Day 827 - Twitter algo tips - https://golifelog.com/posts/twitter-algo-tips-1680850572027
I've been reading threads about the Twitter algorithm which recently went open source. Just to gather up the tips and hacks:
*Caveat: The algo gets updated often, so this might only be a snapshot in time. And take the boosts and weights not as definite results but more as probabilities as the algo is multi-layered and complex. It behaves more like a symphony:*
- Follower:following ratio matters. Ideal is 0.6, where you follow less than 60% of the number of followers you have. But don't rush to unfollow a lot of people quickly, as that kind of behaviour resembles a bot and could get you shadowbanned.
- A tweet decays at a rate of 50% every 6 hours. (Side note: That's why I retweet it every 6h to bump it up and also to show it to folks in different timezones. It's inclusive as folks follow you to see your tweets, and timezones might make it less accessible to them. But might add noise if you RT too much, so do it with care.).
- The algorithm will show more of your tweets if you have a high TweepCred. If your TweepCred is less than 65, the algo will consider up to 3 of your tweets. If your TweepCred is more than 65, the 3-tweet limit is lifted. What TweepCred entails and how it's calculated is still not clear. This aligns with experience where tweeting anything more than 3 per day isn't as effective.
- Tweepcred is also based on the quality of users you interact with. Interact with low quality accounts with bad follower/following ratios and are marked as spam/nsfw/bots/toxic, you get penalized.
- Your account is clustered into a group with other similar profiles. This clustering helps extend your tweet reach beyond your followers to other similar people. Tweets won't do as well if you post “out of network” content, outside your cluster. That's why the advice Twitter is saying 1 thing 1000 ways. Repeating on your niche works (but boring).
- Using keywords in your tweets matter. Because Twitter uses the same algo as Google (PageRank).
- Multiplier boosts:
- Likes (30x)
- RTs (20x)
- Replies (1x) - getting likes + RTs boost more than getting replies
- Trusted circle (3x)
- Adding images/video (2x)
- Tweeting in English
- Follows
- Trending topics (1.1x) - trends, media, news
- Twitter Blue (2-4x)
- Ending tweets with a question mark
- Obvious but worth mentioning: Getting blocked, muted, abuse/spam reports, unfollows would lower your visibility score. Unfollows are only kept on server for past 90 days to prevent permanent shadowbanning.
- Negative feedback like "Show less often", block or mute (-74x)
- Report tweet clicked (-369x)
- The tweets someone sees in the For You feed is based on the probability that they will:
- like/RT the tweet
- click the tweet and reply
- stay for 2+ minutes
- check the account's profile
- Posting these in your tweets also lowers the reach of that tweet:
- only URLs
- no words
- misspelling, making up words, unknown language
- name only
- multiple hashtags
- tweet gets categorised as misinformation or blacklisted
- The myth is verified now: Links gets marked as spam if they are considered "non-news or non-media" links. But this doesn't mean your tweet won't go viral. If engagement on it is high because the link is relevant to the viral content, the tweet can still do well. It just starts off with a handicap.
- Blocking, muting others too much too often might also lower your algo ranking. Not too sure how true this is though. Still rumour level.
Sources:
1. https://twitter.com/IgorBrigadir/status/1641953684564201472
2. https://twitter.com/IgorBrigadir/status/1641945231636545536
3. https://twitter.com/sterlingcrispin/status/1641888492627083264
4. https://twitter.com/sterlingcrispin/status/1641922257470595072
5. https://twitter.com/NFT_GOD/status/1641914923826462721
6. https://twitter.com/steventey/status/1641892397440450560
7. https://twitter.com/NFT_GOD/status/1641925407468404739
8. https://twitter.com/petergyang/status/1642004729390858241
9. https://twitter.com/aakashg0/status/1641976943141699584
10. https://twitter.com/contentkuba/status/1642128038866321408
11. https://steventey.com/blog/twitter-algorithm
*What other Twitter algo tips did you see?*
*Caveat: The algo gets updated often, so this might only be a snapshot in time. And take the boosts and weights not as definite results but more as probabilities as the algo is multi-layered and complex. It behaves more like a symphony:*
- Follower:following ratio matters. Ideal is 0.6, where you follow less than 60% of the number of followers you have. But don't rush to unfollow a lot of people quickly, as that kind of behaviour resembles a bot and could get you shadowbanned.
- A tweet decays at a rate of 50% every 6 hours. (Side note: That's why I retweet it every 6h to bump it up and also to show it to folks in different timezones. It's inclusive as folks follow you to see your tweets, and timezones might make it less accessible to them. But might add noise if you RT too much, so do it with care.).
- The algorithm will show more of your tweets if you have a high TweepCred. If your TweepCred is less than 65, the algo will consider up to 3 of your tweets. If your TweepCred is more than 65, the 3-tweet limit is lifted. What TweepCred entails and how it's calculated is still not clear. This aligns with experience where tweeting anything more than 3 per day isn't as effective.
- Tweepcred is also based on the quality of users you interact with. Interact with low quality accounts with bad follower/following ratios and are marked as spam/nsfw/bots/toxic, you get penalized.
- Your account is clustered into a group with other similar profiles. This clustering helps extend your tweet reach beyond your followers to other similar people. Tweets won't do as well if you post “out of network” content, outside your cluster. That's why the advice Twitter is saying 1 thing 1000 ways. Repeating on your niche works (but boring).
- Using keywords in your tweets matter. Because Twitter uses the same algo as Google (PageRank).
- Multiplier boosts:
- Likes (30x)
- RTs (20x)
- Replies (1x) - getting likes + RTs boost more than getting replies
- Trusted circle (3x)
- Adding images/video (2x)
- Tweeting in English
- Follows
- Trending topics (1.1x) - trends, media, news
- Twitter Blue (2-4x)
- Ending tweets with a question mark
- Obvious but worth mentioning: Getting blocked, muted, abuse/spam reports, unfollows would lower your visibility score. Unfollows are only kept on server for past 90 days to prevent permanent shadowbanning.
- Negative feedback like "Show less often", block or mute (-74x)
- Report tweet clicked (-369x)
- The tweets someone sees in the For You feed is based on the probability that they will:
- like/RT the tweet
- click the tweet and reply
- stay for 2+ minutes
- check the account's profile
- Posting these in your tweets also lowers the reach of that tweet:
- only URLs
- no words
- misspelling, making up words, unknown language
- name only
- multiple hashtags
- tweet gets categorised as misinformation or blacklisted
- The myth is verified now: Links gets marked as spam if they are considered "non-news or non-media" links. But this doesn't mean your tweet won't go viral. If engagement on it is high because the link is relevant to the viral content, the tweet can still do well. It just starts off with a handicap.
- Blocking, muting others too much too often might also lower your algo ranking. Not too sure how true this is though. Still rumour level.
Sources:
1. https://twitter.com/IgorBrigadir/status/1641953684564201472
2. https://twitter.com/IgorBrigadir/status/1641945231636545536
3. https://twitter.com/sterlingcrispin/status/1641888492627083264
4. https://twitter.com/sterlingcrispin/status/1641922257470595072
5. https://twitter.com/NFT_GOD/status/1641914923826462721
6. https://twitter.com/steventey/status/1641892397440450560
7. https://twitter.com/NFT_GOD/status/1641925407468404739
8. https://twitter.com/petergyang/status/1642004729390858241
9. https://twitter.com/aakashg0/status/1641976943141699584
10. https://twitter.com/contentkuba/status/1642128038866321408
11. https://steventey.com/blog/twitter-algorithm
*What other Twitter algo tips did you see?*