Draft comparison

Write two to six versions of a post and see how each is likely to do, as a range rather than a single number. Versions that are too close to call are shown as tied.

Predictions come from a simple placeholder for now: your typical views plus a few hand-set nudges. The 80% ranges are wide, about 4x either way, and were calibrated on one creator's past posts. The trained forecasting model from this project will take its place.

Your typical views per post (the median of your last 10 is a good number)

Version A

Hook (what viewers see or hear in the first seconds)Caption with hashtags

Length (seconds)Posting hour

Version B

Hook (what viewers see or hear in the first seconds)Caption with hashtags

Length (seconds)Posting hour

Ranking

Version A is ahead by more than the model can resolve.

Version A

Tier 1 (top)

7K views expected

80% range 2K to 30K · 1.38x your typical post

Hook creates curiosity+12%
3 to 6 hashtags+7%
Length 7 to 30 s+7%
Evening posting slot+7%

Hook is short, which is good.

Try phrasing the hook as a question to create curiosity.
End the caption with a light call to action, such as a question for comments.

Version B

Tier 2

3K views expected

80% range 757 to 14K · 0.66x your typical post

Hook is long (>12 words)−12%
No hashtags−12%
Over 60 s−12%
Late-night posting slot−7%

Hook is 15 words; viewers decide in 1 to 3 seconds.

No hashtags, so the video is harder to place for new viewers.

75 s is long; retention usually drops past a minute.

Cut the hook to under 10 words, leading with the most surprising part.
Add 3 to 6 hashtags mixing niche and broad tags.
End the caption with a light call to action, such as a question for comments.
Test a cut under 30 s.

Everything runs in your browser. Nothing you type is sent anywhere.