DOI: https://doi.org/10.5281/zenodo.21947412
Canonical: https://thonly.org/research/the-sport-that-says-your-name · Licence: CC0 1.0
Scope and claims division (2026-07-23, first publication): this paper is the first of three published the same day. The present paper owns the sport — the caller mechanism and the rule architecture of sey, the sport the B-Sey℠ app and B-Sey™ ball implement. Its companion The Wager That Isn't owns the challenge/stake/pot grammar (B-Match℠) that nests on this sport's net mode; the third, The Game That Graduates You, owns the third-party game-client doctrine (the witness API). All three share one organizing principle, coined and freed in the present paper: the play/currency wall — competition and play may generate the story; only witnessed giving moves the currency. Earlier corpus publications are cited, not re-claimed: The Quest That Mints Nothing owns the mints-nothing quest class; Proof of Coordinate owns the individuation primitive the ball's chronicle uses. At first publication the sport was specified and unbuilt, and every claim was architectural; a reference implementation has since been built (§§4.3–4.4 record two defects it surfaced). Prior art for the v1 claims dates from 2026-07-23; each later note below dates its own additions.
v2 note (2026-07-25): this revision incorporates the 2026-07-24/25 universalization — the three axes (methods · modes · levels), the call-lead-time ladder with scaffold fading, the three-tier scoring architecture and the implied point, the two speaker formats, the local-first system architecture and the ball's reduced job, the classroom channel and the icebreaker-slot reframe, and place-sponsorship. Claims grow from ten to sixteen and pre-registrations from three to five. The v1 claims are unchanged in substance; §§3.1, 4.1, 5.1, 5.2, 7.1, 9.1, 9.2 are new, and §§11–12 are extended. The prior-art clock for the v1 claims runs from the 2026-07-23 push; for the v2 claims, from this one. (Superseded in part on 2026-07-27 — see the note below; the two speaker formats described here became three, and the claim count moved again.)
v3 note (2026-07-27): this revision adds the seat taxonomy and the two delivery mechanisms that were ratified but never written. §3 is resorted from four seats to five, in two classes — determined (caller, equalizer), whose outputs are pure functions of declared inputs and are checkable by recomputation, and judged (umpire, witness), whose outputs are perceptions and are checkable only by one-sentence explanation — with the seat-migration bar governing both directions. New §4.2 establishes the caller as deterministic given roster and seed, from which auditable fairness, count-only multi-device agreement, and the absence of per-participant state all follow. §5.2 grows from two speaker formats to three named ones — Relay, Call, Free — the relay call having been absent from the specification while it was already the format intended to ship; the pilot constraint is accordingly restated by property (a trial must run a caller-on format) rather than by position in the list. The abstract now leads with the relay rather than the mesh, and P-S5 gains its stated precondition (a device available at recess, no instruction given), without which a null result cannot distinguish an unpopular game from an absent device. Claims grow from sixteen to nineteen. The prior-art clock for claims 17–19 runs from this push, not from either earlier one; §§1–11 claims are unchanged in substance.
v4 note (2026-07-27): a title and abstract correction, no claims change. The paper was titled "An AI-Called Circulation Sport" when the caller was understood as an intelligent intermediary. The v3 seat taxonomy and §4.2 established the opposite — the caller is a determined seat, a pure function of roster and seed, checkable by recomputation and explicitly not a judgement — with the result that the title asserted of the mechanism precisely the property the design had given up in order to make its fairness auditable. Two further readings had gone stale beneath it: the shipping delivery format (the relay call, claim 19) puts a human voice on the announcement, and the arrival of speech synthesis as a pronunciation fallback in the reference implementation gives "AI-called" an available reading — called in a synthetic voice — that was never its sense. The title now reads "An Impartially-Called Circulation Sport for the AI Age": the property rather than the technology, on the same reasoning §4.2 gives for determinism itself — an implementation-independent property outlives the implementation that first carried it. The abstract gains the determinism sentence that justifies the new adjective. The coined term "AI-called circulation sport" remains coined and freed in the Prior-Art statement and is retained in the keywords, together with the patent-class phrase §10 names — retitling does not un-free a term, and the dedication's defensive reach comes from the disclosure rather than from the title. "For the AI Age" is unchanged and means what it always meant: the era whose abundances the sport is designed to spend (§1), not a claim about what runs the caller. Also corrected: the abstract said sixteen claims after the v3 pass took the count to nineteen.
v5 note (2026-07-27): a change of spine, not of specification. Through v4 the paper derived the sport from an abundance — machine intelligence lifting the officiating, matching, and witness scarcities under which all previous sport was designed — and the caller mechanism followed as a way to spend it. That derivation is engineering, and it left the paper unable to answer why the abundance should be spent this way rather than another. The premise is now stated first and the abundance demoted to its enabling condition: a name spoken aloud is an address, an address requires someone other than yourself, and the sport exists to distribute that address equally. §1 is rewritten around it, opening on Carnegie's maxim and inverting it (offered as a technique for advantage; kept here for the recognition, with the advantage engineered out, since nobody chooses whose name is said), with the own-name attention literature supplying the narrower empirical leg that §11 is careful to distinguish from the broader claim. Every mechanism in §§3–8 is unchanged; several are now derived rather than asserted — equal-yet-random, the impartial caller, the absence of per-participant statistics, and the dyadic point all follow from the premise in four lines. New §5.3 states the constraint the premise forces and the prior specification violated: no participant may be called in their own voice. A system holding participant-recorded name audio for pronunciation will reach for it at call time and deliver each player a recording of themselves — not a poor address but no address at all. §5.3 separates the two speech acts a single artifact had been performing (the pronunciation, which belongs to the name's owner; the call, which necessarily does not) and specifies the transfer between them; Call accordingly speaks every name in one uniform voice, which is claimed as a property rather than conceded as a limitation, since a uniform channel addresses all participants identically. Claims grow from nineteen to twenty-one; the prior-art clock for claims 20–21 runs from this push. P-S6 pre-registers the premise itself with two named counter-hypotheses (habituation; celebration), publish-either-way. Two honest limits are added and are the paper's most exposed: the address premise is asserted, not measured, and the uniform synthetic voice satisfies the constraint's letter — it is an other — while only partly satisfying its spirit, since it is not a person who spent attention on you. That gap is handled by ranking the formats (Relay remains the default; Call is the device-plentiful case) rather than by pretending it closes. §5.3 also specifies the gap's partial repair, added at founder request in the same pass and claimed as claim 22: the cross-recording chain, in which each participant records one other participant's name — the circle called in its own voices, with the naive form of that idea (playing each participant their own recording) barred by the constraint and the intuition preserved by inverting the implementation. Two persona references are corrected as stale in the same pass: the call is the bare name, carrying no speaker-attribution prefix — which the address premise independently requires, since a prefix names a speaker who does not exist — and the rabbit mascot no longer teaches this sport. The title changed with the spine, from An Impartially-Called Circulation Sport for the AI Age to A Circulation Sport in Which Everyone Is Called by Another; the v4 note above records the title as it then stood.
v6 note (2026-07-30): one correction the paper owed itself, one new mechanism, and the accumulated backlog folded in. The correction first, stated rather than tightened quietly: §4 and claim 3 gave the equal-distribution guarantee over a session, and a session is a prefix of a round — so the published wording overstated what the mechanism delivers, and §2's classroom scene described precisely the case that fails. Both now read over a complete round. The reference-implementation defect that surfaced this — a caller whose progress reset each session, so that a class of thirty running a ten-minute warm-up called the same ten children every day and never called the other twenty — is specified as new §4.3 and claim 23, round continuity via circle-scoped progress persistence: the mechanism that makes claim 3 reachable under real usage, and one whose absence no session-level test detects, since within any one session the distribution is exactly correct. The prior-art clock for claim 23, and for amended claim 3, runs from this push. Also corrected: the player minimums, written before the relay format and silently assuming the caller was not a person. The abstract and §3 said one to N players and §3.1 gave per-mode minimums as low as one — while a human relay caller is a person who is not receiving, which makes a circle of one impossible in the very format this paper names as its default. The floor is four, now stated once in §3 with its reason, and the per-mode minimums are withdrawn rather than restated, so that one number cannot drift out of step with another. New material, none of it changing an existing claim: §6 gains the two difficulty dials — the circle's radius, and method restriction read upward — with the declare what, never why guard, whose sharpest risk is accessibility rather than fairness (a legible difficulty badge invites a circle to read a body as a declaration); these had shipped to the CC0 rules and never reached the specification, a drift running product→paper rather than the usual direction. §5 states what the two arrows actually do — they relocate the opponent, from a person, to the drop, to the circle's own earlier state. And §9.1 adds P-S7 and P-S8, which test the premise beneath the mission rather than the sport, and which can be run without the sport at all: if P-S7 fails, this sport may pass every other prediction here and still be a well-made game with no mission. Claims grow from twenty-two to twenty-three; pre-registrations from six to eight. Added in the same pass, after a literature check the previous draft of P-S7 had gestured at without performing: §10 now surveys the name-and-belonging literature and states plainly which side of the relation it sits on — Cooper et al. (2017) on perceived instructor knowing, Kohli and Solórzano (2012) on mispronunciation as racial microaggression, Walton and Cohen (2011) as the outcome-side evidence that belonging moves at all, and Goodenow (1993) as the instrument P-S7 and P-S8 specify rather than invent. All of it runs teacher → student, and the authors are aware of no controlled study of student-to-student full-coverage name knowledge — which is P-S7's actual subject, stated as a gap rather than a discovery. §13 also gains a flourish, explicitly marked as one: the Khmer សី sounds like the English say, the two are unrelated, and the paragraph carries its own deletion test.
