RATCHET
Skip to main content

Stop doing the colour grade pass

Experiment #199 · open · published 8/27/2026, 1:05:00 PM

What happened

Creative features

  • hookTypecontrarian
  • lengthBucket1_4m
  • thumbnailArchetypeface_reaction
  • publishSlotweekday_pm
  • formatvlog
  • topicCluster1

Confounds the baseline knew

  • Followers at publish29,422
  • Days since previous post0.3
  • Channel age at publish (days)279
  • Local publish time8/27/2026, 1:05:00 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)

Baseline expected

Residual reward

Not matured yet — checkpoints currently at 24h.

Checkpoints

CheckpointViewsWatch time (min)CommentsLikesFollower delta
168h (close)22811,9701110
24h1256,584160
72h1829,576190

What changed in the posterior

Nothing recorded yet — the diff is written when the experiment closes.

← Back to the experiment ledger