Appearance
【译】推特算法是如何运作的？（How the 𝕏/Twitter Algorithm Works?）【原文】
X/Twitter is undergoing a ton of changes right now including Elon Musk open sourcing the algorithm. If you’re a creator it would make sense to stay on top of these changes so you can implement them immediately.
X/Twitter 正在进行一系列重大改革，其中包括 Elon Musk 将推特推荐算法进行开源。作为内容创作者，了解这些最新变化非常重要，这样你就能及时调整创作策略。
We studied the algorithm and here is what you need to understand so that you can grow on X/Twitter.
我们深入研究了这个算法，现在就来为你解析一下如何利用这些变化在 X/Twitter 平台上实现更好的增长。
X 付费会员值不值得开通？（Is X Premium Worth It?）
Before we get into how the algorithm works and how posts are ranked. It’s worth to answer this most asked question. Paying for X premium extends your reach. Paying the monthly fee give you a boost. So if you are serious about growth on X/Twitter and you can afford $8/month then we suggest that you get yourself premium.
在讲解算法运作原理和帖子排名规则之前，让我们先来回答这个最频繁被问到的问题。订阅 X 平台的 premium 会员服务可以扩大你的内容触达范围。只要支付月费，你就能获得算法优先级的提升。所以，如果你想要在 X/Twitter 平台上实现快速增长，并且觉得每月 8 美元的费用在承受范围内，我们建议你可以考虑开通 premium 会员。
scala
object BlueVerifiedAuthorInNetworkMultiplierParam
extends FSBoundedParam[Double](
name = "scored_tweets_blue_verified_author_in_network_multiplier",
default = 4.0,
min = 0.0,
max = 100.0
)
object BlueVerifiedAuthorOutOfNetworkMultiplierParam
extends FSBoundedParam[Double](
name = "scored_tweets_blue_verified_author_out_of_network_multiplier",
default = 2.0,
min = 0.0,
max = 100.0
)It's clear in the algorithm that X premium users get a boost.
从算法设计来看，X premium 会员确实享有明显的优先级提升。
回复的重要性（Replies Matter More）
It is clear from the algorithm that replies are prioritized a lot and they carry a lot of weight on the scale. Here is how you can boost your replies and get more engagement.
算法数据显示，回复功能在社交平台中占据着极其重要的地位，评分权重很高。下面来看看如何通过提升回复质量来获得更多互动。
scala
object Scoring {
object ModelWeights {
object FavParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_fav",
default = 1.0,
min = 0.0,
max = 100.0
)
object RetweetParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_retweet",
default = 1.0,
min = 0.0,
max = 100.0
)
object ReplyParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_reply", // Replies Matter Here!
default = 1.0,
min = 0.0,
max = 100.0
)
object GoodProfileClickParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_good_profile_click",
default = 1.0,
min = 0.0,
max = 1000000.0
)
object VideoPlayback50Param
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_video_playback50",
default = 1.0,
min = 0.0,
max = 100.0
)
object ReplyEngagedByAuthorParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_reply_engaged_by_author", // Replies Matter Here!
default = 1.0,
min = 0.0,
max = 200.0
)
object GoodClickParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_good_click",
default = 1.0,
min = 0.0,
max = 1000000.0
)
object GoodClickV2Param
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_good_click_v2",
default = 1.0,
min = 0.0,
max = 1000000.0
)
object TweetDetailDwellParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_tweet_detail_dwell",
default = 0.0,
min = 0.0,
max = 100.0
)
object ProfileDwelledParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_profile_dwelled",
default = 0.0,
min = 0.0,
max = 100.0
)
object BookmarkParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_bookmark",
default = 0.0,
min = 0.0,
max = 100.0
)
object ShareParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_share",
default = 0.0,
min = 0.0,
max = 100.0
)
object ShareMenuClickParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_share_menu_click",
default = 0.0,
min = 0.0,
max = 100.0
)
object NegativeFeedbackV2Param
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_negative_feedback_v2",
default = 1.0,
min = -1000.0,
max = 0.0
)
object ReportParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_report",
default = 1.0,
min = -20000.0,
max = 0.0
)
object WeakNegativeFeedbackParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_weak_negative_feedback",
default = 0.0,
min = -1000.0,
max = 0.0
)
object StrongNegativeFeedbackParam
extends FSBoundedParam[Double](
name = "scored_tweets_model_weight_strong_negative_feedback",
default = 0.0,
min = -1000.0,
max = 0.0
)
}
}First of all, there are two types of replies:
首先，回复可以分为两种类型：
- Replies to your own comments
- 回复自己推文下的评论
- Replies to others
- 回复他人的内容
回复自己推文下的评论（Replies to your own comments）
You get 75 points when you reply to comments on your tweets. So you should not disregard when people reply to your tweets.
