Monty Hall, But Make It Physics – Part 3 of 4

A short series that builds intuition for quantum mechanics (without asking you to become a monk)

Series map

  1. File 001 / Part 1: Geometry, vectors, and the death of “50/50”
  2. File 002 / Part 2: Claude Shannon, Bayes, and the price of ignoring likelihoods
  3. File 003 / Part 3: Quantum mechanics and the power of projection
  4. File 004 / Part 4: Wolfram’s multiway systems (and the Ruliad) in toy form

Part 3 — The Quantum Physicist

You’re not dumb, the notation is just terrible

If you look at the Wikipedia page for POVM (Positive Operator-Valued Measure) (https://en.wikipedia.org/wiki/POVM), you will see a bunch of squiggly integrals (integral = a continuous “sum”; we won’t need it for 3 discrete doors) and jargon about “non-orthogonal resolutions of identity” (roughly: ways to carve up a space that are not perpendicular, but still add up to a complete “do nothing” identity matrix). It makes you want to close the tab and accept that some things are just for wizards.

We are fixing that today.

We are going to use the exact same picture from Part 1. The 3D room. The axes (coordinate directions). The arrow.

We are just going to give the parts their “Physics Names.”

Rule for this series: if I use a “physics name” (like ket or POVM), I’ll define it immediately in plain English, and we’ll keep building from simple primitives.

1) “Ket” – Column Vector = The Arrow (State)

$$|\rangle$$

$$|1\rangle$$

$$
\begin{pmatrix} 1 \ 0 \ 0 \end{pmatrix}
$$

It is a Column Vector. It represents “How the world IS.”
It’s an arrow sticking out into 3D space.

That’s it.

If you see a ket, think: “That’s the arrow.”$$|1\rangle$$

2) “Bra” – Row Vector = “The Floor Plan” (Measurement)

$$\langle |$$

In Part 1, we said your Choice was also a vector. If you pick Door 1, you are focused on the x-axis.

Physicists call this a Bra and write it backwards: $$\langle |$$

$$\langle 1 |$$

…literally just means:

$$
\begin{pmatrix} 1 & 0 & 0 \end{pmatrix}
$$

Strictly: the bra is the “question vector.” The measurement is the whole set of allowed questions (operators/effects).

3) The Dot Product = “Are we aligned?”

$$
\langle 1 | 1 \rangle
$$

$$
\langle 1 |
$$

$$
|1\rangle
$$

$$
\begin{pmatrix} 1 & 0 & 0 \end{pmatrix} \cdot \begin{pmatrix} 1 & 0 & 0 \end{pmatrix} = (1\cdot 1) + (0\cdot 0) + (0\cdot 0) = \mathbf{1}
$$

Translation:

If the car is in Door 2 (|2⟩) and you picked Door 1 (⟨1|):

$$
\langle 1 | 2 \rangle = 0
$$

Translation: “No match. You lose.”

The physics notation ⟨φ|ψ⟩ is just checking if the Arrow fits the Floor Plan.

In general, the probability is |⟨φ|ψ⟩|^2. Here it’s 0 or 1 because we’re comparing basis directions.

4) POVM = “a rule-based measurement menu” (Monty as apparatus)

$$
E_k \succeq 0,\qquad \sum_k E_k = I.
$$

Translation: the outcomes cover all possibilities (the sum-to-identity part), but they don’t have to be clean right-angle projectors (they can be “fuzzy” or asymmetric).

\[
|1\rangle=\begin{pmatrix}1\\0\\0\end{pmatrix},\quad
|2\rangle=\begin{pmatrix}0\\1\\0\end{pmatrix},\quad
|3\rangle=\begin{pmatrix}0\\0\\1\end{pmatrix}.
\]

Monty’s rules imply these likelihoods:

That entire rulebook can be encoded as two POVM “effects”:

$$
E_{2}=\mathrm{diag}\!\left(\frac{1}{2},\,0,\,1\right),\qquad
E_{3}=\mathrm{diag}\!\left(\frac{1}{2},\,1,\,0\right).
$$