v7 note (2026-08-13): a floor correction, and the rare one where the shipped product was ahead of the paper. The implementation has been running Call at three players while this document said four globally. §3's Why four is the floor is replaced by Why the floor moves with the format: the surprise condition (three receivers) is restated once as an invariant, and the only variable — whether a body is spent on calling — supplies Relay 4, Call 3. No claim is added; a variable that was always implicit in the derivation is made explicit. This does not reinstate the per-mode minimum table withdrawn in v6, and it does not close the open Team-format question, which is named in §3 rather than left to silence.
v8 note (2026-09-23): the caller contract now has two versions, and the first one had a measured defect this paper did not know about. The published caller (v1) kept a player from being called twice in a row across a round boundary by rotating the colliding opener to the end of the new round — and that rotation made a round's closer close the next round with probability 2/n, two times in three at Call's floor of three, which is the equal-yet-random promise's unpredictable half failing exactly where the circle is smallest. Circles created from 13 September 2026 deal a second version (v2: a 64-hex-character seed, a keyed draw per round, and a boundary swap in place of the rotation); every older circle replays v1 byte for byte, forever, because the seed's type is the version. New §4.4 states the finding, the repair and the versioning rule, with the reference implementation's own measurement (v1 66.7% → v2 49.5% at three players). §4's anyone's name may be next is scoped rather than withdrawn: it is true of the circle and not of every call, since a bag forces each round's last name for anyone counting and the no-repeat rule keeps the previous closer from opening. §11 gains two honest limits (the v1 defect persists in every circle created before the change; the unpredictability is watcher-level, and a locally generated seed is recomputable but never independent of whoever generated it). §10 now says plainly that the recomputable shuffle beneath the caller is not new and points to Decided by No One, the companion paper that generalises the draw and carries its census. No claim is added, withdrawn or reworded; claim 17 reads over both versions of the contract. The body's retired hash-footer sentence is removed under the 2026-08-27 ruling, since that hash is recorded outside the file it would describe.
v9 note (2026-09-26): a repair pass for an examiner-facing mirror; no claim is added or withdrawn. Claim 3(b) and the abstract's anyone's name may be next are scoped to the limits §4.4 already states. The sport is named sey, lowercase and unmarked, wherever the sport is meant; B-Sey℠ (the app) and B-Sey™ (the ball) name only the products, which is the naming the institution adopted after first publication. Unscoped priority and novelty wording is removed or bounded by the censuses now reported in the Prior-Art statement and §10, which also gains the nearest name-game and name-pronunciation prior art. Three Current form notes state where the design as now specified has moved past the text (§3 embodied machines; §7 the mesh and the voiced ball; §9.2 sponsorship), the earlier text being retained as a disclosed variant. A Terms table maps coined names to standard terms, and these notes now stand in date order. Corrected outright: §11's unbuilt and unplayed (a reference implementation exists), §11's placement of the mesh on the first session's critical path (Relay removed it), the dating of golf handicapping, and three citations (the director of Mystic Ball, the full title of Brandt's study, and the title of the companion paper cited as Need-Compiled Questlines, which is The Quest That Mints Nothing).
This paper discloses a group ball game in which an automated caller announces, before each strike, the name of the player who must receive next, choosing receivers by a seeded, recomputable shuffle that gives every player the same number of receptions per round in an order participants cannot anticipate, and applies handicaps only as unequal starting scores. The design premise follows. A name spoken aloud is not a sound but an address, and an address requires someone other than yourself. Carnegie's maxim — that a person's name is the sweetest sound in any language — reads as a fact about acoustics and is really a fact about attention: what a remembered name proves is that another person spent some on you and kept it. This paper takes that as a design premise and asks what a game would be if it guaranteed the sweetest sound to every participant equally, removed the choice of whom to address from human hands entirely, and refused to let the address become a rating. The derivation is tight: everyone addressed (equal-yet-random distribution), nobody choosing (an impartial caller), no evaluation (no per-participant statistics anywhere in the system), the address completed (a dyadic scoring unit credited to no individual), and — the consequence a naive implementation gets wrong — no participant ever called in their own voice, because a recording of yourself saying your own name is not an address but a mirror.
The guarantee was unbuildable until recently, which is the second half of the problem. Sport has historically been designed under three scarcities. Umpires are scarce, so games tolerate bad calls, and the figure of the referee absorbs abuse severe enough to drive large numbers of officials out of school and youth sport in several countries. Fair opponents are scarce, so sport is organized as segregation infrastructure — by age, sex, size, and skill — and the one social unit that most needs shared play, the mixed-age family, is the unit organized sport serves worst (§1 names golf's handicap as the partial exception). Witness is scarce, so games keep score of persons, and the counting hardens into statistics, rankings, and fame hierarchies — the attention economy's atom, present at the playground. Machine intelligence can lift all three scarcities at once. A sport designed for that abundance — rather than an old sport with cameras added — can spend it on what the scarcities stole: inclusion, giving, and story. The order of the two halves is load-bearing and easy to invert: the address is the premise and the lifted scarcities are the enabling condition. An impartial announcer is worth building only if there is a reason every person present should be named aloud.
This paper specifies such a sport. sey — named for its ancestor, the Southeast Asian sey (Khmer សី, the dusk game of collective keep-alive), and implemented commercially as the B-Sey℠ app — is a circle game played by four or more in Relay and three or more in Call with foot or paddle, with or without a net, cooperatively or adversarially, and organized by one mechanism: the caller. A machine intermediary announces, before each strike, the name of the player who must receive next — distributing touches equally yet randomly, so that participation is egalitarian, attention is universal (anyone's name may be next, within the structural limits §4.4 states), and, as a structural by-product, strangers learn each other's names through play. That selection is not a judgement but an arithmetic: the receiver sequence is a pure function of the roster and a seed declared before play, so any participant may recompute a finished session and confirm that nobody was skipped and nobody was protected. The sport's central fairness property is therefore audited rather than trusted — which is why this specification describes the caller as impartial rather than as artificial intelligence, a distinction §4.2 makes structural. Because the caller designates receivers on both sides of the net, targeting a weaker opponent is not merely discouraged but absent from the game's move-set: exploiting the weakest human present is structurally impossible. The point is dyadic — it increments on successful receipt, so every score is a completed give-and-receive and no one scores alone. Handicaps are applied at the start line, never in the stream: unequal teams begin at unequal countdown numbers on a coarse fixed lattice (multiples of three), a quantization that makes the handicap read as a garment size rather than a skill rating. Cooperative play counts up (the collective streak); adversarial play counts down to zero — first to nothing wins. The announcement reaches the circle in one of three named formats, of which the default — the relay call — speaks the name privately to a single rotating human, who calls it aloud: fairness stays machine-held while delivery is human, so the sport runs on one borrowed device with no loudspeaker, no network, and no clock synchronization, and the rotating relay is where participants practise producing the names they otherwise only hear. Where devices are plentiful, a synced-device mesh speaks the call in unison instead, in one voice for every name — never the named player's own recording — so that every participant is addressed identically, which is the equal-yet-random property rendered in the acoustic domain as well as the numeric one. Pronunciation authority nonetheless stays with the name's owner in both formats, transferred to a foreign articulator either by modeled repetition (the relay hears the owner's recording immediately before speaking) or by an owner-supplied respelling that drives the uniform voice — the pronunciation and the call being two distinct speech acts whose rightful speakers differ.
