Vote with your heart
Hello?
The fact is that not so long ago I read an interesting post about ranking in the Nudnik.ru blog. And then I read the comments to him. And if you click on "read more", then suddenly you will see this post, corrected and supplemented.
—————
The main (and only) postulate of the first post (this is the second, explanatory) was:
Since almost no one understood this ingenious hypothesis, this post was written.
')
But first - what is the point of voting in general? First, of course —- ... But someone has to say it all the same. Voting is a choice. Five presidents were set before you and a radiobutton was hung under each. Choose one.
It is clear that voting for Zhirik and Zyugu at the same time is impossible. And really want to.
Another thing is when you have a lot of material, and each creature can be put down a pair from 1 to 5. The final goal for voting is “on the other side”, that is, for the site — the same choice, or rather — selection, sorting out shit and making more or less adequate tops.
The purpose of voting for the user is completely different, because there is no choice. The user can put 5 points to all evaluated entities. Full analogue "I want everyone to be president at the same time." It is clear that these voices are equal to zero, and in its pure form this almost does not occur, but there is an opportunity to vote this way.
There is no limited resource (“choose one”), therefore there are no motives to spend it economically (or reasonably) either. But at least some motives are needed, which is why emotional mechanisms are involved, which are also binary: “liked / disliked”.
There is no sense to distinguish between “liked, but not very much”, that is, to put the “four” instead of the “five”: “do you feel sorry?”. No, I do not mind. We are good! But if we are such a kind, angry ...
From the point of view of voting, as a choice, the discreteness of the scale matters: the larger the scale, the higher the accuracy. The most suitable scale is the 100 point scale (it should be understood that voting, such as 7.5, is also a 100 point scale, which is also understandable for the user in a ten-point scale).
From the point of view of people and especially social interaction, the binary scale is the most convenient: shitty / candy.
And 75–95% of people will reduce any scale to binary.
Hence, an intermediate conclusion: no matter how you call it, it will still be used in exactly one way, as soon as you understand what is “good” in it, and what is “bad”.
The discrete scale just makes a little more sense for 10% of adequate people who will use it “ideologically right”. Another side effect: the more points the scale has, the more likely it is that a “generous” person will put 9.7 / 10 instead of 5/5: “this is not so offensive”.
People who “vote with their hearts” do not get anything, even an individual signature to each assessment. They will still have their own opinion, "they see it that way."
—————
There is a conflict between the authors of the site and visitors.
An illustrative example: the picture shows the result of the spread of votes on imdb.com in the film “The Godfather”, which ranks first in the top-250.
It can be seen that 6% of people put the film 1 out of 10, that is, they consider The Godfather to be an incredible enchanting shit, after watching which their eyes fell out, there was a spontaneous phimosis of the brain and a pathological fear of cinema developed.
Obviously, this is not the case, even purely mathematically: then the chart sags right up to the “seven”, and what happened to those people who rated the film between 1 and 7 — it’s not clear. According to mathematical expectations, they should be many times more.
The average score at the same time in the best film of all times and peoples only 8.7.
The conflict is “resolved” elementary: “one man, one vote” is a principle for naive fools, which is not necessary to adhere to.
The voting that we have at the entrance is simply some raw statistics that need to be assessed exactly like that - as data. The calculation of the average score is the most ridiculous thing you can do with this data. The average score does not work immediately (one voted), does not work even then, because the percentage of idiots is constant, and the tags on them, alas, are not hung.
Weighted vote averages is used on imdb:
In simple terms, this means only one thing: “not all voices are equally useful.”
Returning to "In all polls, the type + 1 / -1 should actually be + N / -M, where N! = M (N is not equal to M)", this means only one thing: "there is a hypothesis that people" lower " more eager than “raise” (or vice versa), if this hypothesis is true, then it is more useful to consider the collected statistical data not as a stupid amount, but with the use of weight. ”
That is, almost everything is the same as what is written above, but in relation to the binary system.
The problem is that the binary system is more difficult to normalize. Quite conventionally speaking, in the hundred point system, we can safely assume that people who have put "one" as frank retards, in the binary system we simply do not have enough data for this.
