What Is Luck
It looks like a fact about the world — of all the ways it could have gone, it went your way. But there are no other ways it went. Luck is a reading of your own predictor, and it settles nothing about what anyone deserved.
A man walks away from a car folded around a tree, and everyone at the scene reaches for the same word. He was lucky. Nobody treats this as a controversial description. It is one of the few words that survives every argument at a dinner table — the careful person who failed will use it, the careless person who succeeded will use it, and both will believe they mean the same thing by it.
The word does real work, too. It is how we stop ourselves crediting a good outcome entirely to skill, and how we stop ourselves blaming a bad one entirely on the person it happened to. Any account of luck that makes those two habits look foolish has probably gone wrong somewhere.
So I want to be careful. I do not think luck is a superstition. I think it is a real measurement, taken accurately, and then filed under the wrong heading.
What the word commits you to
State the ordinary meaning plainly: of all the ways it could have gone, it went the way that favoured you.
That is a comparative claim, and comparative claims need two sides. Favourable is not a property a thing can have on its own; it is favourable relative to something. So the ordinary account of luck requires a set of ways it did not go — alternatives substantial enough to be compared against the one that happened.
Ask where that set is kept.
The intuitive answer is that it is back there, in the past, in the branching that almost happened. And this feels obvious because the near miss is so vivid. You can describe the foot of clearance, the flight you nearly took, the job you almost accepted, in more detail than you can describe most of what actually occurred.
But look at what you consulted to build that description. A skid mark. A boarding pass. A photograph. A memory. Every one of those is a present object, here in the room with you, made of present material. The alternative was never observed by anybody. It was assembled, just now, out of records.
The near miss feels like a memory of what almost happened. It is a construction made this morning out of things lying around the house.
Start from the axiom instead
I work from one axiom — the past is immutable — and one consequence of it that people find much harder to accept.
Actuals only. The Past contains one settled totality, not a bundle of branches with one of them highlighted. Everything that could happen has already happened, and there are no unrealized possibilities, because none were left unrealized. Its formal signature is that the Shannon entropy of the Past is zero: one settled outcome, all of the mass sitting on it, nothing left over to spread across alternatives.
If that is right, then the comparison class luck depends on is not back there. It has never been anywhere but in a present head.
And there is already a standing name for a present head’s set of ways things might go.
It is a prediction
Which puts luck somewhere very specific. In my framework, what a person experiences is a quotient: R = A / E. The numerator is the actual. The denominator is expectation — largely subconscious prediction, and the ideas a person holds. Nobody experiences either term directly. You only ever experience the ratio.
The numerator is settled and write-protected. It can inform what you expect next; it cannot be re-opened. Which leaves exactly one live term in the whole expression.
Surprise is the log of the magnitude of that quotient, S = ln|R|, and it is signed — negative when the actual falls short of what was predicted. Luck is the case where that sign comes out positive, plus approval. Bad luck is the case where it comes out negative, plus complaint.
The identity nobody asks for
Run it forward and something falls out that I did not go looking for.
If luck is a low denominator against a given actual, then so is gratitude. Identical measurement. The two words differ only in where the credit is filed: luck puts it outside the person, gratitude keeps it inside. The number is the same number.
Which means the practice everybody recommends and nobody can quite explain — keep a gratitude list — is not a mood technique. It is denominator maintenance, and it operates on exactly the term that is available to operate on. There is a caveat to that, and I will get to it.
And it has an antipode. A high denominator against the same actual reads as a shortfall, and it reads that way again tomorrow, and the day after. Nothing about the world has to be going wrong for a life to register as a long run of bad luck. Entitlement is not a moral defect in this account. It is a calibration state, and it produces genuine, correctly-measured suffering.
Name the last thing you called lucky. Now say what you had predicted instead — out loud, in numbers if you can get them.
Most people cannot. Expectation is mostly subconscious prediction, and the word lucky is very often the first moment any of it reaches the surface. Luck is the receipt for a bet you never wrote down.
What this costs
Three things, and I would rather name them than let a reader discover them later and feel sold to.
One — you cannot install it
The obvious move, once you see the structure, is to simply expect less and become lucky on purpose. It does not work, and the reason is structural rather than moral. Expectation is a predictor, and surprise is additive — every reading you take afterwards is taken with the same instrument. A denominator tuned for how it feels rather than for what is true degrades every measurement you make with it from then on. You would be trading accuracy for mood, permanently, and paying in a currency you cannot see leaving.
This is also the line between the gratitude practice and its counterfeit, and it is the same line the framework draws everywhere else: direction of travel. Correcting a denominator that was inflated is calibration. Deflating one that was accurate, because the reading looks better low, is not. The first is you finding out what you had been silently pricing in. The second is you tampering with the instrument.