The sport is parameterized on three orthogonal axes — method (the body vocabulary each player declares: hand, paddle, or foot, mixed freely within one circle), mode (what the circle plays against: nothing, its own record, or another circle), and level (the lead time between the call and the object's arrival). The default state has no score at all, and its stated purpose is that everyone learns everyone's name with correct pronunciation; at the longest lead time the called player raises a hand, so the ladder from easy to hard is a scaffold-fading memory curve rather than a skill ladder — the game becomes more difficult exactly as the group's name recall improves. Scoring degrades gracefully across three tiers — implied by the call, validated by a k-of-N vote of the circle's own microphones, or confirmed by an accelerometer in the object — none of which gates play, and all of which are licensed to be loose because the score moves no value. The system is local-first: the entire game runs on one device, offline, with no accounts. Twenty-three claims are enumerated and dedicated to the public domain; eight predictions are pre-registered, including the design's own existential wager and a test of the address premise itself, both of which the authors commit to publishing either way.
Keywords: designed sport, automated turn assignment, receiver selection, seeded shuffle, publicly verifiable random selection, handicap by starting score, AI officiating, AI-called circulation sport, AI-directed pass calling, deterministic turn assignment, circle game, sey, chinlone, sepak takraw, mixed-age play, handicap systems, loneliness intervention, third place, name learning, cooperative games, defensive publication.
This document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. The authors and HeartBank® will not seek patent on any mechanism, architecture, or specification articulated herein, in any jurisdiction, at any time.
The following terms are coined in this paper and simultaneously freed with it: AI-called circulation sport, the caller mechanism, equal-yet-random distribution, the cannot-target-weakness property, the dyadic scoring unit, start-line handicap and quantization-for-dignity, the last-gift rule, the synced-circle mesh, and the play/currency wall. Freed with the v2 revision: the call-lead-time ladder, scaffold-fading difficulty, method-as-declared-handicap, the implied point, circle-scoped adaptation, the acoustic inversion, rosterless classroom mode, and the sponsorable-place test. Freed with the 2026-07-27 seat pass: determined and judged seats, the seat-migration bar, the determined caller, count-only circle coherence, the host-agnostic turn-assignment layer, and the relay call. Freed with the 2026-07-27 address pass: the address premise, the address constraint, the uniform calling voice, pronunciation-authority transfer, and the cross-recording chain. Terms inherited from this corpus's earlier publications (Proof of Coordinate℠; the mints-nothing boundary; the beating credential) are cited, not re-claimed.
Smart balls, computer-vision officiating, and AI sports-coaching applications are active prior art and an active patent field (§10); the present claims are confined to the pre-announced named-receiver, equal-random-distribution, dual-arrow, lattice-handicapped circle-sport architecture and to its dignity-preserving constraint set, which is the part an extractive implementation would omit.
What the censuses found. Two desk surveys bear on these claims, and both narrowed them. (1) The census reported in the companion paper Circulation Sports (22 published warm-up, drill and icebreaker formats across seven sports and the physical-education and icebreaker literature; 2026-08-07; documented practice, not fieldwork) found guaranteed reach common, in four architectures (the queue, fixed role rotation, round completion, and turn-taking norms), and found names already called in a circle: in the soccer Passing Name Game the passer calls the name of their own chosen receiver, and in the Name Toss icebreaker a round closes when everyone has received once. Not found in that aperture on that date: reach that is guaranteed and whose order stays unpredictable across unbounded play, held by a party other than the passer. That conjunction, the external caller, is what claims 1–3 disclose. (2) A survey of name-pronunciation tools (2026-08-17) found the recording of a name's pronunciation by its owner widely practised (§10); claim 21 discloses its use to transfer pronunciation authority to another speaker, not the recording itself. The recomputable shuffle beneath the caller is established prior art (§10). No census has been run on the cannot-target-weakness property, the start-line lattice, or the dual-arrow ruleset, and nothing here asserts that they are new.
"A person's name is to that person the sweetest and most important sound in any language." — Dale Carnegie, How to Win Friends and Influence People (1936)
Carnegie offered that as a technique. The advice was instrumental — learn names, gain advantage — and the century of sales training built on it has made the sentence faintly embarrassing to quote. This paper keeps the observation and discards the instrumentality. In the sport specified below nobody chooses whose name is said, so no participant can deploy a name to their own benefit; the recognition survives and the advantage is engineered out. What remains is the plain finding underneath the salesmanship, which experimental attention research reached independently and from a different direction: a person's own name penetrates a channel they were not attending to, where other words do not (the own-name effect within the cocktail-party phenomenon).
The premise of this sport is one step past that finding, and it is a step Carnegie did not need to take because he was addressing a speaker rather than designing a system.
A name spoken aloud is not a sound. It is an address — and an address requires someone other than yourself.
Your own name in your own voice is not an address at all; it is a recording. Nobody is calling you. The sweetness Carnegie describes is not acoustic — it is the evidence a name carries that another person spent attention on you and kept it. A remembered name is a receipt for attention already given. This is why being greeted by name by a shopkeeper lands differently from a personalized mail-merge, and why the difference is not one of technology but of who is doing the addressing.
From that single premise the sport's architecture follows almost without further argument:
What the sport produces, then, is not fitness and not a winner. It is a quantity of addressed attention, distributed evenly by construction, in a setting where the people doing the addressing are strangers who will now be able to greet each other.
The premise is not new — every coach who learns thirty names by the second practice is acting on it. What is new is the possibility of guaranteeing it, and the guarantee was, until recently, unbuildable. Sport is one of humanity's great social technologies, and like every technology it embodies the constraints of its era. Three of those constraints did most of the shaping, and all three stand between the premise above and any game that could deliver it.
The officiating constraint. A fair game needs a neutral judge, and neutral judges are expensive. Most human play therefore proceeds unofficiated — with the familiar result that disputes end games, dominant personalities adjudicate in their own favor, and children learn early that the loudest voice owns the rules. Where officiating exists, it concentrates in a single fallible human who becomes the game's lightning rod: abuse of referees is severe enough that officiating bodies in the United States and the United Kingdom have reported large losses of officials, attributed largely to the behaviour of adults, with games cancelled or moved for want of officials. The referee crisis is not a culture problem that better manners will fix; it is a structure problem — one human absorbing the full adversarial pressure of everyone else's stakes.
The matching constraint. A good game needs opponents of comparable ability, and comparable opponents are rare. Sport's solution was segregation: leagues by age, divisions by skill, categories by sex and weight. The solution works — and quietly forecloses the most valuable game that could exist: the one where a grandmother, a teenager, and a six-year-old contest the same point and all three matter. The family, the one social unit whose play would repair the most, is the unit sport's architecture serves worst. (The lone mainstream counterexample, golf's handicap system, proves the demand: it is the reason golf is one of very few sports a parent and child can genuinely contest, and it required a century of institutional machinery to maintain.)
The witness constraint. A game's story needs recording, and recording was expensive, so sport recorded the cheapest summary: who scored. From that compression grew the entire statistical culture of sport — batting averages, rankings, ladders, most-valuable-player awards — a fame hierarchy erected on a bookkeeping shortcut. The consequences run downhill to the playground: the child picked last exists because informal sport, too, keeps score of persons; the pickup game has a status economy, and everyone present knows their rank in it.
Machine intelligence can lift all three scarcities. Officiating becomes abundant (a phone judges a line call better than a linesman); matching becomes computable (fairness can be manufactured rather than found); witness becomes total (everything can be recorded, which forces the question of what should be). The default path — already visibly underway in the patent literature (§10) — is to spend this abundance on the old game: more statistics, finer player-surveillance, engagement-optimized training loops. The sport specified below spends it in the opposite direction: on inclusion the matching constraint forbade, on giving the scoring compression erased, and on an output sport has rarely set out to produce — strangers who know each other's names.
