What the figures prove
A dissection whose pieces overlap by two pixels looks exactly as convincing as one that works. A graph captioned as having an Eulerian circuit looks exactly like one that has an Eulerian circuit. The reader cannot tell, and for two phases neither could this site — a figure here carried a printed vertex degree that was simply wrong, through every gate, from the day it was written.
So the assertions live in lib/verify.js and the generators call them
while drawing. A dissection must tile its target with no overlap, no gap and nothing
outside it; congruent pieces must really be congruent; a quoted number must be computed from
the drawing rather than typed beside it. A figure that does not prove its caption throws, and
the build stops.
50 of 50 generator families reach an assertion. Between them they make 462 distinct kinds of claim, tested 775 times in the build that produced this page.
Every line below was collected by running each family at the parameters the essays actually use and recording what it asserted. A family that stops checking something loses the line on the next build; a claim whose wording changes changes here with it. That is the only way this page is worth anything: a hand-written list of what code does drifts in exactly one direction, which is towards claiming more than exists.
What it does not say: that the claim is the right claim. An assertion compares a
drawing against the arithmetic that produced it, and when a generator is handed an argument
outside the range its picture means, both are wrong in the same way and they agree.
scripts/rangecheck.mjs is the check for that, and it still names families that
would accept one.
Geometry
10 families
circle-angle
17 kinds of claim · 7 placements
- A lies on the circle
- and so do the other two
- and the same on the other side
- B lies on the circle
- every apex on the arc gives the same angle
- one angle, not several
- opposite angles add to a straight angle
- the angle on a diameter is a right angle
- the apex is on the major arc
- the apex lies on the circle
- the centre's angle is twice the apex's
- the chord is a diameter
- the exterior angle is twice the base angle
- the quadrilateral's angles add to 360°
- the two halves are the whole angle
- triangle OPA is isosceles
- triangle OPB is isosceles
conic
4 kinds of claim · 6 placements
- a circle needs a cut that produces one
- a ellipse needs a cut that produces one
- a hyperbola needs a cut that produces one
- a parabola needs a cut that produces one
euclid
4 kinds of claim · 12 placements
- the smallest square has side gcd(34, 13) ×4
- the squares tile the 34 by 13 rectangle ×4
- the leftover strip is still there after every pass
- the squares peeled at each pass are the continued fraction of the ratio
figurate
2 kinds of claim · 7 placements
- the first 6 odd numbers sum to 6² ×4
- the L-shaped shells fill the square
golden
7 kinds of claim · 5 placements
- one square per Fibonacci number
- the convergents alternate about φ
- the last convergent is close to φ
- the long side is the next Fibonacci number
- the rectangle is nearly golden
- the short side is the one before it
- the whirling squares tile their rectangle
polyhedron
8 kinds of claim · 11 placements
- 3 triangles at a vertex leave a gap ×3
- 3 hexagons at a vertex cannot leave a gap
- 3 pentagons at a vertex leave a gap
- 3 squares at a vertex leave a gap
- 4 pentagons at a vertex cannot leave a gap
- 4 squares at a vertex cannot leave a gap
- 6 triangles at a vertex cannot leave a gap
- exactly five vertex figures close up
pythagoras
15 kinds of claim · 15 placements
- and half the rectangle — same base, apex on the parallel through the altitude
- four triangles and the tilted square fill the left square
- four triangles and the two upright squares fill the right square
- so the square and the rectangle are equal
- the altitude cuts the big square into exactly these two rectangles
- the angle at C is a right angle
- the eight triangles are all the same
- the long leg is b
- the short leg is a
- the square on the hypotenuse is c²
- the square on the leg has the leg's area squared
- the tilted square equals the two upright squares
- the tilted square is c²
- the triangle is half the square — same base, apex on the opposite side
- the two triangles are the same triangle turned
reuleaux
7 kinds of claim · 6 placements
- the 3-sided Reuleaux curve has one width at every angle ×3
- a Reuleaux polygon needs an odd number of sides
- all the shapes have one width
- every arc bulges away from the centre
- ground and plank stay one width apart
- the width does not depend on how far it has rolled
- the width is the arc radius
unroll
2 kinds of claim · 6 placements
- the rings account for the whole disc
- the triangle of base 2πr and height r has the disc's area
voronoi
6 kinds of claim · 7 placements
- the centre of cell 0 is nearest to site 0 ×20
- every site keeps a cell
