X Algorithm Playbook
The definitive guide to maximizing your reach on X (Twitter), based on reverse-engineering the open-source algorithm.
X Algorithm Playbook
The definitive guide to maximizing your reach on X (Twitter)
Based on reverse-engineering the open-source algorithm
The Algorithm in 60 Seconds
Your Post Score = Σ (weight × P(action))
POSITIVE: copy-link share (20.0), reply/quote/DM-share (5.0), follow (4.0)…
NEGATIVE: report (−234), mute (−58.8), "not interested" (−43.2), block (−31.2)
Key insight: Phoenix predicts 64 action classes feeding ~26 weighted terms — real values now public. Weights multiply predicted probabilities, not raw counts. Maximize positive, avoid negative.
🆕 Updated for the September 18, 2026 release: real published action weights, the production Phoenix model, a 54-rule visibility-filtering engine, SimClusters candidates, VMRanker DPP reranking, a new-author cold-start boost, and Under the Hood transparency reports. See what changed →
Quick Navigation
| Start Here | Go Deeper |
|---|---|
| 10 Golden Rules | Scoring System |
| Pre-Post Checklist | Action Weights |
| Common Mistakes | Filter System |
| What’s New (Sep 2026) | Algorithm FAQ |
The 10 Golden Rules
- Get sent, get replies — Copy-link share (20.0) + replies (5.0; 20.0 from mutuals) lead
- Avoid negative actions — Report −234, mute −58.8, not-interested −43.2, block −31.2
- Space your posts — Author diversity: 2nd post in a feed scores ×0.625
- In-network first — Verified ×0.75 out-of-network discount
- Video > Image > Text — Verified 10s minimum; direct weight is small (0.07)
- Dwell time matters — Longer content = higher engagement signal
- Don’t trigger filters — 17 pipeline filters + 54-rule visibility engine
- Engage authentically — Algorithm tracks your interaction patterns
- Niche down — Semantic IDs + SimClusters reward consistent topics
- Quality > Quantity — One great post beats five mediocre ones
Action Weights
| Action | Real weight | Impact |
|---|---|---|
| Share via copy link | 20.0 | Top positive |
| Reply · Quote · DM share | 5.0 | Strong |
| Follow author | 4.0 | Intent |
| Retweet | 1.0 | Good |
| Like | 0.5 | Baseline |
| Block | −31.2 | Very negative |
| “Not interested” / Mute | −43.2 / −58.8 | Harsher than block |
| Report | −234.0 | Most severe |
Real production values — published Aug 2026 in
params/param.rs. They scale predicted probabilities, not raw counts. Details →
Repository Contents
x-algorithm-playbook/
├── rules/ # 7 core strategy guides
├── checklists/ # 3 actionable checklists
├── case-studies/ # 2 real-world examples
└── reference/ # 5 technical deep-dives
Source
This playbook is based on analysis of:
Star this repo if it helped you grow on X!