The relation between this section's two halves should be stated exactly, because it is easy to invert. The address is the premise; the lifted scarcities are the enabling condition. The abundance did not suggest the goal — a tireless impartial announcer is worth building only if there is a reason every person present should be named aloud, and Carnegie's observation is that reason. What the abundance supplies is the guarantee: a circle can now promise every participant an equal share of being addressed, and keep the promise for the whole session, without asking a human to hold the ledger. Everything downstream — the seats, the arithmetic, the two count-arrows, the start-line lattice — exists to protect that guarantee from the ordinary social forces that erode it. A useful test of any proposed change to this specification is therefore not does it make the game better but does it keep everyone being called by another.
A design axiom governs throughout, shared by all three papers of this sprint and coined here as the play/currency wall: competition and play may generate the story; only witnessed giving moves the currency. Winning, in any game this institution builds, may move narrative, honor, or the address of an already-committed gift — never value itself. The wall's application to stakes and patronage is its own paper (The Wager That Isn't, this sprint); its application here is simpler and prior: in sey, there is nothing to move. What victory produces is a story.
A park, late afternoon. A man finishes a quiet quarter-hour on a bench — the contemplative practice is its own product and no part of this paper — and walks toward a loose ring of people batting a rattan-weight ball into the air. He does not know any of them. Someone widens the circle; he types a name into the app — just Virak — and his phone joins the mesh. A moment later every phone in the circle says, in one voice, one word: "Virak!" The ball comes to him — gently; the woman who sent it saw a newcomer, and delivery is the whole skill of this game. He gets his paddle under it. The circle cheers the streak alive. Within three rounds, everyone in the circle has heard his name four times, and he has heard all of theirs, because the game cannot proceed without saying them. When the streak dies at 41 — nine short of nothing in particular, since the next rite waits at 333 — the circle groans and laughs, and a stranger says, "Again, Virak?" That sentence — a stranger using his name to invite him back — is the sport's actual product. The rally was the packaging.
A living room, that evening. Three generations, one net strung between two chairs. The eight-year-old and her grandmother hold paddles; the parents play with their feet. The family's phones announce in two languages — the grandmother's in Khmer, the child's in English — one circle, one game, two tongues. The teams are uneven, so the app has set the start numbers: the parents begin at 33, the paddle side at 21. First to zero. Nobody chose those numbers; nobody can argue with them; and because the caller decides who receives, the parents cannot serve at the eight-year-old to win — the game has no such move. When the paddle side takes its last point — actively won, as the final point must be — the child performs a victory lap of the sofa. What she has won is the story of having beaten her parents fairly. That is all winning moves here, and it is enough; it has always been enough.
A classroom, the first week of school. Thirty children who mostly do not know each other, one teacher who needs to learn thirty names, one ball. The circle solves both problems, because equal-yet-random guarantees every child is called once the round completes — and because the round is carried across sessions rather than restarted (§4.3), a class playing ten minutes a day reaches every child instead of cycling the same few. The teacher, standing inside the circle, hears every name spoken aloud dozens of times. No child is picked last, because there is no picking. No child dominates, because the caller will not allow the ball to orbit the confident. The physical-education period and the social-integration project turn out to be the same forty minutes.
The sport is a circle game for four or more players in Relay and three or more in Call (the floor is a function of the format axis; see Why the floor moves with the format below), played with a ball designed to be struck by foot or paddle — the athletic adult kicks, the grandparent and the child paddle, and both choices are ordinary equipment decisions, not accommodations. Play proceeds in two modes. Without a net, the circle plays cooperative keep-alive: the score counts up, one per completed reception, and the object is the longest unbroken streak — the collective form. With a net, two sides contest a countdown (§5–6) under volleyball-like rally structure — the adversarial form, and the form on which the companion paper's challenge grammar (B-Match℠) nests. A newcomer joins either form the same way: the circle widens. There are no teams to pick in circle mode and no tryout in either; the topology itself is the invitation. Why the floor moves with the format. The floor is not a constant but a function of one variable — whether the caller is a person. The derivation is single and does not change: three receivers are required, because three is the smallest number that leaves any surprise in the call. With two receivers the caller can only alternate — everyone knows who is next before the name is said, and the mechanism this paper specifies contributes nothing.
What changes is only whether a body is spent on calling. In Relay a human calls and is therefore not receiving, so three receivers cost four players. In Call no human calls, so all present are receivers, and three receivers cost three. One rule, applied to two arrangements:
required receivers = 3 (invariant — the surprise condition)
Relay caller is a person → floor = 3 + 1 = 4
Call caller is not → floor = 3 + 0 = 3This is deliberately not a return to the per-mode minimum table withdrawn in v6. That table restated a number per row, which is exactly how two numbers drift apart. This states the number once, as the surprise condition, and lets the format axis supply the only term that varies. The withdrawal stands; what is added is the variable it was missing.
Provenance, stated asymmetrically because it is asymmetric. The invariant and its derivation are this paper's own, unchanged since v1. Three-in-Call is not — it is prior canon in the product's own design record, where a circle of three standing close with no human caller was specified as the answer for a family of three, and it is the implementation that has been running while this paper said otherwise. The adjacent argument that three is the smallest number at which forward and back differ belongs to the circulation-sport work and is cited here, not claimed.
A gap this amendment does not close, and should not be read as closing. The paper still says nothing about which formats a Team game may run, whether a relay caller belongs to a side, or how rotation crosses a net. The floor above is stated for Relay and Call only. Team ships visibly disabled for exactly this reason, and until the axis is specified the honest position is that the format question is open rather than answered by omission.
The machine intelligence in the sport occupies exactly four seats — including the one the sport is named for — and is barred from the fifth. The seats fall into two classes, and the division is not a taxonomy for its own sake: it decides how each seat's output can be checked, and therefore what an implementer is permitted to put in it.
The determined seats. Their output is a pure function of inputs declared before play, so it is checkable by recomputation — anyone holding the inputs reproduces the result exactly and sees for themselves that the rule was honoured.
The judged seats. Their output is a perception or a narration. It cannot be reproduced from inputs, so it is checkable only by explanation — each judgment stating its ground in one sentence, the standard §5 already imposes on the last-gift ruling.
The barred seat.
Current form. The design as now specified admits an embodied machine into play on one condition: that it displaces no human — not a receiver, not the caller, not an opposing team. Where that holds, a machine may receive (for example as a third receiver when only two people are present, which restores the surprise condition of §3), or may stand as the opposing side in net mode; it keeps the object moving but never keeps the score, since a pass involving a machine counts toward no streak or record; it may never take the relay role, which exists so that people practise saying names; and it plays at a deliberately human level and is never preferred to a person. In Call it may also serve as the single speaker in place of the device mesh. The machine-free rule above is retained as a disclosed variant.
The seat-migration bar. A seat may not take on the other class's mode of operation, and the prohibition runs in both directions because both failures arrive at the same place — a machine-held ranking of persons. A determined seat that acquires judgment becomes a rating engine: a caller that routes around a struggling receiver, or a handicapper that tunes start numbers to measured ability, has evaluated the players, and has done it where no one can see it. A judged seat that acquires distribution authority becomes a caller: an umpire empowered to influence who receives next can express a preference about people through a channel built for judgments about the ball. The rule is short enough to hold in mind while implementing: the determined seats never judge a person; the judged seats never choose a person.
The sport is parameterized on three orthogonal axes. They are not settings; they are a ladder, and an implementation that surfaces them as a configuration screen has already failed the design.
AXIS QUESTION IT ANSWERS VALUES
─────────────────────────────────────────────────────────────────
METHOD what part of you gives? hand · paddle · foot
MODE what do you play against? nothing · your record · another circle
LEVEL how much warning? long lead · moderate · short leadMethod is the body vocabulary a player declares — hand-catch-and-release, paddle, or foot. Anatomy plus implement exhausts the list; there is no fourth. The values are defaults suggested by ability, never assignments by category, and heterogeneous methods are expected within a single circle: a grandparent with a paddle, a child with hands, an adult with feet, one shared object. This is the load-bearing case, not the exotic one — a sport that segregates by method has thrown away the mixed-age property that is the design's principal social claim, and would have reintroduced by the back door the segregation §1 identifies as sport's inherited disease.