- no triangle's circumcircle holds another site
- the bisector is where the two distances tie
- the cells account for the whole box, with nothing double-covered
- the triangulation is not empty
Analysis
7 families
circle-to-sine
3 kinds of claim · 7 placements
- the point is on the circle
- the point's height is the wave's height there
- the two coordinates are a point of the unit circle
exponential
9 kinds of claim · 9 placements
- a base that is not e does not have slope equal to height
- and at base e the slope is the height
- and the gap left is smaller than the last term added
- and the limit of the staircase is e itself
- every partial sum is below e
- more splits is closer to e
- no number of splits reaches e
- the table shown gets e right to three decimals
- the tangent drawn is the curve's own slope
fourier
3 kinds of claim · 7 placements
- every term count is a whole number between 1 and 200
- more terms is closer, away from the jump
- the curve drawn is the sum of the harmonics
harmonic
16 kinds of claim · 13 placements
- block 1 reaches a half ×4
- consecutive partial sums differ by the next term
- each block leans further out than the one below it
- no term is used twice
- the answer is caught between an even and an odd partial sum
- the bound on each block is exactly a half
- the drawn lean is the sum of the drawn steps
- the drawn running total is the partial sum
- the error is smaller than the last term used
- the gap to ln n has settled on γ
- the last term is small
- the plotted partial sums are the series
- the rearranged total settles near the number it was aimed at
- the signs are the ones the series has
- the target is visibly away from the sum the same terms give in order
- the total is not small
riemann
9 kinds of claim · 13 placements
- each panel is at least as close as the one before it
- more bars means less error
- no width, no area
- the accumulation only ever climbs, because f is positive
- the function is one the figure knows
- the panels really are in increasing order of rectangles
- the printed total is the area of the bars drawn
- the quoted exact area is the area under this curve
- the slope of the accumulated area is the height of the curve
secant
6 kinds of claim · 6 placements
- each secant is closer than the last
- every secant from the left has slope −1
- every secant from the right has slope 1
- the function is one the figure knows
- the last secant is near the tangent
- the two sides disagree, so there is no limit
taylor
5 kinds of claim · 5 placements
- the degree-1 sum matches to order 1 at the expansion point ×4
- every term count is a whole number between 1 and 40
- inside the radius, more terms help
- outside it, more terms hurt
- the function is one the figure knows
Algebra
4 families
algebra-tiles
14 kinds of claim · 8 placements
- and the reassembled rectangle has the same area
- the four tiles are the square
- the larger length is larger
- the length really solves the equation
- the missing corner is (b/2)²
- the negative root solves it too
- the picture is about positive lengths
- the positive root is the one a length can be
- the square tile is x²
- the strips are bx however they are cut
- the three pieces and the corner make a square of side x + b/2
- the two pieces are what is left of a² after b² is removed
- which is completing the square, as arithmetic
- which is the identity
complex-turn
7 kinds of claim · 7 placements
- root 0 raised to the 5 is 1 ×17
- and has no imaginary part left
- and the angles add
- every root is on the unit circle
- in both coordinates
- the lengths multiply
- the roots sum to zero
linear-map
3 kinds of claim · 15 placements
- a real 2×2 map has two eigenvalues counted with sign
- Av is λv in the first coordinate
- Av is λv in the second
projection
5 kinds of claim · 5 placements
- a·b is |a||b| cos θ
- a·b is the shadow's length times |b|
- the dropped line meets b at a right angle
- the sign of a·b follows the angle
- the vector projected onto is not zero
Discrete
8 families
clock
19 kinds of claim · 17 placements
- the walk reaches m × n / gcd cells on a 3 by 5 grid ×3
- a dial needs at least three positions
- a row is a permutation exactly when its multiplier is coprime to the modulus
- every non-zero row is a permutation exactly when the modulus is prime
- every residue has a place on the dial
- every row of the addition table is a permutation
- it fills the whole grid exactly when the moduli are coprime
- multiplying `order` times returns to 1
- the arithmetic and the walk agree
- the drawn entry is the arithmetic
- the drawn walk passes the top as often as the division says
- the first modulus is between 2 and 12
- the multiplier is not zero
- the orbit visits each residue once before closing
- the order divides m − 1, which is Fermat's little theorem
- the period is the least common multiple
- the second modulus is between 2 and 12
- the walk is between one and 144 steps
