Calculating Pi with Skittles (and census data!?) - Numberphile

Key Concepts

• Monte Carlo Method: A computational algorithm that uses repeated random sampling to obtain numerical results, in this case, estimating the value of $\pi$.
• Geometric Probability: The principle that the ratio of the area of a circle to the area of its circumscribed square is $\pi/4$.
• Uniform Distribution: The assumption that data points (or objects) are spread evenly across a space, which is necessary for an accurate Monte Carlo estimation.
• Population Density/Clumping: The tendency of human populations to cluster in urban centers, which introduces bias and reduces the accuracy of random sampling models.

1. The Skittles Experiment (Methodology)

The presenters demonstrate a physical application of the Monte Carlo method to estimate $\pi$:

• Setup: A circle is drawn on a surface, and a square is drawn around it such that the circle fits perfectly inside.
• Process: A handful of Skittles are dropped randomly onto the shape.
• Calculation:
  • Count the number of Skittles inside the circle ($N_c$) and the total number of Skittles inside the square ($N_t$).
  • The ratio $N_c / N_t$ approximates the ratio of the areas: $\text{Area of Circle} / \text{Area of Square} = (\pi r^2) / (2r)^2 = \pi/4$.
  • To find $\pi$, multiply the ratio by 4: $\pi \approx 4 \times (N_c / N_t)$.
• Results: Using 562 Skittles inside and 159 outside (Total: 721), the calculation yielded an estimate of approximately 3.11, which the presenters noted was "not bad" but affected by "dodgy edge ones" (Skittles landing on the boundary line).

2. Scaling Up: Population Data Analysis

The presenters transition from physical objects to demographic data to test if the same principle applies to human geography using UK Office of National Statistics data.

• Framework: They select a GPS coordinate, define a radius (e.g., 100 km), and compare the population within that circle to the population within the circumscribed square.
• Case Studies:
  • Nottingham (100 km radius): Population inside: 16.2 million; Square: 20.389 million. Result: 3.18.
  • Anfield, Liverpool (100 km radius): Result: 3.27. The inaccuracy is attributed to the proximity to the coast, where the "empty" sea space disrupts the uniform distribution assumption.
  • Buckingham Palace (10 km radius): Result: 3.28. The smaller radius and high urban density ("spikiness of the data") led to a poor estimate.
  • Center of England (Leicester): Result: 3.049. Despite being a more central location, it still failed to outperform the random distribution of the Skittles experiment.

3. Key Arguments and Observations

• The Importance of Randomness: The presenters argue that the accuracy of the Monte Carlo method is entirely dependent on the uniformity of the distribution. Because human populations are inherently "clumpy" (concentrated in cities), they make for poor subjects for this specific mathematical estimation.
• The "Edge" Problem: In both the physical and digital experiments, boundary conditions (Skittles on the line or coastal geography) introduce significant error.
• Synthesis: The experiment highlights that while the mathematical theory ($\pi/4$ ratio) is sound, its practical application is highly sensitive to the nature of the data being sampled. The presenters conclude that the simple, random distribution of Skittles actually provided a more accurate estimate of $\pi$ than the complex, non-uniform distribution of the UK population.

4. Notable Quotes

• "If I imagine that this has got unit radius... the area of the circle is $\pi r^2$... the square has got area four. So the ratio between the area of the circle and the area of the square is $\pi/4$." — Explaining the mathematical foundation.
• "People tend to clump where they live, don't they? In cities... I suspect we're not going to get a very good estimate of $\pi$." — Identifying the limitation of using population data for random sampling.

Conclusion

The video serves as a practical demonstration of the Monte Carlo method. It illustrates that while $\pi$ can be derived through geometric probability, the accuracy of the result is dictated by the randomness of the sample. The experiment successfully proves that physical, random distribution (Skittles) is a more effective tool for this estimation than demographic data, which is skewed by human settlement patterns and geographical features like coastlines.