A razor, in the thinking world, is not a blade — it’s a rule that shaves away the explanations you shouldn’t reach for first. The most famous one is about seven centuries old, and it does one job: when two stories explain the same facts, it tells you which one to bet on before you’ve gathered more evidence. Not which one is true — which one to bet on. Keep that distinction close; the whole lesson hangs on it.
Before you read — take a guess
Two explanations both fit every fact you have equally well. Occam's razor tells you to prefer the one that…
What Occam’s razor actually says
Picture two suitcases that hold the exact same clothes. One is stuffed with bubble wrap, spare hangers, and a second pair of shoes you’ll never wear; the other holds just the clothes. Same outcome, far less to carry. Occam’s razor tells you to grab the lighter case.
Precisely: among competing explanations that account for the facts equally well, prefer the one that requires the fewest assumptions (the fewest entities posited). It’s named for William of Ockham, a 14th-century English friar, whose principle is usually rendered as “entities should not be multiplied beyond necessity” — or plurality should not be posited without necessity. In plain terms: don’t invent extra moving parts your explanation doesn’t need. This is what philosophers call a principle of parsimony — a preference for leanness, for spending the fewest assumptions to cover the same ground.
An assumption here is anything that has to also be true for your explanation to hold. Each one is a separate bet. The razor says: the explanation that asks you to make fewer separate bets is the one to lead with.
Two words doing all the work
“Equally well.” The razor never says “pick the simple one.” It says “among the explanations that fit the facts equally well, pick the simpler one.” If a simple story leaves facts unexplained, it has already lost — simplicity was never the prize; fitting the facts with fewer assumptions is.
Fill in the core of the razor.
Pick the right option for each blank, then check.
Among explanations that fit the facts , prefer the one with the . The named friar is .
Nuance #1 — simpler, not simplest at all costs
If you cool an engine by removing the radiator, you have certainly made it simpler. You have also made it useless. Simplicity is only a virtue once the explanation still does its job.
Here’s the core nuance: the razor favors the simpler explanation, not the simplest one you can imagine. “The earth is flat” is gloriously simple — and wrong, because it fails to account for the facts (ships vanishing hull-first, time zones, photos from orbit). A simpler explanation that doesn’t fit the facts does not win on a simplicity technicality. It’s disqualified before the razor even gets a vote.
Einstein’s famous paraphrase captures it: “Everything should be made as simple as possible, but no simpler.” The “but no simpler” is the guardrail. You shave assumptions right up to the moment the explanation stops covering the evidence — then you stop.
| Explanation | Simple? | Fits the facts? | Razor’s verdict |
|---|---|---|---|
| ”The earth is flat” | Very | No (ignores orbit photos, horizon, time zones) | Disqualified — simplicity can’t rescue a misfit |
| ”The earth is a sphere” | Less | Yes | Preferred among fitting explanations |
| ”The earth is a sphere held by invisible cosmic ropes” | No | Yes (but adds unneeded entities) | Shaved — extra assumptions, no extra fit |
So the contest is two rounds: first “does it fit the facts?” (a gate everyone must pass), then “which of the survivors is leanest?” (the razor’s actual job).
Someone says: 'My explanation only has one assumption, so by Occam's razor it must be right — never mind that it doesn't explain half the data.' What's wrong?
Nuance #2 — it’s a prior, not a proof
A weather forecast that says “70% chance of rain” is a good bet, not a promise. If you walk outside to clear skies, the forecast wasn’t “wrong” — it was a reasonable lean given what was known. Occam’s razor produces exactly this kind of output: a lean, not a verdict.
Two terms make this precise. A prior is your best starting bet before you gather more evidence — where you’d put your money if you had to guess right now. A proof is a demonstration that something is actually true. The razor gives you the first, never the second. The leaner explanation is the better first bet; it is not the established truth.
This is the single most important — and most violated — point in the whole lesson. People run the razor, land on the lean explanation, and then treat that verdict as if it were evidence, defending it against incoming facts. But the razor’s preference is not data. The moment real evidence arrives, evidence wins; the razor was only ever the placeholder you used while you waited for it.
The central trap
Treating the razor’s verdict as a proof rather than a prior is the mistake this whole topic keeps circling back to. “It’s the simplest explanation” is a reason to bet on it — never a reason to stop checking it.
More likely — and lesson 3 (“Why Fewer Assumptions Win”) makes the probability argument rigorously. The short version: each independent assumption is a separate thing that has to hold, and each one can fail, so stacking assumptions multiplies the ways an explanation can be wrong. Fewer assumptions = fewer ways to be wrong = a higher prior probability. For now, just hold the counting move; the why comes next.
How to use it — count the assumptions
Using the razor is less philosophy than bookkeeping. For each explanation on the table, list every independent thing that must be true for it to hold, then count them. The explanation that needs fewer independent things to be true is the leaner one — your first bet.
“Independent” matters: you’re counting separate, free-standing assumptions, not words or syllables. An explanation can sound elaborate and still rest on one assumption; another can sound crisp while smuggling in five.
Try the counting move directly. For each scenario, expand both explanations and tally the assumptions each one quietly requires.
Occam’s razor
Count the assumptions, then cut
Two explanations for the same thing. Count what each one quietly assumes, then apply the razor:
The website went down at 3:00am.
A deploy we pushed at 2:55am broke it.
Assumes:
- We deployed right before it went down (the logs show this).
1 assumption
A hacker group timed a zero-day attack to our deploy window.
Assumes:
- A group is targeting us specifically.
- They have an unpatched zero-day.
- They struck in exactly our 5-minute window.