- the walk runs at least one full period
complete-graph
10 kinds of claim · 5 placements
- a vertex meets five others
- and in fact at least two, which is Goodman's bound
- EVERY colouring of six people contains a monochromatic triangle
- five edges in the pentagon
- K5 has ten edges
- K6 has fifteen edges
- no triangle has all three edges the same colour
- the three ends make three pairs, and any one of them closes a trio
- this colouring has a monochromatic triangle
- two colours over five edges forces three of one
euler-path
15 kinds of claim · 7 placements
- a closed circuit needs every degree even
- all four landmasses have odd degree
- an even degree pairs every arrival with a departure
- an odd degree leaves exactly one edge unpaired
- an open walk needs exactly two odd degrees
- every vertex drawn is a vertex used
- Königsberg has seven bridges
- landmass E has degree 3
- landmass I has degree 5
- landmass N has degree 3
- landmass S has degree 3
- no Euler walk exists
- the degrees add to twice the number of edges
- the even panel really shows an even degree
- the odd panel really shows an odd degree
map-colour
4 kinds of claim · 6 placements
- an odd rim needs four and an even rim three
- four colours are enough
- no two neighbouring regions share a colour
- the map can be coloured with at most five
number-spiral
8 kinds of claim · 7 placements
- a marked square holds a prime, and an unmarked one does not
- and 41² is where it fails
- consecutive integers are neighbours on the spiral
- Euler's polynomial is prime for forty values in a row and then is not
- no two integers land on the same square
- the first polynomial really is the prime-richer one over the range drawn
- the sieve starts where the primes do
- there are primes to compare against
pascal
3 kinds of claim · 10 placements
- row 0, entry 0 of Pascal's triangle ×45
- the row count is a whole number between 1 and 64
- the subdivision depth is a whole number between 1 and 8
pigeonhole
6 kinds of claim · 9 placements
- each case really does overflow its boxes
- every item is in a box
- more items than boxes forces a box with two
- the fullest box holds ⌈items/holes⌉
- there are at least as many things as boxes
- there is at least one box
triangulation
3 kinds of claim · 5 placements
- C(0) two ways ×9
- a 6-gon has C(4) triangulations ×4
- each cut uses n−3 diagonals
Topology
6 families
euler-solid
14 kinds of claim · 3 placements
- and its twelve edges
- every edge of the cube is shared by two faces
- every edge of the dodecahedron is shared by two faces
- every edge of the icosahedron is shared by two faces
- every edge of the octahedron is shared by two faces
- every edge of the tetrahedron is shared by two faces
- the flattened cube keeps its eight corners
- the flattened cube still gives 2
- the solid is one figure-kit knows
- V − E + F is 2 for the cube
- V − E + F is 2 for the dodecahedron
- V − E + F is 2 for the icosahedron
- V − E + F is 2 for the octahedron
- V − E + F is 2 for the tetrahedron
fixed-point
19 kinds of claim · 10 placements
- and has no direction to choose there
- and it lies inside the disc
- and sits on it
- at the left end the map moves the point right or not at all
- at the right end it moves it left or not at all
- dipole has the index it claims
- every arrow is tangent to the sphere
- every orbit runs into the fixed point
- saddle has the index it claims
- source has the index it claims
- the field is one the figure knows
- the field vanishes at the pole
- the fixed point is unique
- the map is one the figure knows
- the map keeps the interval inside itself
- the map sends the disc into itself
- the point found really is fixed
- the solved point is fixed, first coordinate
- the solved point is fixed, second
knot
11 kinds of claim · 6 placements
- a diagram with k crossings is cut into k arcs
- at every crossing the three arcs are all alike or all different
- the colouring really uses more than one colour
- the figure-eight knot has the crossings it claims
- the knot is one the figure knows
- the table is ordered by crossing number
- the trefoil has the crossings it claims
- the trefoil is three arcs
- the unknot has the crossings it claims
- the unknot is a single arc
- there are three moves, and Reidemeister proved there are no others
mobius
3 kinds of claim · 5 placements
- carrying the surface round once brings it back on the other face
- the band is drawn at a whole number of half-twists
- the gluing arrow matches the twist the figure is about
mobius-cut
3 kinds of claim · 2 placements
- and one lap leaves one loop while two laps leave two
- the cut is inside the band
- the scissors take one lap down the middle and two off-centre
stereographic
6 kinds of claim · 8 placements
- the image of 0 lies on the circle ×14
- the ray to 0 passes through the pole ×14
- the image of -2 lies on the circle ×13
- the ray to -2 passes through the pole ×13
- the image is on the ray from the pole through the point
