RATCHET
Skip to main content

What happens when editing at 2x speed

Experiment #174 · closed · published 7/22/2026, 3:37:00 PM

What happened

Creative features

  • hookTypequestion
  • lengthBucketunder_60s
  • thumbnailArchetypenone
  • publishSlotweekday_pm
  • formatshorts
  • topicCluster1

Confounds the baseline knew

  • Followers at publish26,216
  • Days since previous post2
  • Channel age at publish (days)243
  • Local publish time7/22/2026, 3:37:00 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)

186

Baseline expected

168

Residual reward

+0.17σ

raw 0.17σ · residual sd 0.599 log-views

Checkpoints

CheckpointViewsWatch time (min)CommentsLikesFollower delta
168h (close)1862,604190
24h1021,432050
72h1492,083170

What changed in the posterior

  • Post beat its baseline by 0.17σ. Moved: topic_cluster:topic_1 +0.000, format:shorts +0.000, hook_type:question +0.000.

    • topic_1-0.065-0.065 Δ+0.000±0.25
    • shorts-0.078-0.078 Δ+0.000±0.25
    • question0.1550.155 Δ+0.000±0.26
    • none0.0910.091 Δ+0.000±0.26
    • under_60s0.1210.121 Δ+0.000±0.25
    • topic_0-0.021-0.021 Δ-0.000±0.24
    • interview0.1460.145 Δ-0.000±0.24
← Back to the experiment ledger