear that the first 10 are located in a strange double ladder:

How could it happen that the maximum rating of posts with the number of minuses 1-5 consistently grows, then a fall follows, and then from 6 to 10 again a steady increase? I have not found any explanation for this fact, except that some supernatural forces have tried.

If you look at the same statistics for the best for the month, we will see a normal distribution, without any anomalies:


After reading the articles located on the tops of these teeth, I thought: I wonder how the posts would be distributed, given the number of minuses in different ways? That is, for example, multiply them by 5, so that there are only posts at the top, where there are very few minuses. I conducted such an experiment: suppose that there is a negativity coefficient, and the rating with its account is calculated as

R = P - N * KN

where P is the number of pluses, N is the number of minuses, KN ​​is the negativity coefficient.

For example, consider several abstract posts:

- post with useful or interesting information, for which they put 130 pluses, but some still did not like something, and they put 4 minuses;
- post with congratulations (anniversary of a resource or invention), accepted positively (140 pluses), but since it is in general stupid, it has many minuses, say, 27;
- post on a controversial topic, for example, about copyright, which gains many votes, both positive and negative;
- useful, but very few people interesting post, for example, about the implementation of a specific algorithm, gaining only 27 pluses, while 8 people think that it was necessary to do everything wrong, and they put minuses.

With the usual rating formula, when the coefficient of negativity is KN = 1, the posts will be distributed as follows:

If CH = 5, then the posts, where there are many drawbacks, go down - the “copyright” goes down, giving way to the article about the algorithm:

We are also increasing the CN, and the birthday also goes down, so with an increase in the CN, both thematic posts are at the top, and the controversial remain in the tail:

Now we move the negativity coefficient engine to the other side, thereby reducing the weight of the minuses:

Here the birthday is immediately pulled out, which has collected a lot of advantages. And if we reduce the coefficient:

We will see that the copyright comes first, pushing everything else down. Thus, if desired, optimistic users could find posts with the greatest positive reaction, no matter how many minuses they gained.

I experimented a bit with this coefficient on real data and on statistics of the best posts, built graphs that began to look quite bizarre, as they now crossed the x axis, but it turned out to be not particularly interesting, so I decided to plot the posts in the rating plane. On the X axis - the pros, on the Y axis - the minuses:

Something in this picture is fascinating, as if sparks flying into space, I looked at it for a long time when I first saw it. This is the schedule for the best posts of all time, where it can be seen that they are cut off by the border of the worst rating (170) among the best posts of all time. And at the beginning of the article I put the schedule for the month, where only otzhabrennye are cut off. In both pictures we can clearly see what we all know already: a rare post on Habré reaches up to 500 pluses, and the majority are distributed on a plane almost evenly. Here I was finally convinced that there was no ideal post, and that attempts to find it should be left, otherwise the consequences could be unpredictable.

But what makes a habraiser to click on a plus or a minus? In fact, this idea came to me much earlier, even before I built the very first graph given here. Maybe I'm wrong, but here is what I got the impression:

Although this is a site for IT people who, as many believe, should think like machines, impartially evaluating only useful information, and sweeping aside everything else, it seems to me that people in the assessment of posts are guided not only by the mind, but also by their feelings. They use the buttons under the post for their intended purpose, as written in the hint: "like" or "dislike".

If, after reading, negative emotions have arisen, be it longing for a large number of letters, misunderstanding, resentment, or something else - this is a minus. If positive emotions - and no matter why, because the information can be useful, makes you think, or is just interesting, if it causes nostalgia, enthusiasm or any other positive feelings - this is a plus. Everyone tries to write an interesting article, it is impossible to guess if she will like it, but the author may try to imagine what people will feel when they read it. Today, publishing this article, I hope to find out whether it is so or not.