Experiment #136 · closed · published 5/30/2026, 7:43:00 AM
Creative features
Confounds the baseline knew
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)
277
Baseline expected
148
Residual reward
+1.04σ
raw 1.04σ · residual sd 0.599 log-views
| Checkpoint | Views | Watch time (min) | Comments | Likes | Follower delta |
|---|---|---|---|---|---|
| 168h (close) | 277 | 3,878 | 1 | 14 | 1 |
| 24h | 152 | 2,133 | 1 | 8 | 0 |
| 72h | 222 | 3,102 | 1 | 11 | 0 |
Post beat its baseline by 1.04σ. Moved: topic_cluster:topic_3 +0.027, format:list +0.025, publish_slot:weekend_am +0.024.