A declared method is a start-line handicap (§6) — public, self-chosen, announced before play, never assigned during it — so the axis requires no new fairness machinery; it inherits the machinery already specified. Two consequences follow:
Mode is what the circle plays against, and the values ascend: nothing (Friendship, no scoring of any kind), its own record (Streak, no net, ascending collective count), another circle (Team, net, descending count to zero, and the only mode additionally requiring enough players to form two sides). All three take their floor from the single rule of §3 — three receivers, plus a body for the caller only where the caller is a person (Relay 4, Call 3). The per-mode minimum table given in revisions before v6 remains withdrawn and is not restored: it restated a number per row, which is how two numbers drift apart. One number, stated once, with one variable. The mode names are deliberate: no label on the product surface names rivalry, though the adversarial mode is genuinely adversarial and this paper describes it as such. Naming the third mode for its structural fact — sides exist — rather than for its disposition is a small instance of a general discipline in this corpus, in which vocabulary is held to the same standard as mechanism.
The mode ladder is also the adoption funnel, and an unusually honest one: each rung is entered by the circle's own election, each rung's graduation is the previous rung's success condition, and the entry rung has no score at all.
Level is treated in §4.1, because it is a property of the caller rather than of the players.
The three-value pattern is not aesthetic and should not be read as one. Each axis has an independent reason for its cardinality: method is exhausted by anatomy plus implement; mode is exhausted by the opponent relation; level is coarsely quantized for the same anti-rating reason the start-line lattice is (§6) — a continuously variable difficulty dial would read as a precision skill rating, which is the artifact this design exists to avoid. Where a fourth value would genuinely serve — very large circles may want a fourth lead-time band — the pattern must yield. Three is a finding here, not a constraint.
The sport's central mechanism is small enough to state in a sentence: before each strike, the system announces by name the player who must receive next.
Everything distinctive about the sport follows from that sentence.
Equal-yet-random distribution. The caller's selection policy has two commitments that are individually common and, held jointly by a party other than the passer, were not found in the census reported in the Prior-Art statement. Equal: over a complete round — one pass through the roster — every player in the circle receives the same number of touches, within rounding, and the weakest player is guaranteed the strongest player's share of the game. Revisions of this paper before v6 stated this over a session, and that overstated it: a session is a prefix of a round, and a prefix of a fair sequence is not fair. The guarantee is reachable under real usage only because rounds are carried across sessions rather than restarted (§4.3). Random: the order is unpredictable, so no player can disengage — anyone's name may be next, and the suspense that in ordinary sport attaches only to the ball here attaches to every person simultaneously. Anyone's name may be next is true of the circle rather than of every call, and §4.4 states the exact scope: within a round the bag forces the last name for anyone who counts, the no-repeat rule keeps the previous closer from opening the next round, and the first published version of the caller leaked more predictability than either — a round's closer closed the next one two times in three at the smallest circle. Equality without randomness would be a queue (predictable, boring, and stigmatizing — everyone can see whose turn is being "given"); randomness without equality is every pickup game ever played (the ball orbits the skilled). The conjunction is the mechanism.
The cannot-target-weakness property. In net mode, the caller designates receivers on both sides. The attacking side therefore cannot choose to serve at the weakest opponent — the oldest strategy in net sports, taught openly at every level, and the precise mechanism by which the weakest player on any volleyball court spends the match being hunted. In sey that strategy does not exist. It is not penalized; it is not discouraged; it is absent from the move-set, the way castling into check is absent from chess. Skill, denied the target-selection channel, migrates entirely into delivery — placing the ball so your designated receiver, whoever the caller named, can continue the play. The best player in a sey circle is definitionally the one who makes everyone else look good, because making the named receiver succeed is the only thing the rules allow skill to do. It is disclosed here as the design's sharpest single contribution (Claim 2); no census of net sports has been run for it.
The name as by-product and as point. Because the call is a name, spoken aloud, repeatedly, by every device in the circle (§7), the sport performs introductions as a side effect of its basic loop. This is not a social feature bolted onto a game; it is the game's addressing layer, doing double duty. The design intends the double duty (§9's first prediction measures it): the sport is built so that its standard output includes strangers who know each other's names — the missing primitive of the third place. Name games such as the Name Toss icebreaker pursue the same output without a caller (§10); what the caller adds is that no participant chooses whose name is said.
Hard constraints on the caller (stated here as design law; their governance form lives in the institution's internal directive registry):
The naive difficulty dial for a called sport is the interval between calls. It is the wrong variable: the inter-call interval is confounded by circle size, so the same setting produces a frantic game with four players and a languid one with thirty. The correct variable is call lead time — the interval between the naming of the receiver and the object's arrival. Lead time is invariant to circle size and to method, and it is what a receiver actually experiences as pressure.
Three bands are specified, and what changes across them is not merely speed but how much external scaffolding supports the act of recognition:
LEVEL LEAD TIME SCAFFOLD WHAT IS BEING TRAINED
──────────────────────────────────────────────────────────────────────────
Easy long called player RAISES A HAND nothing yet — the
and may speak their own circle is being taught
name aloud the names
Medium moderate none — the giver must find name → face recall
the named face unaided
Hard short none; the call may land automaticity under
while the object is aloft time pressureAt the longest lead the called player raises a hand, which removes the recall requirement entirely: a giver who does not yet know who Sokha is can simply deliver to the raised hand. The scaffold is then withdrawn — first the hand, then the time — and the withdrawal is enforced by physics rather than by rule, since a hand cannot usefully be raised inside a short lead.
The result is that the difficulty ladder is a memory curve, not a skill curve. A circle can only ascend it as its members' name recall becomes fast and then automatic, which yields the design's cleanest coincidence: the game becomes more fun exactly as everyone learns each other's names. The thing the players want and the thing the sport is for are the same thing, and no exhortation is required to align them.
It also supplies the sport's non-numeric progression. In Friendship mode there is no score, and the circle nonetheless has something that advances: tempo is the score that is not a score.
A constraint the implementer must not violate. It is tempting to advance the level automatically from measured response latency. Doing so per player would reconstruct, invisibly and in the tempo dimension, precisely the artifact §4 exists to prevent — a machine-maintained ranking of who is slow. The rule:
The level is a property of the circle, never of a player.
Permitted implementations, in order of preference: (a) the circle elects the level, with the caller permitted to offer an advance; (b) if adaptive at all, adaptation conditions only on circle-aggregate state, ephemeral and on-device, never per-player, never stored, never displayed. Never permitted: a per-player recall score, in any form, visible or not.
The caller is the seat where an implementer will most want to put intelligence, and the seat where intelligence does the most damage. This design resolves the tension by removing the caller from the domain of judgment altogether: the call sequence is a deterministic pure function of the circle's roster and a seed declared before play. Given the same roster and seed, every conforming implementation emits a bit-identical sequence, which makes the generator a cross-platform contract rather than an internal detail.
Three properties follow. The third was not anticipated when the constraint was adopted, and is now the reason to keep it.
Fairness becomes verifiable rather than trusted. Equal-yet-random (§4) is the sport's central promise and the one a participant is least able to check by watching, since over any short window a fair sequence and a biased one look alike — which is precisely why a sport that merely asserts fair distribution asks to be taken at its word. When the roster and seed are the entire input, any participant, parent, teacher, or competitor can recompute a completed session and confirm that the distribution was equal and that no one was skipped, favoured, or quietly protected. Publishing the generator is therefore part of the mechanism rather than a disclosure concession.
Multi-device play stops requiring a network. Because each device computes the sequence independently, devices in one circle need never be told who is next — only how many calls have elapsed. A single integer, not a name, is the whole synchronization payload; a device rejoining after a lapse fast-forwards to the current index instead of recovering state. Where simultaneous announcement (§7) demands tight clock alignment, sequence agreement demands none at all. The offline-first commitment of §7.1 is thus not merely compatible with the determined caller — it is largely delivered by it.
There is no per-player state to leak, subpoena, or repurpose. The forbidden artifact of §4 and §4.1 — a machine-maintained record of who is slow, who is skipped, who is protected — is not merely prohibited here by policy. A pure function of a roster and a seed has nowhere to keep such a record. The guarantee moves out of the governance layer and into the arithmetic, where a later release cannot quietly relax it and an acquirer cannot monetize it.
The general lesson, offered to implementers of any called game. A fairness property enforced by a rule about the algorithm lasts exactly as long as the organization enforcing it; the same property enforced by determinism and published vectors outlasts the organization. Where a guarantee can be made checkable, making it checkable is worth more than making the underlying decision cleverer — and the two are usually opposed, because the intelligence that would improve a decision is the same thing that makes it unauditable. The corollary an implementer meets later is that a checkable sequence can never be fixed in place: once a seed has dealt a session somebody may recompute, its sequence is a record. §4.4 is the worked case — a real defect, repaired by versioning the seed rather than the algorithm.
