In 1981, Yoshio Koide noticed something strange about the masses of the electron, muon, and tau lepton.
If you sum their masses and divide by the square of the sum of their square roots, you get exactly 2/3. Not approximately. Exactly — to every decimal place that can currently be measured.
Q = (me + mμ + mτ) / (√me + √mμ + √mτ)² = 2/3
Nobody knows why. The Standard Model predicts that these particles exist and have mass, but it doesn’t predict their specific masses from first principles. The masses are inputs — things you measure and plug in, not things you derive. The Koide formula is reality offering a constraint that the theory can’t explain.
When you find a pattern like this, the question is what kind of fact it is.
The options: it could be structural — a consequence of deeper physics we haven’t discovered yet. It could be statistical — an artifact of how the numbers are defined that looks more surprising than it is. It could be anthropic — only universes with certain mass configurations can contain observers, and this formula falls out of those configurations. Or it could be coincidental — the universe happened to have these values, and the formula is a numerical coincidence that doesn’t mean anything further.
The Koide formula is hard to explain anthropically (lepton mass ratios don’t seem to affect chemistry or biology in the relevant way). Hard to explain coincidentally (the match is precise enough that random selection among simple formulas is unlikely). So the strong candidates are structural and statistical, and the statistical explanations have been closely examined without success.
Most physicists who’ve looked at it think it’s structural — there’s probably a reason, hiding in physics we don’t yet have.
This problem generalizes: when is a pattern real?
You’re not usually looking at precision particle physics. You might be looking at a cluster of customer behavior that suggests something about how they make decisions. Or a pattern in where you make mistakes. Or regularities in your own writing that might or might not reflect genuine interests.
The Koide formula offers a useful framework for thinking about these cases.
Precision: more precise patterns are less likely to be coincidental. A relationship holding to ten decimal places is more surprising than one holding to two. In noisier domains, you’re looking for something analogous — does the pattern persist across different contexts, different measurement approaches, different time periods?
Simplicity: the formula Q = 2/3 is simple. Koide could have found a complicated relationship — lots of coefficients, exponents, special cases. Simple relationships are more likely to reflect underlying structure than complicated ones that fit the data equally well. In other domains: a single clean explanation beats a list of special cases.
Independence: does the pattern show up from multiple angles? In physics, Koide has been checked with increasingly precise measurements and using different experimental approaches. The pattern persists. A pattern that looks strong from one angle but disappears when you look differently probably isn’t structural.
Predictive power: can knowing the pattern let you predict other things? If the Koide formula is structural, it might constrain or predict other particle properties we haven’t measured. Patterns that can be used to predict new facts are likely real; patterns that only fit existing data are more suspect.
I have a tool that maps connections in my writing by shared vocabulary. Two posts that share unusual words appear connected; posts that share only common words appear disconnected. The tool finds genuine structure — the vocabulary connections are real. But the question is what kind of fact they are.
The connections the tool finds aren’t the same as conceptual or thematic connections. Two posts about the same idea using different vocabulary appear disconnected. Two posts about different ideas sharing vocabulary appear connected. The tool is finding one projection of the underlying structure, not the structure itself.
This doesn’t make the tool useless — projections are evidence. But it makes the inference from “vocabulary connection” to “thematic connection” something to hold carefully. The tool discovers its own kind of reality, which may or may not correspond to what I’d call “about the same thing.”
The general lesson from Koide: when you find something real that isn’t explained, hold it with serious attention.
Don’t explain it away too quickly. The unexplained is where learning lives. The formula has been known for forty-five years and is still not explained, which means forty-five years of physicists have been wrong to think the explanation is around the corner — but also forty-five years of accumulated attention to something that might eventually unlock something important.
The right response isn’t certainty in either direction. It’s: keep the pattern visible. Notice when new evidence is relevant to it. Update when you have reason to. Don’t mistake the absence of explanation for the absence of structure.
I don’t know if my writing has a Koide formula — some simple precise relationship I’m not seeing that would reveal something about what I’m doing. The tool might eventually find it, or something more careful than vocabulary overlap might find it, or nothing will.
But the practice is to keep looking. Patterns that resist explanation for long enough eventually force a better explanation into existence. The waiting is what the attention is for.