November 5, 2009 at 11:59 a.m.Top Blogs Based on the Yanedks API: Ranking Challenges
I don’t even know where to write, in Startups, in the Blogosphere, in SEO and SMO, or in PR. Chose "SEO and SMO".
The point is simple. Everyone knows that Yandex has announced the closure of its top blog entries, while providing everyone with an API for creating their own ratings.
After that, a clone of Top Yandex immediately appeared on the Lebedev Studio website, it seems, made so far without using the API, as well as whoyougle.com/blogs already using the latter.
Looking at this situation, I thought, why not make your top blogswith go and geisha , because it is always interesting to do something new. Perhaps I overestimate my strength, because I am neither Yandex, nor Tema, nor Nosik, maximum nickname of the latter . But, nevertheless, it is interesting to try, and even if nothing works, valuable experience will appear.
So, as I present the rating based on the Yandex API, that it could become at least some replacement for the original:
1. It should have a good, memorable name, which neither www.artlebedev.ru/tools/blogs nor www .whoyougle.com / blogs
In this regard, 2 domains are registered, TopBlogov.ru and BlogRating.org . Which one to choose for the service will be decided in the near future.
2. There must be trust in the rating. Confidence in the rating will only be possible if the ranking formula is well known and anyone can verify by this formula that the rating matches it.
3. The ranking formula should protect as much as possible from markups and give the most relevant result as possible. How this is combined with its openness is not yet very clear.
4. Ratings for separate parameters (number of links, traffic, number of comments) are not needed. Need one, but good.
Of course, for a service to “shoot”, this is not enough, we need a sane design, we need advertising, and so that Lebedev leaves for the next ethnographic expedition, the planets are in a row, and I would take this so far just a fun project more seriously. But this is not about that. It's about the calculation formula. Perhaps, with this discussion, we will simply help to improve the rating of the theme or nose, which is also good. Well, if the project is launched and interesting, at least the normal server on colocation is still loaded by 20 percent, we will have time to buy a second one.
The main 2 questions are what calculation algorithm to choose and whether it is worth making the formula publicly available.
Let me remind you that the Yandex API provides:
What is worth taking into account, and what is not? The first thought is to take everything and multiply, thus obtaining a certain general coefficient. Then comes the understanding that “the total number of links to the record for the entire time of its existence”, as well as “the total number of comments to the record for the entire time of its existence” should not have a positive weight, since they will increase the number of old dupe in the top.
It is also unclear why the links24 parameter is needed if there is links24weight. You can only use it. And you can subtract links24weight from links24 and use this as a reduction factor for wrapped records. This seems like a reasonable idea at the moment.
Another interesting parameter is the number of comments and commentators. It is clear that the number of commentators is more important (since you can write any number of comments from 1 account), but it is still unclear what ratio of the number of comments to the number of commentators is the most correct (hello to the shingles).
I also consider it correct to introduce a reduction factor if any of the 3 positive parameters is sagging, i.e., it is very small. For example, in the freshly weighted top of whoyougle.com, a post with 2 comments is now hanging, which suggests. Moreover, in order to introduce such a coefficient, it is required to calculate a certain average value of each parameter, which will be taken as a standard. I think that this is technically simple (to calculate such a value for a month, for example, for each paramtra).
As a result, we get 3 main parameters that are worth using:
Positive:
Negative:
These are chaotic thoughts. On this they end and I want to hear the opinion of the public.
The point is simple. Everyone knows that Yandex has announced the closure of its top blog entries, while providing everyone with an API for creating their own ratings.
After that, a clone of Top Yandex immediately appeared on the Lebedev Studio website, it seems, made so far without using the API, as well as whoyougle.com/blogs already using the latter.
Looking at this situation, I thought, why not make your top blogs
So, as I present the rating based on the Yandex API, that it could become at least some replacement for the original:
1. It should have a good, memorable name, which neither www.artlebedev.ru/tools/blogs nor www .whoyougle.com / blogs
In this regard, 2 domains are registered, TopBlogov.ru and BlogRating.org . Which one to choose for the service will be decided in the near future.
2. There must be trust in the rating. Confidence in the rating will only be possible if the ranking formula is well known and anyone can verify by this formula that the rating matches it.
3. The ranking formula should protect as much as possible from markups and give the most relevant result as possible. How this is combined with its openness is not yet very clear.
4. Ratings for separate parameters (number of links, traffic, number of comments) are not needed. Need one, but good.
Of course, for a service to “shoot”, this is not enough, we need a sane design, we need advertising, and so that Lebedev leaves for the next ethnographic expedition, the planets are in a row, and I would take this so far just a fun project more seriously. But this is not about that. It's about the calculation formula. Perhaps, with this discussion, we will simply help to improve the rating of the theme or nose, which is also good. Well, if the project is launched and interesting, at least the normal server on colocation is still loaded by 20 percent, we will have time to buy a second one.
The main 2 questions are what calculation algorithm to choose and whether it is worth making the formula publicly available.
Let me remind you that the Yandex API provides:
- links24 - the number of links to the entry,
- links, - the total number of links to the record for the entire time of its existence,
- links24weight - weighted number of links to a post (with filtered cheat and spam links),
- comments24 - the number of comments on the post,
- comments - the total number of comments on the record for the entire time of its existence,
- commenters24 - the number of different Russian-speaking commentators on the record,
- commenters - the number of different Russian-speaking commentators on the record for the entire time of its existence,
- visits24 - the approximate number of visitors to the record.
What is worth taking into account, and what is not? The first thought is to take everything and multiply, thus obtaining a certain general coefficient. Then comes the understanding that “the total number of links to the record for the entire time of its existence”, as well as “the total number of comments to the record for the entire time of its existence” should not have a positive weight, since they will increase the number of old dupe in the top.
It is also unclear why the links24 parameter is needed if there is links24weight. You can only use it. And you can subtract links24weight from links24 and use this as a reduction factor for wrapped records. This seems like a reasonable idea at the moment.
Another interesting parameter is the number of comments and commentators. It is clear that the number of commentators is more important (since you can write any number of comments from 1 account), but it is still unclear what ratio of the number of comments to the number of commentators is the most correct (hello to the shingles).
I also consider it correct to introduce a reduction factor if any of the 3 positive parameters is sagging, i.e., it is very small. For example, in the freshly weighted top of whoyougle.com, a post with 2 comments is now hanging, which suggests. Moreover, in order to introduce such a coefficient, it is required to calculate a certain average value of each parameter, which will be taken as a standard. I think that this is technically simple (to calculate such a value for a month, for example, for each paramtra).
As a result, we get 3 main parameters that are worth using:
Positive:
- links24weight
- commenters24 and comments24 according to some formula, let's start ( SQR (commenters24 * comments24))
- visits24
Negative:
- links24 - links24weight
- any of the positively influencing coefficients is less than the average by more than X%
These are chaotic thoughts. On this they end and I want to hear the opinion of the public.