What a Positive Test Really Means

Imagine your doctor calls. There's a rare disease that affects about 1 in 1,000 people. The test for it is "99% accurate." Yours came back positive.

How worried should you be?

Most people — including most doctors, in studies that have been run on this exact question — say something like 99%. That feels like the obvious answer. The test is 99% accurate, the test says yes, so you're 99% likely to have it.

The real answer is closer to 9%. Not 99%. Not 90%. Nine.

The gap between those two numbers is what Bayesian thinking is really about. And once you see it once, you can't unsee it.

The Trick: Stop Thinking in Percentages

Probability problems get easier the moment you stop using percentages and start imagining a crowd of people. Psychologists call this "natural frequency" framing, and it's the single most useful trick in this whole post.

Picture 1,000 random people walking into a clinic to get tested. Here's what actually happens:

• 1 person truly has the disease (the base rate is 1 in 1,000). The test is 99% accurate, so it correctly flags them. That's 1 true positive.
• 999 people don't have the disease. The test is 99% accurate, which means it's wrong 1% of the time. 1% of 999 is about 10 people who get a positive result anyway. That's 10 false positives.

So out of the 1,000 people, 11 of them get a positive result. Only 1 of those 11 actually has the disease.

1 out of 11 is about 9%.

That's it. That's the whole calculation. No formulas, no Greek letters, no conditional-probability notation. Just counting people in a crowd.

Why Your Intuition Lied

The reason "99%" feels right is that we tend to ignore the base rate — the underlying frequency of the thing we're testing for. We hear "99% accurate" and anchor on it. But accuracy is only half the story. The other half is how rare the thing is to begin with.

When the disease is rare, the small fraction of false positives gets multiplied across a huge number of healthy people. That handful of false positives ends up swamping the tiny number of true positives. The test isn't broken — it's working exactly as advertised. Your intuition is just missing a step.

Flip the situation around and the same math points the other way. If the disease affected 1 in 2 people instead of 1 in 1,000, a positive test would mean you almost certainly have it. Same test. Same accuracy. Wildly different conclusion. The base rate is doing most of the work.

Bayesian Thinking, in One Sentence

Here's the whole idea: what you should believe after seeing evidence depends on what you believed before.

That "before" belief is your prior. The evidence — a test result, a job offer, a partner texting back fast — updates that prior into a posterior: what you should believe now. Strong evidence shifts your belief a lot. Weak evidence shifts it a little. A rare prior takes a lot of evidence to overturn.

That's it. That's the framework. Everything else is bookkeeping.

Where This Shows Up Outside Medicine

Once you start looking, this pattern is everywhere:

• "This stock screener has a 70% hit rate." Sure — but most stocks don't moon. The base rate of a 10x return is tiny, so most "buys" are still false positives.
• "My ex is texting me again — they must have changed." Maybe. But the base rate of "person who behaved a certain way for years suddenly being different" is low. One text is weak evidence against a strong prior.
• "The interviewer loved me, I'm definitely getting an offer." Interviewers are nice to almost everyone. The base rate of "warm interview" leading to "offer" is much lower than the warmth suggests.
• "My headache must be something serious — I looked it up." Serious causes of headaches are rare. Common ones (sleep, screens, dehydration) are common. The prior matters.

In every case, the trap is the same: weighing the new evidence without weighing how unlikely the conclusion was to begin with.

How to Actually Use This

You don't need to do arithmetic in your head every time. You just need to ask three questions before you update your beliefs:

1. How common is this, normally? (The prior.)
2. How often does this evidence show up when the thing is true? (The true-positive rate.)
3. How often does it show up when the thing isn't true? (The false-positive rate.)

If the prior is small and the false-positive rate isn't tiny, slow down before you panic — or celebrate. The strongest-feeling evidence in the world can't carry a weak prior all the way to certainty in a single step.

Bayesian thinking isn't about getting the exact number right. It's about remembering to include the base rate at all.

If you want to put numbers on it, the Bayesian calculator will do the arithmetic for you. Plug in your prior, the true-positive rate, and the false-positive rate, and it'll show you the same crowd-of-1,000 logic in a tidy posterior. The point isn't to trust the calculator over your gut. It's to give your gut better inputs.