Replying to replies won’t boost your post 75x.
99% of people who copied my thread botched it. I've actually worked on algorithms for 15 years at places like Google.
And I've studied the changes to the algorithm just yesterday. Here's how to ATTACK the latest algorithm:
99% of people who copied my thread botched it. I've actually worked on algorithms for 15 years at places like Google.
And I've studied the changes to the algorithm just yesterday. Here's how to ATTACK the latest algorithm:
1. The algorithm constructs the feed INDIVIDUALLY for each user
Your tweet does NOT have individual stats.
It has stats on a "per-reader level."
The 75x boost only applies IF the system predicts a 100% probability the SPECIFIC reader will reply AND the author replies to that.
Your tweet does NOT have individual stats.
It has stats on a "per-reader level."
The 75x boost only applies IF the system predicts a 100% probability the SPECIFIC reader will reply AND the author replies to that.
2. So here's your point of LEVERAGE:
Increase the probability replies are replied to.
Case 1: The probability a person replies to a specific tweet is 10%, but 0% probability you reply.
→ 10% * 0% * 75x = 0x
Case 2: 100% you reply.
→ 10% * 100% * 75x = 7.5x.
Increase the probability replies are replied to.
Case 1: The probability a person replies to a specific tweet is 10%, but 0% probability you reply.
→ 10% * 0% * 75x = 0x
Case 2: 100% you reply.
→ 10% * 100% * 75x = 7.5x.
3. There is NO direct link penalty
The old de-boost that penalized their existence is GONE.
But, there are still ways links are penalized:
1. Users interact with them less, so their predicted scores are low
2. They reduce your likelihood of getting the 2-min bonus
The old de-boost that penalized their existence is GONE.
But, there are still ways links are penalized:
1. Users interact with them less, so their predicted scores are low
2. They reduce your likelihood of getting the 2-min bonus
4. The algorithm cares DEEPLY about time spent
As the probability increases that a user spends 2 minutes, you get up to a 10x boost.
It takes 400 words to drive 2 minutes of reading.
So, either:
· Write threads
· Write long tweets
· Or having interesting replies
As the probability increases that a user spends 2 minutes, you get up to a 10x boost.
It takes 400 words to drive 2 minutes of reading.
So, either:
· Write threads
· Write long tweets
· Or having interesting replies
5. Nothing beats QUALITY
The way you are really going to bend the trajectory of your tweet is by getting people who don't normally like or reply to do so.
That will increase the probability of those boosts to other people who are stingy with their love.
The way you are really going to bend the trajectory of your tweet is by getting people who don't normally like or reply to do so.
That will increase the probability of those boosts to other people who are stingy with their love.
6. Retweets are still HUGE
The retweet 1x boost shouldn't fool you. It doesn't mean retweets aren't valuable.
They are very valuable.
Yes, you only get a 1x boost if the system predicts a user will retweet. But, the retweet broadcasts your tweet to many more users.
The retweet 1x boost shouldn't fool you. It doesn't mean retweets aren't valuable.
They are very valuable.
Yes, you only get a 1x boost if the system predicts a user will retweet. But, the retweet broadcasts your tweet to many more users.
7. The algorithm does like videos people WATCH
The .005x boost seems insignificant.
But it can add up over time and on the margin.
It's one of the few "characteristics" that you can easily control to push up your numbers.
The .005x boost seems insignificant.
But it can add up over time and on the margin.
It's one of the few "characteristics" that you can easily control to push up your numbers.
8. You CAN'T report someone out of the algorithm
The -74x and -369x penalties look ominous.
But if you piss off someone who does lots of reporting, that's not gonna affect the probabilities for others.
It's only if you piss off someone not easily pissed off.
The -74x and -369x penalties look ominous.
But if you piss off someone who does lots of reporting, that's not gonna affect the probabilities for others.
It's only if you piss off someone not easily pissed off.
The algorithm is CONSTANTLY changing. Just take a look at these commits.
No one else on Twitter is following the GitHub repo as closely as me. Or producing guides as easy to read.
Follow me for weekly updates: @aakashg0
No one else on Twitter is following the GitHub repo as closely as me. Or producing guides as easy to read.
Follow me for weekly updates: @aakashg0
Here was my original thread that went viral (and 99% of copycats botched the read-out of):
https://twitter.com/aakashg0/status/1641976869460275201
And here's last week's update that 90% of those people never even saw, so they were operating on old information.
You can expect a similar update next week:
You can expect a similar update next week:
https://twitter.com/aakashg0/status/1645133881824415744
How to ATTACK the latest Twitter algorithm:
• • •
Missing some Tweet in this thread? You can try to
force a refresh
