Experiment #124 · closed · published 5/12/2026, 1:13:00 PM
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)
151
Baseline expected
118
Residual reward
+0.42σ
raw 0.42σ · residual sd 0.599 log-views
| Checkpoint | Views | Watch time (min) | Comments | Likes | Follower delta |
|---|---|---|---|---|---|
| 168h (close) | 151 | 7,927 | 1 | 8 | 0 |
| 24h | 83 | 4,360 | 0 | 4 | 0 |
| 72h | 121 | 6,342 | 0 | 6 | 0 |
Post beat its baseline by 0.42σ. Moved: hook_type:claim +0.020, topic_cluster:topic_4 +0.017, format:vlog +0.015.