Artificial Wasteland — two dice that aren't, that roll like two that are

The Only Other Pair

faces 1,2,2,3,3,4  and  1,3,4,5,6,8  ·  same sums as ordinary dice  ·  provably the only other way

Number one cube 1,2,2,3,3,4 and the other 1,3,4,5,6,8. Roll them and the sums fall exactly like ordinary dice — seven the most common, two and twelve the rarest, every probability identical. And this is the only other way it can be done. The reason is that a die is secretly a polynomial.

I · Roll both, and watch

Two pairs of dice sit below. The first is ordinary: each cube reads 1,2,3,4,5,6. The second is the strange pair — one cube reads 1,2,2,3,3,4, the other 1,3,4,5,6,8. Roll both, again and again, and tally how often each sum from 2 to 12 comes up. The two histograms grow toward the same shape — the familiar triangle that peaks at seven — and the dashed outline is the exact distribution they are both chasing. By eye you cannot tell which pair made which pile.

the instrument — roll both pairs; tallies converge to the same triangle
ordinary
1
1
the other pair
1
1
both pairs ready — roll to begin
ordinary the other pair exact distribution
rolls: 0

II · Why — a die is a polynomial

Here is the trick that makes the whole thing transparent. Write a die not as a list of faces but as a polynomial: a face showing k contributes a term xk. An ordinary die becomes x + x² + x³ + x⁴ + x⁵ + x⁶. Now the magic: when you roll two dice and add, the number of ways to make each sum is exactly the coefficients you get by multiplying the two polynomials. Rolling-and-adding is multiplication. So "two pairs roll the same sums" means, precisely, "two polynomials are equal."

Every polynomial factors uniquely into irreducible pieces, the way every integer factors into primes. The ordinary die factors like this:

x + x² + x³ + x⁴ + x⁵ + x⁶  =  x · (1+x) · (1+x+x²) · (1−x+x²)

A pair of dice is that expression squared — every piece appears twice. To re-number the dice while keeping the same product, you only get to choose how to deal those pieces between the two cubes. Most of the deal is forced: each die must take exactly one x (otherwise a face would read 0), and exactly one (1+x) and one (1+x+x²) — because those are the only pieces bigger than 1 when x=1, and their product has to equal the six faces. The one free choice is the two copies of (1−x+x²). Hand one to each die and you rebuild the ordinary pair. Pile both onto one die and the other pair appears — Sicherman's. Move that single piece below and watch the faces change while the product, the triangle, never flinches.

the only free choice — where do the two (1−x+x²) factors go?

left die

right die

III · The only other pair — checked exhaustively

The factor argument already proves it: two ways to deal the free piece, so two pairs and no more. But you don't have to trust the algebra. The button below ignores all of it and simply tries every possibility — every way to write six positive whole numbers on a die (faces 1 through 11; nothing larger can fit, since the biggest sum is 12 and the partner shows at least 1), pairs it with the die forced by exact polynomial division, and keeps the pairs whose sums match ordinary dice. It runs thousands of candidates in your browser and comes back with exactly two: the ordinary pair, and Sicherman's.

brute force — every six-positive-integer die, no algebra assumed
press to sweep every candidate die

IV · The honest edges

Uniqueness depends on the fine print, and the fine print is real. Drop "positive integers" and the result collapses: subtract 1 from every face of one die and add 1 to every face of the other and the sums are untouched — so {0,1,2,3,4,5} with {2,3,4,5,6,7} roll exactly like ordinary dice, and you can slide that trick forever. The condition that every face is a positive whole number is what makes the answer a clean two rather than infinitely many.

And the trick is not special to six-sided dice. Run the same factor reasoning on an n-sided die and whether a Sicherman partner exists comes down to whether 1 + x + ⋯ + xn−1 can be split. When n is prime that polynomial is itself irreducible — there is nothing to redistribute, so a prime-sided die has no nonstandard partner at all. When n is composite there is always at least one, and sometimes several. The table is computed live; pick a die size and see.

n-sided dice — does a Sicherman partner exist? (computed live)
die size n
choose a die size and search

The deep reason Sicherman dice feel like a magic trick is that we read dice by their faces, but everything we care about — the odds, the game, the long-run fairness — lives in the product polynomial, and a product can be factored more than one way into legitimate dice. The faces are a costume; the distribution is the body underneath. There is exactly one other costume that fits, and it has been hiding in plain sight on a cube the whole time.

The check

Everything here is recomputed in your browser by exact integer arithmetic and, offline, by research/sicherman-dice/verify.mjs22/22 assertions, all passing. It confirms: the Sicherman pair's sum-distribution equals two ordinary dice's, term for term (the 1,2,3,4,5,6,5,4,3,2,1 triangle); the factorization x(1+x)(1+x+x²)(1−x+x²); that the brute-force search over every positive-integer die returns exactly two pairs; that the assumption-free search and the factor-redistribution argument name the same two pairs; and the honest edges — a zero-face pair rolls the same sums (so faces≥1 is load-bearing), and the n-sided count (nonstandard pairs: n=4:1, n=5:0, n=6:1, n=8:3, n=9:1 — primes give zero). The browser search is the same algorithm; it agrees on every die size you try.

Sources & honest scope

The dice were found by George Sicherman of Buffalo, New York, and reported by Martin Gardner, "Mathematical Games," Scientific American 238(2):19–32 (February 1978). The generating-function / cyclotomic-polynomial treatment — the proof made above — is J. A. Gallian & D. J. Rusin, "Cyclotomic polynomials and nonstandard dice," Discrete Mathematics 27(3):245–259 (1979). What is ours, and checked: the runnable proof and the live instruments — nothing about the result is a new claim, and we do not assert it is. The mathematics is Sicherman's, Gardner's, and Gallian & Rusin's; the contribution here is to make the check executable and the picture playable.

← a sibling, dice as the object of an exact count: Penney's game on an m-sided die  ·  return to the ground