当你回复自己推文下的评论时，可以获得75分。因此，对于他人在你推文下的评论，一定要重视。
Here is a practical advice that you can follow:
这里有个实用的操作建议：
- Take 15 minutes
- 每天抽出15分钟
- Go to notifications
- 打开通知栏
- Go to Mentions
- 查看提及信息
- Reply to every comment that you get
- 回复每一条收到的评论
Bonus if you get them to reply back to you again so you keep the ball rolling.
如果能够引导对方继续回复你，形成良性互动，那就更好了。
回复他人的内容（Replies to other accounts）
You get 13.5 points when you reply/comment to other accounts.
当你回复或评论其他用户时，可以获得13.5分。
You need three types of accounts to reply to:
建议你关注并回复以下三类用户：
- Bigger accounts than you – they give you impressions
- 粉丝量比你多的账号 - 可以借助他们获得更多曝光
- Same size as you – they engage back with you
- 粉丝量与你相近的账号 - 容易产生互动
- Smaller accounts than you – they become your loyal followers
- 粉丝量比你少的账号 - 他们可能成为你的拥趸
Here is another practical advice that you can follow:
这里再分享一个实用技巧：
- Make three private lists
- 建立三个单独的关注列表
- Add 10 people to each (Big, same, and smaller accounts)
- 每个列表添加10个目标账号（分别对应大号、同量级账号和小号）
- Visit those lists daily and add value to their accounts
- 每天查看这些列表，为这些账号提供有意义的互动
One thing to keep in mind when replying is to add value. If you don’t have any value to add then do not reply.
记住一个原则：回复一定要有实质性内容。如果没什么有价值的内容可说，就不要回复。
发布更多视频和图片（Post More Videos and Images）
Elon Musk’s vision for 𝕏/Twitter is to become the everything app. From live streaming to payment. In attempts of this, X is prioritizing more long form videos.
Elon Musk 期望将 𝕏/Twitter 打造成一个全能型社交平台 (everything app)，集直播、支付等功能于一体。在这一战略下，X 平台正在重点推广长视频内容。
If you are active on other video platforms such YouTube, Instagram and TikTok, you can repurpose your video content and share it on X/Twitter.
如果你已经在 YouTube、Instagram 和 TikTok 等其他视频平台上活跃，你可以将这些视频内容进行适当调整后分享到 X/Twitter 上。
You can use a tool like Hypefury to help you post videos on X/Twitter and cross-post as reels to Instagram.
你可以使用 Hypefury 这类工具来协助你在 X/Twitter 上发布视频，同时还能将其作为短视频 (reels) 同步发布到 Instagram。
If you’re not making videos already then this is your chance to dive into it.
如果你此前没有制作过视频，现在正是一个很好的尝试机会。
If you can’t make videos at all then you can post tweets with images. They have a close weight to videos.
即使你完全没有制作视频的条件，你也可以选择发布带图片的推文，因为在平台算法中，图片内容的推荐权重与视频相差无几。
时间线内容排序机制揭秘（How Posts are Ranked on the Timeline）
Knowing how posts are ranked in social media is the number one topic that every user is interested in. With X/Twitter open sourcing the algorithm, it is no longer a secret how posts are ranked on X/Twitter.
在社交媒体领域，用户最关心的莫过于平台如何对内容进行排序。随着 X/Twitter 开源其算法，我们终于可以深入了解其背后的运作机制。
Here are the factors of how posts are ranked on the timeline:
影响时间线内容排序的主要因素包括：
- Graph Jet (图计算引擎)
- Sim Clusters (相似聚类)
- TwHin (推特异构信息网络)
- Real Graph (真实社交图谱)
- Tweep Cred (用户信用评分)
- Trust & Safety (信任与安全)
Now all these words may seem confusing but it’s simple. Let’s break them down.
这些专业术语看起来可能有点晦涩难懂，但实际上很容易理解。让我们来逐一解析。
1. Graph Jet (图计算引擎)
This is a graph that maps your relationship with other accounts.
这是一个用来展示你和其他账户之间关系网络的图谱系统。
Example:
举个例子：
How many of your followers are also following Hypefury?
假设要分析你的粉丝中有多少人同时也关注了 Hypefury？
If the score is high, you’ll see more of Hypefury’s tweets on the timeline. Even if you are not following Hypefury.
如果这个重合度很高，即使你没有关注 Hypefury，你的信息流中也会经常看到 Hypefury 的推文内容。
This also works the other way.
这种关系是双向的。
If a large numbers of Hypefury followers are also following you, then Hypefury will see a lot of your tweets on the timeline even if you are not following Hypefury.