3 assumptions
The move in one line
Don’t ask “which story is more exciting?” Ask “which story needs fewer separate things to be true?” Then bet on that one — and stay ready to drop it.
Worked example A — the 3am outage
Walk the outage scenario all the way through, because it shows both halves of the razor: choosing the lean bet, and abandoning it the instant evidence arrives.
The fact: the site went down at 3:00am.
- Explanation 1 — bad deploy. Assumes: we deployed at 2:55am. That’s one assumption, and it’s not even speculative — the deploy log either shows it or it doesn’t. Checkable in seconds.
- Explanation 2 — coordinated zero-day attack. Assumes: (1) a group is targeting us specifically, (2) an unpatched zero-day exists, and (3) they struck in our exact five-minute window. That’s three independent assumptions, none of them yet evidenced.
Occam prefers Explanation 1: same fact explained, one assumption instead of three, all checkable. So the bad deploy is where you start looking.
Now the crucial second half. Suppose you pull the access logs and find an intrusion an hour before the outage. You drop the lean guess immediately. The razor was a starting bet, not a verdict — and the moment the logs (real evidence) contradict it, evidence wins. That’s not the razor failing; that’s the razor working exactly as designed. It got you looking in the most probable place first, cheaply, and then stepped aside when data showed up.
The logs reveal a genuine intrusion before the outage. You'd been betting on the bad deploy. What should you do?
Worked example B — hoofbeats
There’s a maxim taught to medical students: “When you hear hoofbeats, think horses, not zebras.” You hear hoofbeats outside (the fact). It could be a zebra that escaped a zoo — but the “horse” explanation needs almost nothing extra to be true (horses are everywhere), while “zebra” needs a whole stack of unusual assumptions (an escaped zebra, here, now). Same hoofbeats, far fewer assumptions for “horse.”
In diagnosis this is shorthand for common diseases are common. A headache is far more likely to be tension or dehydration than a rare brain tumor, so a good clinician investigates the common cause first — while staying alert for the zebra if the evidence starts pointing that way. This rests on base rates: how often each cause occurs in the population to begin with (the subject of an earlier course, thinking in probabilities / base rates). The lean explanation is usually the one with the higher base rate, which is why it’s the better prior — a thread lesson 3 will pull all the way through.
Sort each explanation by what the razor does with it — the lean first bet ('horse') or the assumption-heavy long shot ('zebra').
Place each item in the right group.
- Your friend is late because of traffic
- The headache is a rare brain tumor
- The outage was a timed zero-day attack
- Your friend is late due to a serious accident
- The headache is from dehydration
- The outage was our 2:55am deploy
How it connects to other models
The razor doesn’t work alone. It sits inside a small family of thinking tools you’ve already met or will meet soon — and knowing the boundaries between them keeps you from misusing each.
| Model | What it does | How it relates to Occam |
|---|---|---|
| First-principles thinking | Builds an explanation from scratch by stripping inherited assumptions | First principles constructs a fresh explanation; Occam chooses among ready-made ones. Both refuse to inherit bad assumptions — one by rebuilding, one by selecting. |
| Base rates / thinking in probabilities | Weighs how common each cause is to begin with | The lean explanation usually has the higher prior — base rates are why “horses” beats “zebras.” |
| Confirmation bias | The pull to favor what flatters our existing story | We tend to inflate the assumption-heavy explanation because it’s dramatic and confirms our narrative. The razor is a deliberate counterweight. |
The first-principles distinction is worth dwelling on. First-principles thinking is about building an explanation: you tear a problem down to what you know is true and reason up. Occam’s razor is about choosing among explanations already on the table. You reach for first principles when no good explanation exists yet; you reach for the razor when several do and you must pick which to bet on.
Match each model to the job it actually does.
Pick a term, then click its definition.
The pitfall — razor as proof, and over-shaving
The dominant pitfall, again, is treating a razor as a proof rather than a prior. Signs you’ve fallen in: you defend the lean explanation against incoming facts; you say “it’s the simplest, so it’s settled”; you stop checking once the razor “decided.” The fix is a habit — when you state the lean explanation, append “…for now, until evidence says otherwise.”
There’s an inverse trap worth flagging: over-simplifying — shaving away an assumption the explanation genuinely needs, so it stops fitting the facts. That’s the “no simpler” guardrail being violated from the other side. (Lesson 5 handles this failure mode in full; here, just know that the razor cuts unnecessary assumptions, never necessary ones.)
Spot the trap. Which statement misuses Occam's razor?
Recap
Big picture
Occam’s razor at a glance
- Occam’s razor
- What it says
- Among explanations fitting the facts equally well…
- …prefer the fewest assumptions (parsimony)
- Named for William of Ockham (14th c.)
- Two nuances
- Simpler, not simplest-at-all-costs ('but no simpler')
- A prior / first bet, NOT a proof
- How to use it
- List each explanation's independent assumptions
- Count them; lead with the fewest
- Drop the lean guess when evidence overrides it
- Connections
- First principles: builds vs. Occam: chooses
- Base rates: why the lean guess is the higher prior
- Confirmation bias: the force the razor resists
- Pitfall
- Treating the verdict as proof, not prior
- Over-shaving a needed assumption (lesson 5)
- What it says
Occam’s razor — check yourself
Two explanations fit every known fact equally well; one needs two assumptions, the other needs five. What does the razor recommend?
Check your answer to continue.
You can now count assumptions and lead with the leanest explanation. But we’ve been asserting that fewer assumptions makes an explanation more likely to be true — without proving it. Next, in “Why Fewer Assumptions Win,” we’ll make that case with probability: why each extra assumption is another way to be wrong, and how that quietly stacks the odds against the bloated explanation.