- the image lands on the plane
Probability
5 families
bayes
14 kinds of claim · 12 placements
- and at three doors it is two thirds
- everybody has it, so everybody positive has it
- nobody has it, so nobody who tests positive has it
- one of the two strategies wins, and only one
- staying wins as often as the first pick was right
- switching gets better with more doors
- switching wins two thirds against a host who knows, and half against one who does not
- the answer rises with the base rate
- the four cells fill the square
- the game needs at least three doors
- the ignorant host's story throws two of the six worlds away
- the ignorant-host variant is only drawn at three doors, where switching has one meaning
- the shaded fraction is Bayes' theorem
- the two strategies exhaust the possibilities
birthday
5 kinds of claim · 4 placements
- 22 is still under a half
- 23 is over it
- one person shares with nobody
- the half-way point is 23 people
- the pair count is k choose 2
buffon
7 kinds of claim · 7 placements
- at least one needle is dropped
- every needle claimed was drawn
- so a needle crosses with probability 2/π
- the area under half a sine wave is 1
- the box has area π/2
- the estimate is in a plausible range for pi
- the printed estimate is 2Ln / dc
galton
6 kinds of claim · 8 placements
- a path bounces once per row
- and are centred under the funnel
- and ends in a bin the board has
- the exact bin shares add to one
- the path ends over the bin its right-turns name
- the row count is a whole number between 1 and 24
walk
15 kinds of claim · 6 placements
- a walk in space escapes, and the simulation sees it
- a walk on a line comes back, and the horizon barely hides it
- a walk on a plane comes back too, but more slowly
- an endpoint has the parity of the step count
- and their spread is the square root of the number of steps
- each step moves one place along one axis
- every step is drawn
- every step is one place, either way
- every walk is counted once
- some walk goes outside the envelope, because √n is typical and not maximal
- the position has the parity of the step count
- the quoted constant is Pólya's
- the spread of the endpoints is of the order of the square root of the steps
- the walk comes back to where it started at least once
- the walks are centred on where they started
Number
10 families
approx
10 kinds of claim · 4 placements
- |qα − p| comes out below 1/N
- and every convergent from the first onward breaks a record
- every record-breaking fraction is a convergent
- neighbouring convergents differ by a determinant of one
- so p/q is within one over qN of the number itself
- the constant is one this figure knows
- the largest denominator is a whole number between 8 and 200
- the number of boxes is a whole number between 3 and 14
- there are record-breaking denominators to mark
- two of the N + 1 points share a box, as they must
continued
18 kinds of claim · 9 placements
- the quotients rebuild 34/13 ×3
- the number of terms is a whole number between 2 and 12 ×2
- √2 has a convergent above √5, as Hurwitz says every irrational must
- and the last of them has closed on √5
- between two and five constants
- e has a convergent above √5, as Hurwitz says every irrational must
- each convergent is closer than the one before
- neighbouring convergents differ by a determinant of one
- the constant is one this figure knows
- the convergents alternate above and below the value
- the denominator is a whole number between 1 and 10000000
- the golden ratio's convergents never leave the neighbourhood of √5
- the numerator is a whole number between 1 and 10000000
- the tower's last convergent is a number
- there are enough convergents to draw
- π has a convergent above √5, as Hurwitz says every irrational must
- π has a convergent that beats √5 by two orders of magnitude
- φ has a convergent above √5, as Hurwitz says every irrational must
descent
11 kinds of claim · 4 placements
- the side of the big square is a whole number between 2 and 200 ×2
- the side of the small squares is a whole number between 1 and 200 ×2
- and the discrepancy never changes size
- and the new pair is strictly smaller
- every step is strictly smaller than the one before
- the descent carries the discrepancy to minus itself
- the descent runs out of room before the discrepancy runs out
- the descent takes at least three steps
- the pair is one the descent can start from
- the two small squares overlap and still reach the corners
- the two small squares, less their overlap and plus the corners, are the big one
factor
28 kinds of claim · 19 placements
- 6 is called what it is ×7
- the bar for 6 is its divisors laid end to end ×7
- the lattice draws every divisor of 30 exactly once ×4
- 496 is perfect ×2
- the number is a whole number between 4 and 100000 ×2
- and so do the norms of two and three
- between one and eight whole numbers above one
- both first splits divide the number
