Recently, due to work reasons, I need to get in touch with Twitter's recommendation algorithm. I looked it up online and found that Musk had already made the main algorithm code public a year ago. The summary of the big guys online is as follows:

retweetCountParams retweet 20x

replyCountParams reply 1x

Here you can see that the effect of 20 replies is equal to 1 retweet

reputatiomParams reputation value 0.2x

It means that the bias towards big Vs is not too serious

LuceneScoreParams is the scoring mechanism of Lucene 2x

Lucene is a search engine used to discover relevance, so you can understand that this is a weight used to set content relevance

textScoreParams pure text 0.18x

UrlParams content with url 2x

isreplyParams reply to others' tweets 1x

favCountParams likes 30x

LangEnglishUIBoost English interface 0.5x

LangTweetBoost tweet in English 0.2x

LangDefaultBoost Default language weight 0.02x

unknownLanguageBoost unknown language weight 0.05x

The weight of the language can be ignored.

offensiveBoost offensive content 0.1x

intrustCircleBoost trust circle content 2x

multipleHashtagsOrTrendsBoost add multiple tags 0.6x

inDirectFollowBoost someone saw your tweet and followed you from then on 4x

tweetHasTrendBoost hot spot 1.1x

selfTweetBoost self-retweet 2x

tweetHasImageUrlBoost tweet with image 2x

tweetHasVideoUrlBoost tweet with video 2x

To sum up, it is to perform three-click one-click on the tweet, that is, like, forward and follow

In addition, the tweet must have pictures and videos, and you can also choose to reply to other people's tweets.

You can also forward your own tweets.