🎲Learn Bayes' Theorem and Actually Use It to Update Beliefs
Update beliefs the way doctors, rationalists, and spam filters actually do — starting with frequency trees, finishing with a real prior-posterior update on a coin you've flipped yourself.
Phase 1Why Your Intuition About Evidence Is Broken
Frequency trees first, formulas second — build the intuition.
Even Harvard doctors get the mammogram question wrong
7 minHuman intuition systematically ignores the base rate. When the thing you're testing for is rare, even accurate tests produce mostly false alarms — and almost no one's gut can see it.
Draw the tree before you trust the number
7 minA frequency tree — imagining 1,000 people and walking them through each branch — turns every Bayesian problem into simple counting. No formulas required.
A belief is a probability you're willing to update
7 minThe prior is what you believed before the evidence. The posterior is what you believe after. Bayes' Theorem is the rule for moving from one to the other — and it's the only consistent rule there is.
P(H|E) = P(E|H) × P(H) / P(E), and why it's just bookkeeping
6 minThe formula looks like intimidating algebra but it's just the tree from yesterday written in compressed notation. Every term has a meaning you can see on the tree.
Phase 2Running the Update — Spam, Tests, and Coins
Apply Bayes to inboxes, medical tests, and fair coins.
Your inbox runs on Bayes and you never noticed
7 minModern spam filters are Bayes' Theorem applied at scale. Each word in an email updates a prior belief about spamminess — a long chain of tiny posteriors.
Positive result, but maybe ask for a second test
7 minWhen the prior is low, a single positive test is unreliable. But a second independent positive test — run in sequence, with the first posterior as the new prior — changes the math dramatically.
Is this coin fair? Flip it and find out — Bayes-style
7 minEven with just 10 flips, you can distinguish a fair coin from a biased one if you start with a clear prior and update with each flip. Bayes gives you a principled answer where null-hypothesis testing gives you p-values.
Odds are easier than probabilities — use them
6 minBayes' Theorem in odds form is just: posterior odds = prior odds × likelihood ratio. No division, no normalization, no headaches. It's the format real Bayesians use.
How much is a single data point actually worth?
7 minThe strength of any piece of evidence is exactly the likelihood ratio between the hypotheses it's being used to distinguish — no more, no less. This is the only honest answer to "how much should this change my mind?"
Phase 3Bayes in the Wild — Likelihood, MAP, and A/B Tests
Use Bayes to settle real-world decisions and experiments.
Your manager keeps changing the launch date — what's their pattern?
8 minThe likelihood function turns raw observations into a weighted verdict about competing hypotheses — the bridge from data to calibrated belief.
A user reports one bug — is it a regression or a fluke?
8 minMAP estimation picks the hypothesis with the highest posterior probability — it's the Bayesian answer to "what's the most likely explanation?" given data plus prior.
Your A/B test has 500 users — stop now or keep going?
8 minBayesian A/B testing gives you P(B > A) directly — a meaningful number you can act on — instead of a binary "significant or not" verdict.
Your doctor says "definitely not cancer" — should you believe her?
8 minCalibration is the test of whether expressed confidence matches Bayesian reality — and most "99% sure" statements are either under- or over-claiming.
Phase 4Updating a Real Belief With Real Data
Flip a coin ten times and defend your posterior.
Flip your coin 10 times and defend your conclusion
8 minRunning a full Bayesian analysis end-to-end on real data is how Bayes becomes a habit instead of a formula — and the habit is the thing that changes how you think.
Frequently asked questions
- What does 'prior' mean in Bayes' theorem?
- This is covered in the “Learn Bayes' Theorem and Actually Use It to Update Beliefs” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- Why does Bayes feel counterintuitive for medical tests?
- This is covered in the “Learn Bayes' Theorem and Actually Use It to Update Beliefs” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- How is Bayesian statistics different from frequentist?
- This is covered in the “Learn Bayes' Theorem and Actually Use It to Update Beliefs” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- Can Bayes' theorem be applied outside math problems?
- This is covered in the “Learn Bayes' Theorem and Actually Use It to Update Beliefs” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
Related paths
∫Understand the Fundamental Theorem of Calculus
Build the Fundamental Theorem of Calculus from accumulation first — feel why differentiation undoes integration before you ever see the formal statement. You'll finish by pulling a real dataset and verifying the theorem numerically with your own code.
🧮Understand Vector Spaces: The Rulebook Behind Linear Algebra
See why polynomials, pixel images, and 3D arrows all live in the same mathematical world — then design your own vector space of polynomials up to degree two and compute its dimension from scratch.
🔗Understand Correlation vs Causation With Real Examples
Go beyond 'correlation is not causation' — learn to spot confounders, recognize when randomization or natural experiments are doing the real work, and leave with a 30-second rule of thumb you'll actually use on tomorrow's headline.
🔵Learn Set Theory Basics: The Language Every Math Class Assumes
Stop squinting at ∪, ∩, and ∁ — shade Venn diagrams first, then translate them into clean notation, until you can model anything from music genres to probability events as sets you can actually picture.