This is the capstone. Six lessons built one relentless idea: a probability is a degree of belief, and every scrap of evidence updates it. You started with the prior, multiplied by the likelihood, and renormalized into the posterior. You learned to invert correctly — never confusing the chance of the data given the hypothesis with the chance of the hypothesis given the data — and to anchor on base rates so the prosecutor’s fallacy never fools you. You met conjugate priors where a Beta turns into a posterior Beta by simple counting, the Normal-Normal precision-weighted average that blends a return prior with noisy data, shrinkage and Black-Litterman as Bayes wearing a portfolio disguise, and credible intervals and MCMC for when the posterior is too gnarly to integrate by hand. No formula sheet, no hints, no take-backs: every answer locks the instant you submit, the wrong options are the exact traps that fool real analysts, and your score stays hidden until the end.
How this exam works
This is a graded exam. Questions arrive one at a time. Once you submit an answer it is final — there is no going back, no second try, and a wrong answer simply fails that question. Your score stays hidden until the very end, where you need 70% to pass. Read every option before you commit.
In the Bayesian view, what does a probability of 0.7 that a strategy is profitable actually represent?
Select an answer to continue.
Passed? Here's what you now own
You can read any probability as a belief and update it correctly: prior times likelihood, renormalized into a posterior. You never invert P(data given hypothesis) into P(hypothesis given data) by accident, you anchor on base rates with natural frequencies, and you treat a Beta prior or a return prior as pseudo-data blended with evidence by precision weighting. You recognize shrinkage and Black-Litterman as Bayes in portfolio clothing, you keep credible and confidence intervals straight, and you reach for MCMC — convergence-checked — when the posterior refuses to be integrated.
Big picture
Bayesian Finance — the whole toolkit
- Bayesian Finance
- Prior, likelihood, posterior
- Probability is a degree of belief
- Posterior is proportional to prior times likelihood
- Strong data overwhelm a weak prior
- Bayes' rule & base rates
- Never confuse P(D given H) with P(H given D)
- Prosecutor's fallacy ignores the base rate
- Natural frequencies keep the base rate visible
- Conjugate priors
- Beta in, Beta out: just add wins and losses
- Prior parameters are pseudo-counts
- Sequential equals batch updating
- Updating return & volatility
- Normal-Normal precision-weighted average
- Precisions add, so the posterior tightens
- More data sharpen the estimate
- Shrinkage & Black-Litterman
- James-Stein beats raw means for three or more
- Equilibrium returns are the prior
- Views are evidence weighted by confidence
- Credible intervals & MCMC
- Credible gives a direct probability; confidence does not
- Metropolis samples what you cannot integrate
- More samples never fix non-convergence
- Prior, likelihood, posterior
That’s Bayesian finance, end to end. You now own the machinery that turns belief and evidence into calibrated, updatable estimates — and the discipline to invert correctly, weight by precision, and quantify uncertainty honestly instead of pretending it away.