Tips to be a better problem solver [Last live lecture]

Q: How do the triangle areas turn into an infinite series, and why does ln(2) appear?

Each region for floor(x/y)=k is a triangle whose base is the difference between two intercepts (e.g., 1/(k+1) − 1/(k+2) style differences) and whose height is 1 in the chosen coordinate orientation. So each probability term is (1/2)·base·height, giving a sequence like 1/2, 1/6, 1/12, … with alternating structure when expressed as (1/2) times an alternating sum of reciprocals. The lecture then recognizes the alternating sum as ln(2) by converting it into a function with powers (a geometric-series form) and evaluating an integral from 0 to 1, where ∫(1/(1+x)) dx produces ln(2).

Q: What role does the “gut check” play, and how does it correct the final probability?

After deriving a closed form, the lecture checks reasonableness numerically and against the diagram’s area bounds. A deliberate earlier mistake leads to an incorrect candidate probability (about 0.35), which fails because the region for floor(x/y)=0 alone already has probability 1/2. The correction comes from noticing the alternating sum’s sign pattern isn’t exactly the standard ln(2) series; adjusting the decomposition changes the result to 2 − ln(2) ≈ 0.653. This corrected value then matches both area intuition and simulation.

TL;DR

Treat unfamiliar problems by extracting and using the defining features of the setup (e.g., equal radii on a circle) before reaching for advanced tricks.

Briefing Cornell Notes

Briefing

A practical way to become a better problem solver is to treat every unfamiliar math puzzle as a chance to exploit definitions, symmetry, and “two descriptions of the same object”—then verify the result with reasonableness checks and (when possible) computation. The lecture builds this toolkit through geometry, trigonometry, and a probability problem that ends with a clean closed-form answer.

It starts by reframing “problem solving” as something hard to teach: progress must happen even when the problem is new. Nine principles are promised, but the through-line is consistent—small, deceptively simple habits compound into major breakthroughs. The first major example is the inscribed angle theorem in circle geometry. Rather than relying on memorized facts, the approach emphasizes using defining features: points on a circle share the same radius, so triangles formed with those radii are isosceles. Naming angles and then writing down the angle-sum constraints turns a diagram into manipulable equations. By canceling shared angle variables, the lecture derives a key relationship: the large inscribed angle equals twice the small inscribed angle (θ_L = 2θ_S). That “burn it into memory” identity becomes a reusable engine for later problems.

The same identity then powers a geometric proof of the double-angle identity for cosine: cos²(θ) = 1/2(1 + cos(2θ). The method again follows a general strategy: find one object that can be described two ways. On the unit circle, cosine is a coordinate (a length), but cosine squared is interpreted via projection symmetry—projecting the same unit-length segment in two equivalent ways multiplies two cosine factors, producing cos²(θ) as a length rather than an area. To introduce 2θ, the lecture constructs a right-triangle-inside-a-circle setup where the inscribed angle theorem (and Thales’ theorem as a special case) relates the relevant angle to half of a diameter. The resulting geometry yields the algebraic identity without invoking complex numbers.

The lecture’s capstone is a probability question: pick two independent uniform random numbers x and y from [0,1], consider the ratio x/y, take its floor, and ask for the probability that this integer is even. The solution is built from the same problem-solving habits: translate the setup into geometry by treating (x,y) as a random point in the unit square; convert “floor(x/y)=k” into inequalities that define regions bounded by lines like y = x/n and y = x/(n+1); compute probabilities as areas of infinitely many triangles. The resulting infinite alternating sum connects to ln(2) via a calculus trick (turning the sum into an integral using geometric-series structure). A deliberate mistake is introduced earlier to motivate a final “gut check”: the first closed form fails a reasonableness test, and correcting the sign logic yields the final probability 2 − ln(2) ≈ 0.653.

To close the loop, the lecture uses programming as a verification tool: Monte Carlo sampling with NumPy estimates the probability near 0.65, and histograms show the distribution of floor(x/y). The overall message is that strong problem solving is less about rare genius and more about repeated exposure to patterns—definitions, symmetry, “same object, two descriptions,” area/inequality geometry, and disciplined verification.

Cornell Notes

The lecture argues that progress on unfamiliar problems comes from repeatable principles: use definitions, add structure (often by drawing or naming), exploit symmetry, and look for a single object that admits two different descriptions. Geometry provides the template: using circle definitions and isosceles triangles leads to the inscribed-angle relationship θ_L = 2θ_S. That identity then enables a geometric proof of the trigonometric double-angle formula cos²(θ) = 1/2(1 + cos(2θ) ) by interpreting cosine squared as a length via projection symmetry and using inscribed angles to bring in 2θ. The probability problem is solved by converting floor(x/y) into regions in the unit square, summing triangle areas into an alternating series that evaluates to ln(2), then correcting a sign error through a reasonableness check. The final probability is 2 − ln(2) ≈ 0.653, confirmed by simulation.

How does the lecture turn an abstract geometry diagram into solvable equations?