So the honest move is not to lower the denominator. It is to notice that it was never a fact about the world in the first place.
Two — it settles nothing about desert
This is the expensive one, and it takes a favourite move away from both sides of an old argument.
“He didn’t earn it, he got lucky” and its humble cousin “I was just lucky” are both claims about ordering — about whether rank is present and whether it is permitted to stand. That is a real question and it has real grounds. But luck does not supply them. Luck reports the gap between one person’s prediction and one outcome. It says nothing whatever about whether the ordering is right.
So the deflation and the humility are both honest feelings resting on a claim that is not there. Whether a distribution of rewards should stand is a question you have to answer on its own terms, with arguments about ordering, and the word luck has been quietly doing borrowed work in that conversation for a very long time.
Three — it does not even out
The folk consolation is that luck balances in the end. I have a version of a balance in my framework, and I want to be exact about what it does and does not offer, because this is the place where a reader most wants to be comforted and most deserves not to be.
The claim in the system is that every difference is paired at the node — the totals of the two sides cancel there. That is a statement about the totality. It is not a statement about your side of it.
No counterpart is realized here, none is owed to you, and none can be drawn on by anybody on this side. History is one-sided by construction. The books balance where you are not, and that is true rather than consoling.
And I should say that this particular result currently carries an asterisk in my own foundation: whether the geometry it rests on is doctrine or illustration is an open question I have not closed. So I will not lean on it. Either way, the answer to it evens out is no — not for you, not here, not in any form anyone can collect.
It is also grading the wrong thing
One more, quieter cost.
Luck is a verdict on success — on whether the outcome cleared the bar. But in this framework every artifact anybody makes is a miss. The denominator is not ours to command, so nothing lands exactly where it was aimed, and the evaluative question was never whether you hit. It is fidelity: how close the thing you made sits to the thing it was made in relation to.
Luck has no access to fidelity. It measures the distance between a prediction and an outcome, which is a live and useful number, and then gets asked to rule on a question that number has nothing to say about.
Where the claims come from
So the argument can be checked rather than taken on trust.
| Move | Rests on | Register |
|---|---|---|
| One settled totality; no unrealized possibilities | The actuals-only primitive | Doctrine |
| Shannon entropy of the Past is zero — all mass on one outcome | R4 | Doctrine |
| The actual is write-protected and cannot be re-opened | The axiom, R2 | Doctrine |
| A person experiences the quotient, never the numerator | P4, R8 | Doctrine |
| Surprise is the signed log of the magnitude of R | Locked definition | Doctrine |
| Therefore luck reports the denominator, not the world | The five rows above | Consequence |
| Luck and gratitude are one measurement, filed twice | P4, R8 | Consequence |
| A denominator tuned for comfort degrades later readings | R3 | Consequence |
| Desert is a question about ordering; luck does not address it | The real axis | Consequence |
| Every artifact is a miss; the question is fidelity, not success | R10, R11 | Doctrine |
| The totals cancel at the node, and nothing is collectable here | R7 — asterisked | Consequence |
| The car, the flight, the dinner table | — | Illustration |
Dissolving luck into a property of the observer is remarkably convenient for a framework built on the immutability of the past. Is this reasoning, or is it the prejudice of the idea showing?
Partly the second, and I hold that every idea is biased toward itself — including this one. Orientation and prejudice come as a pair and cannot be separated. What I can offer instead is the table above: the derivation is short enough to attack, and the place to attack it is the actuals-only primitive. Deny that, and luck goes back to being about the world.
On an ordinary Tuesday
Here is the whole of the practical residue, and it is smaller than the argument but it is not nothing.
When you catch yourself saying lucky or unlucky, you have just published a reading of your own instrument. Two questions follow, and they take about fifteen seconds.
What did I actually predict? Usually you will not know, and the not-knowing is the finding. You have been running a predictor you have never inspected.
Would I defend that prediction now? If yes, keep it and take the surprise as information about the world. If no, then what happened was not luck — it was a correction, and correcting is the entire job. Low fidelity under continuing correction is not a failure. The only failure available is to stop correcting.
- There are no other ways it went. The alternative you narrate so vividly was assembled today, out of present records.
- Luck is a reading of the denominator. Same actual, different prediction, different luck — with nothing about the world differing at all.
- Luck and gratitude are the same measurement. One files the credit outward, the other keeps the surprise inward.
- It settles nothing about desert. Whether rank should stand is a different question, on a different axis, and it needs its own grounds.
And one line to keep if the rest of it goes:
There is no luck in the world. There is the distance between what happened and what you were quietly betting on — and only one end of that measurement was ever yours.