Five small machines built from the mathematics which I found interesting about including
simulations that run entirely in your browser, hand-written in plain
JavaScript with no libraries. Throw darts at π, watch the bell curve
assemble itself, argue with a game-show host, calibrate your eye against
Pearson's r, and set Conway's cells loose.
Fig. 1
Monte Carlo Estimation of π
Scatter random points on the unit square. The fraction landing inside the
quarter-circle approaches π/4 — so four times that fraction approaches π.
Estimate of π
—
error —
Points
0
Inside
0
π̂ over time
The estimator converges at rate O(n−½): each extra digit of π costs a
hundred times more points. In higher dimensions this gets stranger —
see The Curse of Dimensionality.
Fig. 2
The Galton Board
Each ball takes eleven fair coin-flips on the way down and lands in the bin
that counts its rights. The histogram is binomial(11, ½) — and the
silhouette of the central limit theorem appears.
Balls dropped
0
Mean bin · expected
— · 5.50
Eleven pegs, eleven coin-flips. The dashed curve is the expected
binomial(11, ½) count.
Francis Galton built the original in 1873 to show that order emerges from
pure chance. With enough balls, the binomial is indistinguishable from a Gaussian.
Fig. 3
The Monty Hall Problem
Three doors, one car, one stubborn intuition. Conditional probability says
you should always switch — play a few rounds, then let the simulator keep
the score honest.
Pick a door. One hides the car; two hide goats.
Running tally
Games
Won
Obs.
Theory
Stay
0
0
—
33.3%
Switch
0
0
—
66.7%
Your first pick is right ⅓ of the time — so switching wins the other ⅔.
When Marilyn vos Savant published the correct answer in 1990, around ten
thousand readers — including statisticians — wrote in to tell her she was wrong. She was not.
Fig. 4
Guess the Correlation
Sixty draws from a bivariate normal distribution. Read the cloud and
estimate Pearson's r by eye — calibrating intuition against the
statistic is harder than it looks.
Your guess for r
−10.00+1
Read the cloud, then commit.
Round
1
Score
0
Mean error
—
Best round
—
Anscombe's quartet is the standing warning: one value of r can hide very
different clouds. Always plot the data.
Fig. 5
Conway's Game of Life
A cellular automaton with four rules and no randomness, yet rich enough to
be Turing-complete. Draw on the grid with your pointer, then press Run.
Generation
0
Population
0
The preloaded R-pentomino — five cells — runs for 1,103 generations before
settling. Gosper's glider gun fires forever.
Every machine on this page is hand-written vanilla JavaScript on
<canvas> —
no libraries, no frameworks. View source to see the workings.