It starts from defining features: points on a circle are all the same distance from the center, so radii are equal. That makes triangles formed with radii isosceles, forcing equal base angles. Then it names angles (including “prime” angles created by adding radii lines) and writes angle-sum constraints: triangles sum to π radians (180°). With three equations in hand, shared angle variables can be canceled to relate the inscribed angles directly, yielding θ_L = 2θ_S.

What is the general “two descriptions of the same object” strategy, and how is it used for cos²(θ)?

The strategy seeks one geometric quantity that can be computed in two different ways, producing a non-obvious identity. For cos²(θ), cosine is first tied to a unit-circle coordinate (a length). Cosine squared is then reinterpreted as a length via projection symmetry: projecting a unit segment onto the x-axis scales by cos(θ), and projecting in the complementary way introduces another cos(θ) factor, so the product becomes cos²(θ). A second description uses inscribed-angle geometry to bring in 2θ, leading to cos²(θ) = 1/2(1 + cos(2θ)).

Why does converting (x,y) into a point in the unit square make the probability problem tractable?

The ratio floor(x/y) is awkward algebraically, but geometrically it becomes about regions. Choosing x and y uniformly on [0,1] means (x,y) is uniform over the unit square. Conditions like floor(x/y)=0 translate to inequalities such as 0 ≤ x/y < 1, which becomes x < y (boundary y=x). More generally, floor(x/y)=k becomes k ≤ x/y < k+1, which can be rewritten as k·y ≤ x < (k+1)·y, producing wedge/triangle regions bounded by lines y = x/(k+1) and y = x/k. Probabilities become areas.

How do the triangle areas turn into an infinite series, and why does ln(2) appear?

Each region for floor(x/y)=k is a triangle whose base is the difference between two intercepts (e.g., 1/(k+1) − 1/(k+2) style differences) and whose height is 1 in the chosen coordinate orientation. So each probability term is (1/2)·base·height, giving a sequence like 1/2, 1/6, 1/12, … with alternating structure when expressed as (1/2) times an alternating sum of reciprocals. The lecture then recognizes the alternating sum as ln(2) by converting it into a function with powers (a geometric-series form) and evaluating an integral from 0 to 1, where ∫(1/(1+x)) dx produces ln(2).

What role does the “gut check” play, and how does it correct the final probability?

After deriving a closed form, the lecture checks reasonableness numerically and against the diagram’s area bounds. A deliberate earlier mistake leads to an incorrect candidate probability (about 0.35), which fails because the region for floor(x/y)=0 alone already has probability 1/2. The correction comes from noticing the alternating sum’s sign pattern isn’t exactly the standard ln(2) series; adjusting the decomposition changes the result to 2 − ln(2) ≈ 0.653. This corrected value then matches both area intuition and simulation.

Review Questions

In the inscribed-angle derivation, which equal-length facts (radii) create isosceles triangles, and how do those equalities translate into angle-sum equations?
For the probability problem, how do inequalities like k ≤ x/y < k+1 become line boundaries in the (x,y) unit square?
Why does interpreting cos²(θ) as a projection length (instead of an area) make it possible to connect to an inscribed-angle theorem involving 2θ?

Key Points

1
Treat unfamiliar problems by extracting and using the defining features of the setup (e.g., equal radii on a circle) before reaching for advanced tricks.
2
Add structure when needed—often by drawing extra lines—and then name angles/points so relationships become algebraic.
3
Exploit symmetry aggressively: isosceles triangles and projection symmetry can turn hard expressions into simple equalities.
4
Look for a single object that admits two descriptions; equating those descriptions can produce identities that aren’t obvious algebraically.
5
Convert non-visual algebraic conditions (like floor(x/y)=k) into geometric regions in a coordinate plane so probabilities become areas.
6
When summing infinitely many contributions, expect special functions (like ln(2)) to appear through recognizable series/integral patterns.
7
Always verify final answers with reasonableness checks (bounds, quick approximations) and, when feasible, numerical simulation to catch sign or logic slips.

Highlights

The inscribed-angle theorem is derived from circle definitions and isosceles triangles, yielding the reusable identity θ_L = 2θ_S.

A geometric proof of cos²(θ) = 1/2(1 + cos(2θ)) comes from interpreting cosine squared as a projection length and using inscribed angles to introduce 2θ.

The probability that floor(x/y) is even becomes an area-sum over infinitely many triangles in the unit square, leading to the closed form 2 − ln(2).

A deliberate sign/series mistake is caught by a gut check: the answer must exceed 1/2 because floor(x/y)=0 already has probability 1/2.

Monte Carlo simulation with random uniform samples confirms the analytic probability near 0.65.

Topics

Problem Solving Principles
Inscribed Angle Theorem
Geometric Proofs
Double-Angle Identity
Probability via Areas

Mentioned

Ben Eater
Cam Christensen
Eli shared

Tips to be a better problem solver [Last live lecture] | Ep. 10 Lockdown live math