None of this disarms the machine; it relocates it. Perception and narration — was the ball in, was the final point actively won, what is this circle's story — are irreducibly judgments, and they are where sensing and inference legitimately belong. The boundary is worth restating at the hardware layer beside the warning already given there (§7.1): instrumentation serves the judged seats only. An accelerometer reporting a touch supplies the umpire with evidence. It must never supply the caller with a reason.
The equal-distribution guarantee holds over a complete round — one pass through the roster — and not over an arbitrary slice of one. The distinction is not pedantry. It is the difference between a working guarantee and a broken one, and the reference implementation shipped the broken version first.
A class of thirty playing a ten-minute warm-up calls perhaps ten children. If the caller's progress resets when the session ends, the next session calls the same ten, and the twenty children at the far end of the round are never called at all — every day, indefinitely, with nothing on any surface to indicate it. The distribution is perfectly equal within each session and catastrophically unequal across them. An invariant enforced only within a unit is silently void across units, and the unit boundary is where nobody looks.
The repair is to persist the round rather than the session. The caller retains a circle-scoped integer call index — never a per-participant record, never a timestamp — so that a truncated session resumes where it stopped rather than restarting: whoever has not been called yet is still uncalled tomorrow. This is what makes the guarantee reachable under real usage, in which play is chopped into ten-minute pieces by bells, weather, and attention.
Two properties are worth stating for implementers. First, the persisted state is a single integer per circle, which is why it introduces no per-participant artifact and nothing to leak (§4.2). Second, no amount of session-level testing detects the defect, because every test anyone would write runs one session, and within one session the distribution is exactly correct. A competitor implementing from this specification would ship the same bug.
§4.3 recorded a defect that lived at the boundary between sessions and that no session-level test could see. This section records a second defect of the same family — invisible inside any one round, visible only across the boundary between two — and the way it was repaired without altering a single sequence that had already been dealt.
What the first published caller did at a round boundary. Equal-yet-random is dealt as a bag: each round is a fresh shuffle of the roster, dealt to its end before the next is drawn, which is what makes the counts equal. Two independent shuffles will sometimes deal the same player last in one round and first in the next — called twice in a row, which the circle hears as a stutter and the player experiences as being singled out. The first version of the cross-platform contract (v1: a 32-bit seeded generator, a descending Fisher–Yates shuffle, and the bag, published with conformance vectors in @333eco/sey) forbade the repeat with a rotation: if a new round would open with the previous round's closer, that player was moved to its end. The rule does what it was written for. Its cost was not measured until 13 September 2026, when a documentation example printed two identical consecutive rounds and the explanation first offered for it was checked rather than accepted. Rotation maps every shuffle that would have opened with the closer onto a shuffle that ends with them, so the previous closer is never first, takes each middle position with probability 1/n, and closes the next round with probability 2/n — twice the rate of any middle position. A round repeats the previous round's order exactly with probability 2/n! rather than 1/n!.
PLAYERS v1 ROTATION — closer closes again v2 SWAP — closer closes again
analytic 2/n measured * analytic 1/(n−1)
─────────────────────────────────────────────────────────────────────────────
3 66.7% 66.8% 50.0% (49.5% †)
4 50.0% 49.9% 33.3%
7 28.6% 28.5% 16.7%
12 16.7% 16.6% 9.1%
* 40,000 seeds per row.
† measured in the reference implementation's own engine, which measured
the v1 form at 66.7% under the same conditions.Why it was this sport's defect before it was anyone else's. The effect is sharpest where the roster is smallest, and the smallest roster this paper permits is Call's floor of three (§3) — a floor derived from the requirement that some surprise remain in the call. At three players a participant who has watched one round could name the second call of the next round correctly two times in three under v1. Under the repaired form the same guess succeeds one time in two, which is the best any rule forbidding back-to-back calls allows: with the previous closer barred from the opening, the four permitted orders are equally likely. Equal counts were never touched — every round still dealt every player exactly once, and long-run averages remain equal by symmetry — so neither a count nor the recomputation audit of §4.2 could see the defect. It was a failure of unpredictability, the half of equal-yet-random that §4 says keeps a player from disengaging, and it was worst at the very floor that exists to protect that half.
The repair. The second version (v2) changes the seed, the generator and the boundary rule, and makes one structural change to Relay. The seed is 32 random bytes written as 64 lowercase hexadecimal characters. Each round is drawn from its own key — an HMAC-SHA-256 of the seed, a stream label (receivers, or relay callers) and the round number — expanded by an HMAC counter stream and rejection-sampled into uniform indices for the same descending Fisher–Yates shuffle, so there is no 32-bit state. And the boundary rule is a swap with a uniformly drawn later position instead of a rotation: the previous closer still never opens the next round, and is equally likely to land anywhere else in it. In Relay, a caller's tenure becomes exactly one receiver round, callers are drawn in rounds of their own over the whole circle, and each receiver round deals everyone except the current caller. The formats, the floors, the level fields, count-only catch-up and the refusal to keep per-player state are unchanged. The normative text is the sport's contract (@333eco/sey 1.1.0, The caller sequence, §8), which calls the round key and the keyed shuffle from the general-purpose CC0 package @333eco/primitives (its specification, Parts I–II); both ship vectors, and the contract is sufficient without the package.
Version the seed, never the algorithm. A sequence that anyone may recompute is a record, and the v1 contract had already said what that forbids: changing the sequence for an existing seed is a breaking change, because a circle saved on one device would replay differently on another and every past audit would silently stop reproducing. So v1 was not edited. Circles created from 13 September 2026 carry a v2 seed; every older circle keeps its numeric seed and replays v1 byte for byte, indefinitely. A client tells the versions apart by the seed's type alone — a number is v1, a 64-hex string is v2, anything else is not a seed — so no stored circle changed, no version field can be lost or mistyped, and a client must implement both to replay every circle. The contract's earlier format axis was the first instance of the rule, added by making the absent value mean the old behaviour; v2 is the larger one, and the one that shows why the rule exists. A repair that edited v1 would have fixed the future by falsifying the past.
What nobody can predict means, scoped. The property is watcher-level. It says that no one attending to the play can know whose name is next, and it holds within three limits that are structural rather than defects. (a) The bag forces each round's last name for anyone who counts, and more generally a player already called in the current round knows they will not be called again before it ends; this is the price of equal counts, and it is paid knowingly, since a draw that could call the same player again at once would give up the equality that is the sport's other half. (b) The previous closer never opens the next round — the no-repeat rule's intended and bounded cost. (c) The sequence is not secret from a device. Count-only coherence (§4.2) works because every device in the circle holds the seed and computes the same order, so any of them can compute ahead; the unpredictability the sport needs is the absence of a pattern a player can read, and v1's closer effect was exactly such a pattern. Beyond these, the seed is generated on the device that opens the circle. Recomputation proves that a session was dealt faithfully from the declared seed; it cannot prove that the opener did not reopen the circle until an order they preferred appeared. For play order that is an acceptable residue. For anything of value it is not, and the companion paper Decided by No One specifies the form that removes it — a seed no party controls, a roster committed before the seed can be known — where the draw is generalised beyond this sport.
The dyadic scoring unit. A point increments when the named receiver successfully receives. The unit of scoring is therefore not a strike, a kill, or a goal — acts one player performs — but a completed give-and-receive: a delivery good enough to be received, and a reception good enough to continue. Nobody scores alone, structurally. The compression error of historical scorekeeping (§1) — recording the last touch and calling it the achievement — is not corrected by tracking assists (a patch several sports have tried) but by defining the point as the pair. In the institutional grammar this corpus builds elsewhere, the point is a witnessed micro-gift; within this paper it is enough that it is a dyad, and that the sport's atomic event requires two people succeeding together.
Count up (the circle). Cooperative keep-alive counts consecutive completed receptions. The streak is a built thing — collective, cumulative, mortal. When it dies, it dies for everyone at once, and the circle starts again: the game's only loss condition is shared by all players equally, which is why circle-mode sey has no losers at all, only a streak that was and the next one. Streak milestones are celebrated at the repdigits — 3, 33, 333, 3333 — an exponential rationing that keeps each rite roughly ten times rarer than the last (a celebration every 33 would inflate to wallpaper across a long streak; a celebration at repdigits stays an event for the rest of the sport's life). Between rites, the circle's record — the park record, the school record — carries continuous motivation. Records belong to circles and places, never to persons.
