June 16, 2026
Zero drama? Statistically impossible
The Null Is Always False (Except When It Is True) (2014)
Math nerds melt down over whether “nothing” can still be something
TLDR: The post argues that “no real difference” can still be a useful idea, even if tiny random gaps appear in data. Commenters turned it into a stats cage match, fighting over whether science keeps mistaking meaningless specks for big discoveries.
A deceptively dry stats post turned into full-blown comment-section theater after it challenged one of science’s most annoying party fights: is the “no difference” result ever actually real? The article’s answer was basically: yes, sometimes the null — meaning “there’s really no meaningful difference” — can still make sense, even if tiny wiggles show up in samples. The key point is that random noise is not the same thing as a real effect, and a microscopic difference doesn’t magically become important just because you collected mountains of data.
That’s where the community split into camps. One side came in swinging with the classic hot take: “the null is always false” and p-values are just a factory for fake certainty. The other side pushed back hard, saying this is exactly how people confuse “not exactly zero right this second” with “not zero on average over time.” In plain English: if the world is messy and always changing, you can still ask whether there’s any stable difference worth caring about.
And yes, the jokes were relentless. Commenters had a field day with the idea that proving a laughably tiny effect might need 31 million people, with quips about needing “everyone on Earth and a clipboard.” Others mocked the old stats wars as a forever-beef between people who want numbers to be clean and a universe that absolutely refuses. Even the line about a stray electron ruining perfect truth got the nerds cheering — because nothing says internet drama like philosophers, scientists, and pedants arguing over whether zero is real.
Key Points
- •The article argues that although sample differences are almost never exactly zero, this does not mean the population null hypothesis is never true.
- •It distinguishes between variable sample mean differences and a population-level null hypothesis that assumes the true mean difference is exactly zero.
- •A quoted criticism from Jacob Cohen says the null is always false in the real world and that sufficiently large samples can reject even tiny deviations from zero.
- •The article responds that detecting extremely small effects can require impractically large samples, giving an example of about 31 million participants to detect a Cohen’s d of 0.001 in a t-test.
- •Citing Richard Hagen, the article says that when the null is true, increasing sample size does not guarantee that ever-smaller observed differences will become statistically significant.