From Amida Buddha to Office Lunch: A Short History of the Ladder Game
The ladder game most Koreans know as 사다리타기 goes by many names: amidakuji (阿弥陀籤) in Japan, ghost leg or 畫鬼腳 in China and Taiwan, and "ladder lottery" in mathematics papers. The Japanese name is the oldest documented one and carries a surprising origin. In the Muromachi period (14th–16th centuries), lots were drawn not with parallel lines but with lines radiating from a center, like the halo drawn behind Amida Buddha in Buddhist iconography — participants paid different hidden amounts written at the center, and the radial "kuji" (lottery) decided who paid what. The halo shape gave the game its name centuries before it straightened into today's ladder of vertical rails and horizontal rungs.
The modern form spread through East Asian schoolyards and offices in the 20th century precisely because it needs nothing but a scrap of paper: draw one vertical line per person, add horizontal rungs at random while the outcomes are hidden, fold the bottom over, and let everyone trace their path. It is a randomizer, a small ritual, and a piece of theater at once — the slow trace down the ladder builds suspense in a way a die roll cannot.
Interestingly, the game barely exists in Western tradition, where drawing straws or flipping coins fills the same niche. That makes amidakuji one of the clearest examples of a "cultural algorithm": a mathematically nontrivial procedure — as the next section shows, it secretly performs permutation composition — preserved and transmitted purely as folk custom.
Why Nobody Ever Shares a Prize: The Bijection Hidden in the Ladder
Play the ladder game a thousand times and you will never see two players land on the same outcome, no matter how carelessly the rungs were drawn. This is not luck — it is a theorem, and the proof fits in a paragraph.
Consider what one horizontal rung does. A path arriving at either of the two rails it connects crosses to the other rail; paths on all other rails pass straight through. In the language of permutations, one rung is a transposition of two adjacent elements — it swaps exactly two positions and touches nothing else. A swap is trivially reversible and one-to-one. A full ladder is just a stack of these swaps applied top to bottom, and a composition of one-to-one maps is itself one-to-one. Therefore any ladder, with any number of rungs in any arrangement, defines a bijection: every player reaches exactly one outcome and every outcome is reached by exactly one player. Collisions are structurally impossible, which is precisely why the game works with the results hidden.
There is an elegant converse, too: every possible assignment of n players to n outcomes — all n! permutations — can be built from adjacent swaps, so a sufficiently rich ladder can realize any of them. (This is the same fact that makes bubble sort correct, since bubble sort only ever swaps neighbors.) The minimum number of rungs needed to realize a given permutation equals its inversion count — the number of pairs that finish in reversed order — with a maximum of n(n−1)/2 for a full reversal. A trace is deterministic chaos: fixed rungs give a fixed answer; the randomness enters only through what you cannot see.
4 rails (A B C D), 3 rungs, traced top to bottom:
A B C D
|---| | | rung 1 swaps A,B -> B A C D
| |---| | rung 2 swaps A,C -> B C A D
| | |---| rung 3 swaps A,D -> B C D A
1 2 3 4
A->4 B->1 C->2 D->3 (a bijection: no outcome repeats)
Each rung = one adjacent swap (transposition).
Any of the n! assignments is reachable with enough rungs;
minimum rungs for a permutation = its inversion count
(full reversal of n rails needs n(n-1)/2 rungs).Is a Hand-Drawn Ladder Actually Fair? Rung Count and Probability Bias
Here is the uncomfortable truth about paper ladders: a bijection is not the same thing as a fair lottery. The ladder guarantees everyone gets a different outcome, but it says nothing about all outcomes being equally likely — and with few rungs, they are not even close.
The extreme case makes it obvious. Zero rungs means player 1 gets outcome 1 with certainty. Add one rung and only two players' fates change. In general, a player's path can migrate at most one rail per rung it meets, so reaching an outcome k rails away requires at least k well-placed rungs. Simulations and analysis of ladder lotteries show that with a typical hand-drawn ladder (roughly 1.5 to 2 rungs per rail), each player's distribution stays strongly peaked near their own starting column — you are far more likely to land directly below where you started than at the far end. Studies of randomly-placed-rung ladders find the probability matrix approaches uniform only slowly as rungs grow; getting close to truly equal odds takes on the order of n² or more rungs for n players, far beyond what anyone draws by hand.