Count down (the net). Adversarial sey inverts the scoreboard: each side begins at a start number (§6) and every rally won subtracts one from the winner's remaining count. First to zero wins — first to nothing. The arithmetic is identical to counting up; the symbolism is its inversion, and a scoreboard is nothing but symbolism: the winning team is the team that first gives away everything it started with. Rally structure follows volleyball (serve rotation; a point every rally). Two refinements complete the mode:
Both arrows, one ruleset. Cooperative sports count up; adversarial sports count up; darts and a few others count down. A single sport in which the mode selects the arrow — building in the circle, emptying at the net — is disclosed as architecture (Claim 5): the two arrows are not a gimmick but the two halves of the institution's founding game (a children's balance game in which kindness counts up and gratitude counts down and the ideal score is zero), expressed as sport.
What the arrows do is move the opponent. In an ordinary net sport the opponent is a person, and every point is taken from someone. The arrows relocate that. Counting up, the circle plays against the drop — gravity, entropy, its own prior record. Counting down, it plays against another circle, by agreement. A third position exists and is the one the level ladder of §4.1 occupies: the opponent as the circle's own earlier state, which is competition that costs no one anything. This is why the entry rung carries no score at all and why advancement between modes is elected rather than triggered (§3.1). The sport does not remove competition; it makes the choice of opponent explicit and puts the adversarial option last.
Scoring in a called sport admits an architecture unavailable to every other game, and it follows from one observation: the caller has already assigned identity. Every other sport must determine who touched the object; this one chose who touches it next. What remains to be established is only whether the touch occurred — a binary event, not a localization problem.
Three tiers implement that, and each degrades gracefully into the one below:
TIER MECHANISM HARDWARE
────────────────────────────────────────────────────────────────────
default the point is IMPLIED by the call — none
assume the dyad completed
validated k-of-N vote across the circle's own none
phone microphones: "was there a strike
in this window?"
confirmed accelerometer in the object reports instrumented
{object_id, touch, t} objectSensing never gates play. A circle with one phone and a rolled sock plays the whole sport; instrumentation improves fidelity and nothing else.
The microphone tier deserves note because it is cheap for a non-obvious reason. Localizing a strike acoustically across phones — time-difference-of-arrival — is a research problem: it needs tight clock synchronization and degrades badly in wind and traffic. Detecting one does not. A strike is a sharp transient, separable from ambient noise on a single microphone; a k-of-N agreement rule across the circle tolerates a phone in a pocket, a gust, and a passing motorcycle; and the vote requires only a coarse shared window, which the mesh's announcement synchronization (§7) already supplies. The reduction from localization to detection is what moves this tier from research to implementation, and it is available only because the caller supplies identity.
The implied point is an epistemic claim, not a semantic one — and the distinction is load-bearing. The scoring unit remains what §5 defines: a completed delivery-and-reception pair, credited to no individual. The default tier does not redefine the point; it assumes the dyad completed and permits a human, or later a sensor, to say otherwise. An implementation that treated the call itself as the scoring event would have quietly built a different game — one in which the caller scores — and would have destroyed the dyadic unit that §5 claims. The caller never scores.
Why such loose epistemics are licensed here. Because of §8: no point mints, moves, or gates value. A score that moves nothing requires no audit. Were the score coupled to money, the instrumented tier would be mandatory and the default tier negligent; because it is not, trust is the correct default and sensing is an amenity. This is a case of a constraint adopted for dignity reasons removing an engineering dependency — the play/currency wall paying for itself in hardware.
One consequence for the record architecture (§12, claim 10): a trusted-by-default streak can silently inflate. Records therefore carry their evidence tier — unvalidated, microphone-validated, or object-confirmed — so that a circle's record states not only its magnitude but how it was witnessed.
A circle has one acoustic channel and two candidates for it: the name of the next receiver, and the count. They cannot both be the headline. Interleaving them — "Sokha! … fourteen! … Dara! … fifteen!" — buries the name in numerals, and the name is what the sport is for.
The specification therefore inverts the usual convention of scoring sports:
Scoring sports call the score. This one calls the person.
The formats below are named rather than numbered, deliberately: one constraint stated at the end of this section must survive any future reordering of the list, and a rule that says "format 1" breaks the moment a format is inserted ahead of it.
Relay (the default, and the format a first deployment should ship). The system speaks the name privately to a single human relay — through that person's own device, audibly to them alone — and that person calls it aloud to the circle. Selection remains machine-held, so every distribution property of §4 survives intact; what changes is only who moves the air. The consequences are larger than the modification: there is no loudspeaker, no mesh, no clock alignment, and no requirement that participants own devices, so a single borrowed handset serves an entire circle. That matters most exactly where the sport is most wanted — a school that prohibits phones, a park with no signal, a group in which one person has a device.
Two further properties were not the reason for the arrangement and are now the reason to keep it. First, the relay hears the participant-approved pronunciation immediately before producing it, which makes correct articulation a modeled repetition rather than a reading task; a name read off a screen is mangled precisely for the names this design exists to protect. Second, and more consequential: the relay produces names aloud, where a player only ever hears them. The role is therefore where the sport's stated purpose is actually practised, and it follows that the role must rotate under the same equal-yet-random assignment that distributes receptions. Left to the circle, the confident participant holds the device and the quiet one never says a name — picked-last rebuilt through the back door of the very role meant to distribute the practice. Rotation also disposes of the objection that the relay is excluded from play: the role is a turn, not a sacrifice. The honest cost is a tempo ceiling — the repeat beat is absorbed at long lead times and binds at short ones — which is what the synchronized mesh (§7) is for, as a later upgrade rather than a launch prerequisite.
Call — the name format, spoken to the circle. The system addresses the whole circle directly over the mesh in one voice for every name — never in the named participant's own recorded voice, for the reason given at the end of this section — and counts silently. The count is visible, not audible: the mesh already places a screen in every pocket, and adversarial mode's countdown must remain readable for tension and tactics. Silence is not concealment. Numbers are spoken only at the rationed celebration thresholds of §6, which means a spoken number is always a rite rather than ambient noise. The chant, if the circle wants one, remains entirely human — the system never leads it and never competes with it. An accessibility setting must permit a spoken count, off by default, for players who cannot read the screen; the pre-announcement of §4 is already an auditory pre-cue and the same constituency is served by both.
Free — the count format, with the caller off. The caller is disabled entirely; the system counts aloud with the circle, and nobody is told where to pass. This is free play with a shared, trusted counter — and it exists because the design's central mechanism is unproven (§9, §11). It is also why the microphone tier is specified: with no call, the system has nothing to count from unless it hears the strike. The two are one feature.
The formats are mutually exclusive by construction — a session runs exactly one — which resolves the apparent contradiction among them. The rule was never that numbers are undesirable; it is that the name must never compete with numerals — one voice per channel, and the channel's content follows what the mode is for.
Across every format, one constraint governs articulation rather than selection, and it follows directly from §1:
The voice that says a participant's name must never be that participant's own.
The constraint has to be stated because the obvious implementation violates it. A system that teaches correct pronunciation will hold a recording of each name made by its owner — that is the only honest source of the pronunciation — and at call time the nearest artifact to hand is exactly that recording. Playing it delivers to each participant a recording of themselves saying their own name. The result is not a poor address; it is not an address at all, and no amount of audio quality repairs it.
Naming the constraint separates two speech acts that a single artifact had been performing at once, and which have different rightful speakers:
| The pronunciation | The call | |
|---|---|---|
| Whose is it? | the name's owner | anyone but the owner |
| What is it for? | how the name sounds | you, now |
| Failure mode | anglicization | self-address; no speaker |
| Where it lives | a stored model, approved by its owner | the moment of play |
The two are joined by transfer, not by identity: the owner's authority over how the name sounds is carried into a voice that is not theirs. Two mechanisms accomplish this, one per format. In Relay, the human caller hears the owner's own recording immediately before speaking, and reproduces it — modeled repetition, the highest-fidelity transfer available, and the reason the default format is also the best one. In Call, where every participant is a player and no human other is free to speak, the owner supplies a respelling — the name written as it sounds — which the single system voice articulates. The respelling is not an accessibility fallback; it is the transfer mechanism itself, and the only route by which a participant's authority over their own name reaches a synthetic articulator.