同样地，如果 Hypefury 的粉丝中有很多人也关注了你，那么即便你们互不关注，Hypefury 的信息流中也会经常出现你的推文。
2. Sim Clusters (相似聚类)
X/Twitter puts you in groups with similar accounts based on your activities.
X/Twitter 会根据你的行为将你归类到相似用户群组中。
If you engage with posts related to “NBA”, then you will be added to the “NBA” cluster.
当你点赞、转发或评论 "NBA" 相关的内容时，系统就会将你划分到 "NBA" 兴趣群组中。
You will be show more NBA posts.
你的信息流就会显示更多 NBA 相关内容。
Pro Tip: Talk about one thing (niche down) so that you’re added in a cluster and accounts on that cluster will get to see your posts on the timeline.
小贴士：建议你专注于单一话题领域发帖，这样系统就能将你归入特定群组，让对相同话题感兴趣的用户在首页上看到你的内容。
3. TwHin (推特异构信息网络)
TwHin creates a document which stores all your actions such as likes, retweets, people you follow, etc.
TwHin 会生成一份用户行为档案，记录你在平台上的各种活动，例如点赞、转发以及关注用户等互动行为。
This document is then used to recommend posts to you and which accounts to follow.
系统会基于这份行为档案，为你推荐可能感兴趣的内容和值得关注的账号。
4. Real Graph (真实社交图谱)
This collects all the posts that you engage with (engagement data) then predicts similar posts that you are likely to engage with.
系统会收集你与所有帖子的互动记录（互动数据），并据此预测你可能感兴趣的相似内容。
Then those posts are show more in your timeline.
这些预测出的内容会在你的信息流中获得更高的展示优先级。
Example:
举个例子：
If you engage with your friend, no matter the topic, your friend’s posts will be show more on your timeline.
当你与某位朋友的内容发生互动时，无论内容话题是什么，该朋友的后续帖子都会在你的信息流中获得更多展示机会。
5. Tweep Cred (用户信用评分)
This is your reputation or “score” that X/Twitter gives you based on your
这是 X/Twitter 根据以下几个方面为你计算的信用分数：
- Avg engagement
- 平均互动数据
- Quality of content
- 内容质量
- Follower to following ratio
- 粉丝和关注人数的比例
- Age of your account
- 账号使用时长
- ...
The higher the score, the more you appear on the timeline.
这个分数越高，你的内容在信息流中展示的机会就越多。
Here is how you can increase your credibility (Tweep Cred):
想要提高你的用户信用分数 (Tweep Cred)，可以：
- Post good quality content
- 发布优质内容
- Don’t post sensitive content
- 避免发布敏感内容
- Unfollow low quality content
- 取消关注低质量账号
- Have a mix of engagement bait
- 保持多样化的互动内容
6. Trust & Safety (信任与安全)
As the name suggests, this filters sensitive content and shows you whatever binds to the policies of X/Twitter.
顾名思义，这个功能主要负责过滤敏感内容，确保显示的内容都符合 X/Twitter 的平台政策。
We talk more about this on what to avoid down below.
关于具体哪些内容需要注意规避，我们会在后面详细说明。
想要在 𝕏 平台上获得更多关注必须要避免的事情（Things to Avoid if You’re Serious About Growing on 𝕏）
These are things that you should work on avoiding at all costs.
以下这些事情是你必须要极力避免的。
1. NSFW 内容（NSFW Content）
This goes without saying but X/Twitter has been ruthless about their policies. Content about sensitive topics are downgraded.
这一点不言而喻，X/Twitter 在执行相关政策时非常严格。涉及敏感话题的内容会被压制展示。
2. 被举报（Getting Reported）
If someone reports you or you get flagged as “Not interested” you get deboosted.
如果有人举报你，或者将你标记为"不感兴趣"，你的内容曝光度会被大幅降低。
Now you may not be able to control the people may flag you as “Not interested” but you have control on not getting reported.
虽然你无法阻止别人把你标记为"不感兴趣"，但你可以避免被举报。
Be a good person, do not argue and spam other and do not impersonate anyone.
保持友善，不要与人争吵或骚扰他人，也不要冒充他人身份。
- The penalty when you get reported is -369 points
- 被举报会被扣除 369 分
- The penalty for “Not interested” is -74 points
- 被标记"不感兴趣"会被扣除 74 分
The move by Elon Musk to open source the X/Twitter algorithm is a smart one because creators know what they need to do and avoid for their posts to perform well and grow.
Elon Musk 将 X/Twitter 算法开源是个高明的决定，因为这样创作者就能清楚地知道该做什么、避免什么，才能让自己的内容获得更好的表现和成长。
More creators are likely to join 𝕏 because of this transparency and it’s a step closer towards achieving the everything app.
这种透明度很可能会吸引更多创作者加入 𝕏，这也是向打造超级应用 (Super App) 目标迈进的重要一步。