Zhihu Good Things - Matrix Account Group Thinking Gameplay

Before solving the problem, let me first insert a digression: Why does brushing likes affect ranking?

In the process of optimizing Zhihu's ranking, I used more SEO thinking. Let me tell you how I used it.

① Zhihu = search engine

② Questions to be answered = Keywords in search engines

③Our own answers = our webpage

④All answers to the question = the ranking of web pages after the search engine inverted index

⑤Agree = external link

The main ranking rules of search engines are: relevance, timeliness, and authority.

Relevance: Relevance is a priority. If the answer is not relevant, it will be folded and punished. I will not comment on this (Zhihu is a high-quality question-and-answer platform. High-quality original + illustrated text will also affect the ranking)

Timeliness: The timeliness of Zhihu lies more in the fact that after we have just finished writing an article, it will be displayed in the comprehensive list for a period of time. Zhihu will decide whether to continue pushing it to more users based on the performance during this period.

If you observe carefully enough, you will find that the answer here is not the first one in this question. Especially if you have just finished writing an article on a question that is not very popular, there is a high probability that it will appear in the comprehensive article. (So ​​is there another way to play with timeliness? Have you thought of it? Prepare N accounts in batches and regularly guide them to the article or answer under the answer, so that exposure can be maintained for a long time.)

Authority: I divide this authority into two parts. One part is the weight of the account itself (similar to the search engine's score for the domain name. This score is multi-dimensional, and the most core and critical part is the center of the link farm, that is, there are a large number of reverse anchor text links, which can be understood as fans on Zhihu), and the other part is the weight of the answer to the question. (Similar to the weight of web pages in search engines)

The answerer himself = domain name (or the entire website)

Answer the question = a link (page) in the domain name or website

Why are some answers, which are clearly not agreed with, ranked higher than those that are agreed with? (Excluding relevance, it is probably because the account of the answerer has a high weight. In the eyes of search engines, a domain name with high weight and high trust value will also have a high trust value and high weight on one of its internal pages or pages. The same is true for Zhihu!)

When we are operating the Zhihu Good Things project, the [Agree] that is brushed by optimizing the ranking of answers in questions is actually the anchor text external link in the search engine. We can call this algorithm: recommendation algorithm.

The more users who recognize us, the higher our ranking will be. The higher the weight and trustworthiness of the account, the more we will be ranked. The more weight and trustworthiness of the account give us a [Like], the higher our ranking will be. (Zhihu ranking is different from search engines. There is an additional [opposition] button. It is found that opposition has a huge impact on ranking. You can also operate [opposition] to influence the ranking.)

Regarding the issue of losing likes by cheating

As mentioned above, you can use a third-party part-time job platform or ask friends to click "Agree" on your answer to influence the ranking of your answer in the question. Recently, due to the upgrade of Zhihu platform rules, the rate of brushing likes and losing likes is very high. After actual analysis, Jiahang found that it is mainly caused by the following three reasons:

① A large number of people agree with this question in a short period of time.

② The users who give likes are all low-salt and have few fans.

③ Part-time platform users give multiple likes in a short period of time

①Reason analysis:

Here I think of the strategy of posting external links back then. If I had 100,000 external links, I would not post them all at once, but would post them evenly over a certain period of time. The same is true for Zhihu. If a new question is directly visited and liked in a short period of time, it will be detected by the risk control system. Please remember that TA knows everything you do on Zhihu!

①Solution:

Suppose we want to influence the ranking of this answer in this question:

On Double Eleven 2020, which mobile phones priced around 1,000 yuan are worth recommending?

Step 1: Open https://www.zhihu.com/

Step 2: Search for "Which mobile phones around 1,000 yuan are worth recommending for Double Eleven 2020?" and click on the question that needs to be sorted.

Step 3: Find the answer you want to sort, scroll to the end of the answer, and click [Agree]

In addition, it should be noted that it is best to complete about 10 tasks each time. After completing these 10 tasks, you will basically not lose likes if you wait 10-30 minutes before completing another 10 tasks.

②Solution:

You can explain the situation when submitting the task. Requirements: 500+ Zhihu users with more than 1,000 fans will like the task. Note: The main reference is still based on fans!

Search engines, QQ groups, WeChat and other channels also have a large number of high-follower accounts that provide likes services. It has been tested that high-follower accounts have a more obvious impact on the results of question and answer ranking than ordinary accounts.

③Solution:

When submitting tasks on a part-time platform, it is best to leave an interval of more than 30 minutes, and you can submit them on multiple platforms.

The station group thinking solves the problem of mutual praise and account blocking

When we are doing a site group, we know that search engine sprockets and interconnection also require skills. If we connect without thinking, the final result will be annihilation.

In order to ensure the reading experience, I try not to use jargon as much as possible!

When operating a station group, we all know that the same C segment (ABCD) cannot be interconnected, that is, 111.123.111.7 cannot be linked to 111.123.111.8 (just understand it)

Zhihu matrix account should try to use one machine, one card and one number. Do not use virtual number segment mobile phone number to register. These are hard requirements to ensure that each account is independent. (In the search engine, we operate by using the station group IP, one IP corresponds to one domain name)

The process of maintaining an account is relatively simple. Just check the recommendations every day, and randomly agree + collect + comment, etc. to simulate a real person as much as possible.

In fact, mutual liking of answers is a sprocket process. Imagine that I have 10 accounts that like each other's answers from these 10 accounts every day. After a long time, it will be detected by Zhihu's risk control and the accounts will be 100% blocked. Our goal is to increase the data, increase randomness, and reduce the risk of being identified by Zhihu's risk control system.

​

This is a sprocket diagram of a site cluster. All the small sites are not interconnected. Instead, the weight is centralized to the five sites, ABCDE. These five sites then centralize the weight to the main site to increase the weight of the main site. This sprocket diagram can only be used as a reference, because what we need to consider is to increase the weight of all matrix numbers, rather than just concentrating power on one.

Our operations are as follows:

Split the matrix accounts into groups of 10, and divide them into five groups, A, B, C, D, and E. Each group of accounts will like each other, but accounts in different groups will not like each other!

During the mutual praise process of group A, the accounts in group A randomly give single likes to the accounts in the other four groups, such as BCDE (single anchor text in the search engine), namely:

A1 Agree with A2 Agree with A3 Agree with A4 Agree with A5 Agree with A6 Agree with A7 Agree with A8 Agree with A9 A10 Agree with B5 Agree with B8 Agree with C8 Agree with D9 Agree with E10 Agree with D2.

A2 agrees with A1 Agrees with A3 Agrees with A4 Agrees with A5 Agrees with A6 Agrees with A7 Agrees with A8 Agrees with A9 Agrees with A10 Agrees with B1 Agrees with B2 Agrees with C4 Agrees with D8 Agrees with E3 Agrees with D5.

This cycle repeats in sequence. During the mutual likes process of Group A, accounts that do not like each other are interspersed to disrupt Zhihu's risk control system. Note: the number of single likes must be N times greater than the mutual like matrix grouping to minimize the risk perceived by Zhihu's risk control. If there are not so many matrix accounts, you can randomly recommend and agree with other people's answers during the mutual like process. The more single likes you have for non-mutual likes, the smaller the chance of your mutual like matrix account being blocked. This applies to all accounts!

Author: Jiahang

Source: Jiahang Thinking