They’re positive matrices. And they add up:

$$
E_2 + E_3 = I.
$$

The Matrices are POVM Elements

$$E_2 $$, $$E_3$$

In the quantum Monty Hall model, we operate in a 3-dimensional Hilbert space, where the basis states are

$$
E_{2}=\mathrm{diag}\!\left(\frac{1}{2},\,0,\,1\right).
$$

$$
E_{3}=\mathrm{diag}\!\left(\frac{1}{2},\,1,\,0\right).
$$

Properties of a POVM 

The user noted key properties that define these matrices as valid components of a POVM: 

The Born Rule and Probability Calculation

The final equation you provided uses the Born rule (specifically, the generalized version for POVMs) to calculate the probability of a given measurement outcome k( Monty opening door k)

$$
P(\text{Monty opens }k)=\mathrm{Tr}(E_k\,\rho),
$$

$$
P(\text{Monty opens }k)=\mathrm{Tr}(E_k\,\rho),
$$

5) Correlation = “The Shape of the Room Changed”

This is the final link to the “Switching” strategy.

When Monty opens Door 2 (say he opens Door 2), he hasn’t just given you information. He has chopped off part of the room.

Before Monty opens anything, the car is in exactly one door — the world is one of
|1⟩, |2⟩, |3⟩.

What’s “spread out” is not reality. It’s your knowledge.

If you want the quantum name for “uncertain but classical,” it’s a mixed state:

$$
\rho=\frac13\Big(|1\rangle\langle 1|+|2\rangle\langle 2|+|3\rangle\langle 3|\Big).
$$

(If you don’t like density matrices, read $$\rho$$ as “a probability distribution wearing a tuxedo.”)

  1. Before Monty opens a door, your prior belief over doors is uniform.
  2. Monty opens Door 2.
  3. This action is incompatible with the Car being at Door 2.
  4. It’s like a laser blast that deletes the y-axis (Door 2 axis).

Now look at your choice (Door 1) vs the World.

BUT HERE IS THE TRICK:

The laser didn’t hit Door 3. But the reason the laser fired at Door 2 was highly likely because the car was at Door 3.

If the car was at Door 1, the laser could have hit Door 2 or Door 3.

If the car was at Door 3, the laser had to hit Door 2.

The “Beam of Probability” on Door 3 is twice as thick as the beam on Door 1.

Because Door 3 forced Monty’s hand.

Correlation just means: The shape of the remaining room tells you where the arrow came from.

6) Sanity check (visceral correlation, no physics required)

Imagine you’re debugging a production outage.

There are three possible root causes:

You have no reason to favor any of them, so you start with a uniform prior (your starting belief before the test):

$$
P(A)=P(B)=P(C)=\frac{1}{3}
$$

Now you run a diagnostic test, and the test is rule-based:

(That’s Monty: he never opens the car door.)

The test prints: “B is not the problem.”

A novice brain says:

“Cool, B is eliminated. Two causes remain: A or C. Must be 50/50.”

But a careful engineer asks:

“Wait—does the test’s output depend on the actual cause?”

Define the rule:

So the likelihoods are:

$$ P(\text{prints “not B”} \mid A)=\frac{1}{2}, \quad P(\text{prints “not B”} \mid B)=0, \quad P(\text{prints “not B”} \mid C)=1 $$

$$
\mathrm{Score}(A)=\frac{1}{3}\cdot\frac{1}{2}=\frac{1}{6},\quad
\mathrm{Score}(C)=\frac{1}{3}\cdot 1=\frac{1}{3}
$$

Normalize by the $$\frac{1}{6}+\frac{1}{3}=\frac{1}{2}$$:

(“Normalize” just means: rescale the scores so they add up to 1 again.)

$$P(A\mid \text{not B})=\frac{\frac{1}{6}}{\frac{1}{2}}=\frac{1}{3},\quad P(C\mid \text{not B})=\frac{\frac{1}{3}}{\frac{1}{2}}=\frac{2}{3}$$

Translation back to Monty Hall:

If you only take one thing from this sanity check, take this:

Summary (no hidden jargon)