RATCHET
Skip to main content

5 things editing at 2x speed

Experiment #99 · closed · published 4/8/2026, 9:24:00 PM

What happened

Creative features

  • hookTypenumber_list
  • lengthBucket20m_plus
  • thumbnailArchetypeobject_hero
  • publishSlotweekday_late
  • formattutorial
  • topicCluster6

Confounds the baseline knew

  • Followers at publish18,721
  • Days since previous post1.4
  • Channel age at publish (days)138
  • Local publish time4/8/2026, 9:24:00 PM

The residual reward

Raw views mostly measure how big the channel already was. The baseline predicts what this channel “should” have earned given the confounds above; the residual is how the post actually landed relative to that prediction — measured in standard deviations.

Observed (168h)

73

Baseline expected

134

Residual reward

-1.01σ

raw -1.01σ · residual sd 0.599 log-views

Checkpoints

CheckpointViewsWatch time (min)CommentsLikesFollower delta
168h (close)7338,325040
24h4021,079020
72h5830,660030

What changed in the posterior

  • Post missed its baseline by 1.01σ. Moved: topic_cluster:topic_6 +0.010, format:tutorial +0.004, hook_type:number_list +0.004.

    • topic_6-0.314-0.304 Δ+0.010±0.35
    • tutorial0.0610.065 Δ+0.004±0.28
    • number_list-0.349-0.344 Δ+0.004±0.29
    • 20m_plus0.0060.011 Δ+0.004±0.30
    • object_hero-0.263-0.259 Δ+0.004±0.29
    • weekday_late-0.271-0.267 Δ+0.004±0.29
    • topic_30.3000.298 Δ-0.003±0.31
    • weekend_pm0.2540.252 Δ-0.002±0.30
    • under_60s0.1660.164 Δ-0.002±0.29
    • topic_40.1810.179 Δ-0.002±0.28
    • topic_70.2620.261 Δ-0.001±0.33
    • question0.1200.119 Δ-0.001±0.29
    • topic_2-0.211-0.212 Δ-0.001±0.29
    • text_dominant0.0640.063 Δ-0.001±0.27
    • topic_0-0.111-0.112 Δ-0.001±0.29
    • claim0.0160.015 Δ-0.001±0.30
    • commentary-0.217-0.217 Δ-0.001±0.29
    • list-0.080-0.081 Δ-0.001±0.31
    • interview0.1550.155 Δ-0.001±0.29
    • before_after0.1210.120 Δ-0.001±0.31
    • story_cold_open0.2950.294 Δ-0.001±0.29
    • weekday_am-0.081-0.080 Δ+0.001±0.29
    • 10_20m0.0490.049 Δ-0.000±0.28
    • demo_first-0.130-0.129 Δ+0.000±0.28
    • 4_10m-0.241-0.241 Δ+0.000±0.29
    • shorts0.1170.116 Δ-0.000±0.29
    • 1_4m0.0310.031 Δ-0.000±0.27
    • vlog-0.025-0.025 Δ+0.000±0.29
    • contrarian0.0590.059 Δ+0.000±0.29
← Back to the experiment ledger