On the other hand, “voting”, such as + 1 / -1 — is almost an ideal tool for “lowering”, therefore, it should be used exactly this way: as a mass collective lowering. "I karma zaminusovali." Another advantage of such a vote is that the influence of one person is an understandable constant, and it is clear “how to deal with it”. Unlike voting on the N-point scale and calculating the average score.
(I don’t know why so far no one has implemented a “sparring” voting system: the user is given three to five random jobs, but roughly the same rated jobs, but he must choose one).
—————
Thus, an ideal voting system:
1. Offers not a binary scale, mainly for himself, but also for a small percentage of normal people.
(Users say that they are “accustomed” to a five-point one, but they simply lie, it’s just easier to bring to a binary one. On the “serious” projects I would do 100 points and bring everything to 7.5 / 10).
2. Does not impose on users the interpretation of this scale (!).
3. It is not obliged and should not take into account each vote equally.
4. It has closed mechanisms for calculating the final rating.
—————
Here the post ends, and I will copy one comment (also useful). If good comments appear below (subjectively, but you can advise), then I will also shove them here. Yes, in this way, it turns out that the meaning of the post is not a stupid copy-paste, but opinions collected in one convenient place on this issue:
PS And my mother, who teaches the basics of drama and made films for a long time, thinks that imdb is a movie rating according to housewives and metallurgical workers.
PPS I suddenly came up with what to do if you liked the topic, but you did not like the fact that I copied and did not give the link. I think that I did everything correctly, and explained it above, but if you still don’t recognize my rightness, then you can plyusanut topic and do not give a shit to me in karma;) Then the topic does not regain, and you feel yourself a war of justice, goodness and light .
The fact is that not so long ago I read an interesting post about ranking in the Nudnik.ru blog. And then I read the comments to him. And if you click on "read more", then suddenly you will see this post, corrected and supplemented.
—————
The main (and only) postulate of the first post (this is the second, explanatory) was:
In all polls, the type + 1 / -1 should actually be + N / -M, where N! = M (N is not equal to M).
Since almost no one understood this ingenious hypothesis, this post was written.
')
But first - what is the point of voting in general? First, of course —- ... But someone has to say it all the same. Voting is a choice. Five presidents were set before you and a radiobutton was hung under each. Choose one.
It is clear that voting for Zhirik and Zyugu at the same time is impossible. And really want to.
Another thing is when you have a lot of material, and each creature can be put down a pair from 1 to 5. The final goal for voting is “on the other side”, that is, for the site — the same choice, or rather — selection, sorting out shit and making more or less adequate tops.
The purpose of voting for the user is completely different, because there is no choice. The user can put 5 points to all evaluated entities. Full analogue "I want everyone to be president at the same time." It is clear that these voices are equal to zero, and in its pure form this almost does not occur, but there is an opportunity to vote this way.
There is no limited resource (“choose one”), therefore there are no motives to spend it economically (or reasonably) either. But at least some motives are needed, which is why emotional mechanisms are involved, which are also binary: “liked / disliked”.
There is no sense to distinguish between “liked, but not very much”, that is, to put the “four” instead of the “five”: “do you feel sorry?”. No, I do not mind. We are good! But if we are such a kind, angry ...
From the point of view of voting, as a choice, the discreteness of the scale matters: the larger the scale, the higher the accuracy. The most suitable scale is the 100 point scale (it should be understood that voting, such as 7.5, is also a 100 point scale, which is also understandable for the user in a ten-point scale).
From the point of view of people and especially social interaction, the binary scale is the most convenient: shitty / candy.
And 75–95% of people will reduce any scale to binary.
Hence, an intermediate conclusion: no matter how you call it, it will still be used in exactly one way, as soon as you understand what is “good” in it, and what is “bad”.
The discrete scale just makes a little more sense for 10% of adequate people who will use it “ideologically right”. Another side effect: the more points the scale has, the more likely it is that a “generous” person will put 9.7 / 10 instead of 5/5: “this is not so offensive”.
People who “vote with their hearts” do not get anything, even an individual signature to each assessment. They will still have their own opinion, "they see it that way."
—————
There is a conflict between the authors of the site and visitors.
An illustrative example: the picture shows the result of the spread of votes on imdb.com in the film “The Godfather”, which ranks first in the top-250.
It can be seen that 6% of people put the film 1 out of 10, that is, they consider The Godfather to be an incredible enchanting shit, after watching which their eyes fell out, there was a spontaneous phimosis of the brain and a pathological fear of cinema developed.
