What Is the Mathematics of Juggling? A Notation That Found New Patterns
By the BrainSnail editorial team. How these articles are written and checked, and how to tell us when one is wrong.
In the mid 1980s several people independently invented a way of writing down juggling patterns as sequences of numbers. The notation turned out to be more than a record-keeping device: it made it possible to enumerate every valid pattern, to prove which sequences are possible, and to discover patterns that no juggler had ever thrown because nobody had thought to look for them.
What the numbers mean
The notation, called siteswap, describes a pattern by the sequence of throws, with each number giving how many beats later that object will next be thrown. A throw of three means the ball comes back to be thrown again three beats later, and a steady sequence of threes is the ordinary three-ball cascade. The key assumptions are that throws alternate between hands and that exactly one throw happens per beat. From that, several things follow immediately and are genuinely useful:
- •The average of the digits in a sequence equals the number of objects being juggled, so 531 uses three balls because the digits average three
- •Even numbers are thrown and caught by the same hand, and odd numbers cross to the other hand, which follows from the alternation
- •A zero means an empty beat with no ball in that hand, and a two means a ball held briefly rather than thrown
- •Higher numbers mean higher throws, since a ball that must stay in the air longer must be thrown further up
- •A sequence is valid only if no two balls are scheduled to land on the same beat, which is a precise condition that can be checked
- •Not every sequence with the right average is valid, which is exactly why the mathematics is interesting rather than trivial
The validity test
The condition for a sequence to be juggleable has a clean formulation. Take the sequence, add to each digit its position in the sequence, and reduce each result modulo the sequence length. The sequence is valid precisely when those results are all different, meaning they form a permutation of the positions. That works because the calculation says which beat each throw lands on, and two balls landing on the same beat is a collision. Applying it to 531 gives results that are all distinct, confirming a valid three-ball pattern, while 521 fails because two throws collide. This test converts a physical question into an arithmetic one that can be checked by hand in seconds or by computer exhaustively, which is what made enumeration possible: the number of valid patterns of a given length with a given number of objects can be counted exactly, and the counting formula turns out to be surprisingly elegant, which has connected juggling to work in combinatorics on permutations and on the affine Weyl group.
What it produced
The practical consequence was new tricks. Enumerating valid sequences generated patterns that were mathematically sound and had never been performed, and jugglers worked through the lists and learned them, with 441, 531, 97531 and the five-ball pattern 744 among those that entered the standard repertoire from the notation rather than from tradition. The mathematics also clarified relationships between patterns that jugglers had treated as unrelated, showing which are transformations of each other and how to move between them, which changed how the skill is taught. Software followed quickly, with juggling simulators that animate any valid sequence and let a performer see a pattern before attempting it. The notation has been extended to cover multiplex throws, where more than one object leaves a hand at once, synchronous patterns where both hands throw simultaneously, and passing patterns between several jugglers, each extension requiring its own validity condition.
The limits of the model
Siteswap describes timing and nothing else, which is both its power and its boundary. It says nothing about where in space a throw goes, so a pattern thrown in a column, in a cascade and behind the back are identical in the notation, and the entire visual character of a performance is outside the model. It says nothing about the objects, so balls, clubs and rings are equivalent even though clubs add rotation that must be timed separately. It assumes perfectly regular beats, which real juggling only approximates, and it assumes exactly one throw per beat unless extended. It also says nothing about difficulty, since a valid pattern may be physically impossible for a human because the throws are too high, too fast or require hands to be in two places. None of these is a defect: the model captures one dimension cleanly, which is what allows it to prove things, and the parts it omits are the parts that make juggling a performance rather than a computation.
The takeaway
Siteswap notation writes a juggling pattern as a sequence of numbers giving how many beats until each object is thrown again, so the digits average to the number of objects, odd numbers cross hands and even ones do not. A sequence is valid exactly when adding each digit to its position and reducing modulo the length produces all different values, which means no two balls land together. Enumerating valid sequences produced patterns nobody had thrown, several of which entered the standard repertoire.