- both trees end in the same multiset of primes
- every divisor including the number itself adds to twice the number
- no divisor is drawn twice
- no element of this ring has norm two or three, so neither factor can split further
- the cells add up to the divisor sum
- the divisor count is the product of one more than each exponent
- the divisor sum factorises as the two row totals multiplied
- the exponent is a whole number between 2 and 7
- the first tree's leaves multiply back to the number
- the lattice half-width is a whole number between 2 and 5
- the Mersenne number for this exponent is prime
- the number has a factor tree to draw
- the number has at least three prime factors, so its trees can differ
- the number has at most three distinct primes
- the number is a product of exactly two prime powers
- the powers of two below 2^k add up to the Mersenne number
- the rectangle holds every divisor
- the second tree's leaves multiply back to the number
- the two conjugate factors have norms multiplying to 36
- the two trees start differently
ferrers
9 kinds of claim · 7 placements
- the partitions of 8 into odd parts and into distinct parts come out equal ×2
- and it is the largest one that fits
- every listed partition adds to the number it partitions
- the conjugate is a partition of the same number
- the number being partitioned is a whole number between 3 and 12
- the parts are a descending list of whole numbers
- the square, the arm and the leg account for every dot
- there is a Durfee square to draw
- turning the diagram over twice gives it back
lattice-circle
11 kinds of claim · 10 placements
- 3, 4, 5 is a Pythagorean triple ×10
- 5 = 2² + 1² ×5
- the points on the circle of radius √25 are 4(d₁ − d₃) ×5
- a prime is a sum of two squares exactly when it is one more than a multiple of four
- and the point it came from is on the unit circle
- every marked point really is on the circle
- no two slopes give the same triple
- the largest prime to test is a whole number between 10 and 200
- the lattice half-width is a whole number between 1 and 16
- the number is a whole number between 1 and 200
- the slopes are proper positive fractions
mediant
11 kinds of claim · 7 placements
- the circles on 0/1 and 1/7 touch ×62
- 0/1 and 1/7 are Farey neighbours ×24
- 1/1 arrives already in lowest terms ×15
- the sequence has one entry per coprime pair up to 7 ×2
- circles that are not neighbours stay clear of each other
- no fraction is produced twice
- reading the drawing left to right reads the fractions in increasing order
- the depth is a whole number between 2 and 5
- the drawing scale is a whole number between 200 and 400
- the order is a whole number between 2 and 12
- the tree is full to its stated depth
necklace
13 kinds of claim · 4 placements
- the length is a whole number between 3 and 7 ×2
- the number of colours is a whole number between 2 and 4 ×2
- and exactly a of them are single
- and the number of non-constant classes is (a^p − a)/p
- every class is a single string or a full ring of p
- every other class holds exactly p strings
- every string of the given length is drawn
- p divides a^p − a, which is what the count just showed
- the classes account for every string
- the classes of size one are exactly the constant strings
- the figure draws every string, so there is a ceiling on how many
- the length is prime
- the length is prime — the whole argument needs it to be
reciprocity
16 kinds of claim · 4 placements
- (p | q) comes out of the count above it
- (q | p) comes out of the count below the line
- 3 folds gives the same answer as Euler's criterion
- and every one of them lands in the bottom half
- and their product is minus one to the power of the whole rectangle
- both are odd primes and they are different
- every point is counted exactly once, on one side or the other
- no lattice point lies on the diagonal, because the primes are coprime
- the first prime is a whole number between 3 and 31
- the folded values are all different
- the modulus is a whole number between 5 and 31
- the modulus is an odd prime
- the multiplier is a whole number between 2 and 30
- the multiplier is not a multiple of the modulus
- the points below the line are the sum of the floors of kq/p
- the second prime is a whole number between 3 and 31
sieve
12 kinds of claim · 11 placements
- 211 leaves remainder 1 on division by 2 ×10
- the limit is a whole number between 10 and 400 ×2
- the staircase ends at π(1000) ×2
- a square left unstruck holds a prime, and a struck one does not
- every prime factor of the constructed number is outside the list
- one gap between each consecutive pair
- the column count is a whole number between 4 and 25
- the count runs above x / ln x over this range
- the gaps add up to the distance from 2 to the last prime
- the list is between two and six primes
- the pass number is between 0 and 8
- the squares left standing below 100 are the primes