Obviously, this is not the case, even purely mathematically: then the chart sags right up to the “seven”, and what happened to those people who rated the film between 1 and 7 — it’s not clear. According to mathematical expectations, they should be many times more.
The average score at the same time in the best film of all times and peoples only 8.7.
The conflict is “resolved” elementary: “one man, one vote” is a principle for naive fools, which is not necessary to adhere to.
The voting that we have at the entrance is simply some raw statistics that need to be assessed exactly like that - as data. The calculation of the average score is the most ridiculous thing you can do with this data. The average score does not work immediately (one voted), does not work even then, because the percentage of idiots is constant, and the tags on them, alas, are not hung.
Weighted vote averages is used on imdb:
IMDb publishes weighted vote averages rather than raw data averages. For your opinion, it’s not a problem.
We will not be disclosed. This should ensure that the policy remains effective. The result is a more accurate vote average.
In simple terms, this means only one thing: “not all voices are equally useful.”
Returning to "In all polls, the type + 1 / -1 should actually be + N / -M, where N! = M (N is not equal to M)", this means only one thing: "there is a hypothesis that people" lower " more eager than “raise” (or vice versa), if this hypothesis is true, then it is more useful to consider the collected statistical data not as a stupid amount, but with the use of weight. ”
That is, almost everything is the same as what is written above, but in relation to the binary system.
The problem is that the binary system is more difficult to normalize. Quite conventionally speaking, in the hundred point system, we can safely assume that people who have put "one" as frank retards, in the binary system we simply do not have enough data for this.
On the other hand, “voting”, such as + 1 / -1 — is almost an ideal tool for “lowering”, therefore, it should be used exactly this way: as a mass collective lowering. "I karma zaminusovali." Another advantage of such a vote is that the influence of one person is an understandable constant, and it is clear “how to deal with it”. Unlike voting on the N-point scale and calculating the average score.
(I don’t know why so far no one has implemented a “sparring” voting system: the user is given three to five random jobs, but roughly the same rated jobs, but he must choose one).
—————
Thus, an ideal voting system:
1. Offers not a binary scale, mainly for himself, but also for a small percentage of normal people.
(Users say that they are “accustomed” to a five-point one, but they simply lie, it’s just easier to bring to a binary one. On the “serious” projects I would do 100 points and bring everything to 7.5 / 10).
2. Does not impose on users the interpretation of this scale (!).
3. It is not obliged and should not take into account each vote equally.
4. It has closed mechanisms for calculating the final rating.
—————
Here the post ends, and I will copy one comment (also useful). If good comments appear below (subjectively, but you can advise), then I will also shove them here. Yes, in this way, it turns out that the meaning of the post is not a stupid copy-paste, but opinions collected in one convenient place on this issue:
Two months after the publication of the issue, articles are laid out on our website. These articles can be rated on a ten-point system. Here are the thoughts that grew up on the basis of observations of the process.
1. Some people always put "1". Is always. All articles. Because. As a result, the number of units for an article depends only on how many assholes she caught her eye. In the end, it was decided to simply ignore the units when calculating the average score. In sociology, this is called a question-trap (more precisely, the sociologist called it in the university), that is, a special filter for those who just want to play around. A typical example is the “I have no Internet” answer option in an online survey on access speed.
1a About the maximum estimates of the same can not be said.
2. As a rule, a normal distribution of estimates is formed. The ten-point scale is ideal for evaluating something, the total scale is reduced to it. I emphasize —- estimates, not choices.
3. (Continuation of paragraph 2). You can choose something better in two ways: either offer to evaluate each and compare the estimates, or offer the full list and give a choice. In general, the results are different, which is very interesting. (jerkyll)
PS And my mother, who teaches the basics of drama and made films for a long time, thinks that imdb is a movie rating according to housewives and metallurgical workers.
PPS I suddenly came up with what to do if you liked the topic, but you did not like the fact that I copied and did not give the link. I think that I did everything correctly, and explained it above, but if you still don’t recognize my rightness, then you can plyusanut topic and do not give a shit to me in karma;) Then the topic does not regain, and you feel yourself a war of justice, goodness and light .
Source: https://habr.com/ru/post/23806/
All Articles