How I filtered this, and why it matters

The raw rules (under 100k subs, at least 5× views-to-subs, past 6 months) matched 502 videos, and most were worthless. Three things were polluting the list.

Paid ads. The highest ratio in the raw set was a channel with 11 subscribers and 1.4 million views on a GLP-1 video. That is a bought campaign, not a breakout. Like rate catches these cleanly: genuine viral videos run a 1 to 5 percent like rate, while the ad hits sat at 0.00 percent across millions of views. Anything under 0.15 percent is gone.

Shorts are in, but labelled. Read the two formats separately, because the ratio does not mean the same thing in each. The Shorts feed pushes to strangers by design, so a Short's views-to-subs number largely reflects the algorithm handing it an audience. A long-form video has to earn its click, so its number reflects the packaging. That is why the top of the combined list is Shorts: the best Short here is 1834.3×, while the best long-form is 252.9×. Use the format buttons to compare like with like.

Telling them apart is harder than it looks. YouTube raised the Shorts ceiling from 60 seconds to 3 minutes in October 2024, so length settles nothing: plenty of these Shorts run well over a minute. Every video here was checked against YouTube itself rather than by duration.

Keyword accidents. Search does not understand meaning. "SEO" is a common Korean surname, so badminton matches pulled in; lo-fi "coding music" matched AI. A cheap model read every surviving title and dropped anything not genuinely on topic, or not in English.

There is also a 500-subscriber floor, since the very bottom of YouTube is ad accounts rather than channels, and a 50,000-view floor so that 5× means something. What survived is 358 real organic outliers, and every one of them is in the table below. Nothing is trimmed to a round number: you can sort and filter, so throwing away qualifying videos would only cost you examples.