Paper players exploit this, consciously or not: picking the column directly above a desirable visible outcome is a real edge on a sparse ladder. The countermeasures are the folk fixes everyone half-knows — hide the outcomes before columns are chosen, let each participant add extra rungs after choosing, or shuffle which outcome label goes where. This digital version cuts the knot differently: it does not simulate rungs at all, but directly shuffles the outcome list with Math.random(), giving every player a genuine 1/n chance at every outcome — the distribution a hand ladder only approximates.
4 players, rungs drawn between random neighbors.
Rough chance that player 1 ends at their own column:
0 rungs -> 100%
2 rungs -> ~55%
5 rungs -> ~35%
many -> approaches 25% (= 1/4, the fair value)
Sparse ladders favor "straight down".
Fair play checklist for paper ladders:
1. hide outcomes BEFORE columns are picked
2. everyone may add rungs after picking
3. shuffle the outcome labels too
This tool skips the ladder: it shuffles outcomes directly,
so every assignment is 1/n from the start.Ladders, Braids, and Bubble Sort: Where Amidakuji Shows Up in Science
A children's lottery turns out to be a respectable mathematical object. Combinatorists study "ladder lotteries" as a structure in their own right: a classic result identifies optimal ladders — those realizing a permutation with the fewest possible rungs — with reduced decompositions of the permutation into adjacent transpositions, and researchers have counted and enumerated them and built algorithms that list every optimal ladder for the reverse permutation. The number of rungs you need is governed by the inversion count mentioned earlier, and sliding a rung past its neighbors corresponds to the braid relations from knot theory: amidakuji diagrams are essentially braid diagrams with the over/under crossing information forgotten.
The same picture appears across disciplines. In physics, the world lines of particles that scatter off each other trace an amidakuji pattern through spacetime. In sorting theory, a bubble sort execution is literally a ladder: each comparison-and-swap of neighbors is one rung, and sorting networks used in parallel hardware are ladders with rungs arranged so any input permutation exits sorted. In music theory, voice-leading diagrams that track how chord tones move between voices form the same crossing diagrams.
For a puzzle-flavored taste of the mathematics, try this with pen and paper: draw a 3-rail ladder that sends A to 3, B to 1, and C to 2. You will find it needs exactly two rungs — that permutation is a 3-cycle with two inversions — and no arrangement with a single rung can do it. Then try swapping only A and C while leaving B fixed: the minimum is three rungs, because the middle rail must be crossed twice.
Puzzle 1: send A->3, B->1, C->2 (needs exactly 2 rungs)
A B C
|---| | swap A,B -> B A C
| |---| swap A,C -> B C A
1 2 3 A->3, B->1, C->2 done
Puzzle 2: swap A and C, keep B fixed (minimum 3 rungs)
A B C
|---| | -> B A C
| |---| -> B C A
|---| | -> C B A
1 2 3 A->3, C->1, B->2 ... B is back home
Fewer than 3 is impossible: A and C must each
cross the middle rail, and every crossing is a rung.Running a Good Draw: Etiquette, Group Psychology, and When Not to Use a Ladder
Random selection is social technology. The reason teams reach for a ladder instead of letting the manager assign the unpleasant task is procedural fairness: research on how people judge decisions consistently finds that outcomes feel more acceptable when the process is visibly neutral, even to the person who loses. A ladder draw is neutral, transparent, and — crucially — theatrical enough that everyone witnesses the same result at the same moment. That shared moment is what kills the post-hoc grumbling a private coin flip invites.
A few practices keep digital draws clean. Decide the stakes before pressing start, and write them into the result labels themselves ("buys coffee", "presents first") so the outcome is unarguable. Agree in advance that the first run is final — re-rolling "best of three" after seeing a result is the classic way fairness dies. For recurring duties like cleaning rotations, consider excluding last week's loser from this week's worst slot; a memoryless 1/n draw will occasionally hand the same person the chore three weeks running, which is statistically normal (probability 1/n² for two repeats, not astronomical for small groups) but socially corrosive.
Know the tool's limits, too. A ladder assigns n people to n distinct outcomes — a permutation. It is the wrong shape when you need one winner from many (use a single "win" label among blanks, which works but wastes the structure), weighted odds (someone should carry double chance), or team splitting into unequal groups. And for anything with real money or legal weight, folk randomizers are inappropriate altogether: regulated lotteries exist precisely because stakes change how much fairness must be provable, not just felt.