One voice for every name is a property, not a limitation. A circle whose calls were assembled from twelve different recordings would address each participant in a different timbre — the texture of being called would vary by whose recording happened to be playing. A single voice calling everyone means every participant is addressed identically: equal-yet-random rendered in sound. The uniformity of the acoustic channel is the audible form of the fairness the caller already guarantees numerically.
A note on what this does not claim. A system voice is an other, which is what the premise strictly requires, but it is not a person, and §1's account of why a name is sweet — evidence that someone spent attention on you — is only partly satisfied by a machine. This is a real gap and the specification does not paper over it: it is the reason Relay is the default format and Call is the device-plentiful case, rather than the reverse. Where a human can say your name, one should.
The gap above admits a partial repair, specified here as an option within Call rather than as a fourth format, because it changes which audio plays and not what anyone does during a rally.
The intuition is that a circle calling itself in its own members' voices would be warmer than a circle called by one synthetic voice, and it is correct. The naive implementation of it is nonetheless barred by the constraint above: playing each participant the recording they themselves made is precisely the mirror this section exists to prevent, and the warmth is illusory because the one voice a participant never hears as an address is their own. The intuition survives; the implementation must be inverted.
The mechanism is a cross-recording chain. At join, after a participant records their own name — the pronunciation model, which stays theirs — the device passes to the next participant, who is shown that name, hears the model, and says it themselves. That second recording is the call audio. The device continues around the circle, so each participant records exactly one other participant's name and is recorded by exactly one other participant: a cyclic permutation of the roster, costing one additional beat of roughly three seconds per person and producing exactly N artifacts rather than N².
Four properties follow, none of which the uniform voice provides:
The owner's approval (§5.3, and claim 21) governs the call audio as it governs every other pronunciation artifact: a chain recording a participant has not approved does not enter play, and the participant may ask for another take without stating a reason.
When to offer it, and when not. The two voicings answer different circles and the specification declines to rank them absolutely. Among strangers, the uniform voice is the better default: an identical acoustic channel is itself a fairness signal, and a newcomer has no relationship to the voice that would say their name — a stranger's recording of it carries no more warmth than a machine's while carrying more opportunity for mockery. Among people who already know each other — a family, a classroom after its first week, a returning circle — the chain is better, because the voices mean something and the sport's remaining work is warmth rather than fairness. The honest cost is stated plainly: the chain forfeits the addressed-identically property, and a circle that adopts it has traded a fairness signal for a warmth signal, which is a reasonable trade only once fairness is no longer the live question.
Two failure modes and their handling. A participant who leaves mid-session takes their recording's referent with them but not their recorded voice; a participant joining late has no one in the chain. In both cases the affected name falls back to the uniform voice for the remainder of the session rather than to the participant's own recording — the constraint holds under degradation, which is where constraints are actually tested.
Three constraints on Free, because it switches the sport's dignity guarantees off. Equal-touch distribution and the cannot-target-weakness property both live in the caller; with the caller off, humans choose receivers again and the picked-last dynamic returns. Accordingly: (a) a caller-on format remains the default everywhere; (b) Free must never be represented to an institution as equivalent — the anti-exclusion properties an administrator approves are precisely the caller's, and a free-play session does not carry them; (c) any trial of the caller mechanism must run a caller-on format — Relay or Call — or it measures nothing. That constraint is stated by property rather than by position in this list, which is why the formats carry names: a future revision may add or reorder formats, and the rule must not silently come to designate the wrong one. On which point see §9.1: Free is not merely a fallback but the natural control arm for the experiment that tests the caller itself.
Mixed-ability play fails two ways: with no handicap, the game is a rout; with a visible skill-rating handicap, the game is fair and humiliating — the rating announces precisely how much worse you are, to two decimal places, forever. Golf's handicap system, the great success of the genre, threads the needle partly by its own crudeness and its social conventions; explicitly skill-rated matchmaking (chess Elo, ranked ladders in esports) threads it in the opposite direction, by making the rating the point. A family sport needs a third path.
The start-line handicap. In net mode, the two sides may begin at different start numbers — the stronger side at 33, the beginners at 21; first to zero still wins. The handicap is public, singular, and ceremonial: one number per side, announced before play, arguable by nobody because chosen by no player (the system sets it, and its rubric is one-sentence-explainable: fairness plus desired game length). It is applied at the start line, never in the stream: once play begins, the caller and umpire treat both sides identically (§4's hard constraint), and any correction of a misjudged handicap waits for the next game — the rubber band lives between games, never inside one. A mid-game rubber band (catch-up mechanics, dynamic difficulty) is the standard videogame solution and is rejected here by name: dynamic mercy is detectable, and detected mercy is humiliation delivered by an algorithm.
Quantization-for-dignity. Start numbers are drawn from a coarse fixed lattice: multiples of three only (3, 6, 9, … 21, … 33). The coarseness is the point. If start numbers could be any integer, the pair (26, 24) would communicate a precision measurement — a skill rating in countdown clothing, recreating exactly what the start-line design exists to avoid. Quantized to the lattice, the handicap can only say what a garment size says: roughly this, roughly that, next size up. The minimum expressible gap (three) is large enough to be read as a category rather than a diagnosis. Handicap design ordinarily treats coarseness as a loss of precision; this design takes the opposite view — in social play, quantization is a dignity mechanism — and registers the inversion as a claim (Claim 4). Two free properties follow: the start number doubles as the duration dial (a park quickie from 9, a festival final from 33), and the lattice happens to be the institution's own numerological signature (three, the heart; thirty-three, the founding number) — a brand fact noted for honesty and carrying no mechanism weight.
The difficulty dials live in the world, not in the system. The anti-rating guards of §4 constrain what the system may hold; they do not constrain what the circle may vary, and a dial the system cannot see cannot rate anyone. Two follow, neither visible to the caller.
Three constraints govern both, and the first is the one an implementer will not derive. Declare what, never why. The sharpest risk at these dials is not fairness but accessibility: if one hand only becomes a legible difficulty badge, then a player who is one-handed by anatomy reads as playing hard mode, and the circle has been invited to read a body as a declaration. The surface says one hand — you choose which, never your weak hand. You narrow your own: a circle may step down together, but each player chooses what they give up, or a shared restriction excludes exactly the people mixed-method play exists to include. And at the start line, never in the stream, as with every other handicap here.
The synced-circle mesh. The call must be heard outdoors, over wind and traffic, by everyone at once — the mundane detail on which the whole design stands or falls in a real park. The solution is already standing in the circle: every player's phone joins the session and speaks the call in unison. N players are N loudspeakers arranged, by the game's own geometry, in a ring around the play — a surround system that assembles itself and scales with the circle. The mesh deletes the alternative's failure modes (a single device at the circle's center is a trip hazard, a theft target, a single point of failure, and one weak speaker), and it enables a property no single device could offer: the multilingual circle. The call — the name — is one utterance in one voice; but score announcements and celebrations can be rendered per device, in each phone's own language. A grandmother hears the milestone in Khmer as her grandchild hears it in English, in the same circle at the same moment. For diaspora families — the institution's first constituency — the feature is not cosmetic; it is the difference between a game the whole family plays and a game the grandchildren play while the elders watch. The engineering cost is stated honestly: multi-device simultaneous playback requires tight clock synchronization or the call smears into echo; local time-sync is a solved class of problem, but it is the mesh's hardest component.
The two-stage hardware split. Version one is software plus any ball on earth — the app is the caller, umpire, and equalizer, and the ball is whatever the circle owns. Nothing in §3–6 requires instrumented equipment. The manufactured B-Sey™ ball — designed for the foot-and-paddle dual strike, and eventually voiced, so the call emanates from the play itself — follows as a second-stage artifact, and brings with it the ball chronicle: the ball carries a durable identity (per this corpus's Proof of Coordinate℠ publication) and keeps its own biography — the circles it has played, the records it has witnessed, the dedications spoken over it. The chronicle records collective history only (a ball remembers parks and streaks, never individual performance), and a well-traveled ball becomes what sporting goods rarely are: a commons artifact whose value grows with use, passed between circles as a gift that carries its own story. The split is a deliberate adoption strategy: the sport must never gate on hardware; the hardware, when it comes, must only deepen a game already alive.
Current form. Two parts of this section have been superseded by the design as now specified. The synced-device mesh is not on version one's path: the relay call (§5.2) is the shipping default, runs on one device, and needs no clock alignment, and the mesh is the later upgrade that serves Call. And the object carries no speaker: §7.1 reduces its job to an accelerometer, a clock and a radio used as a messenger, since the circle's devices already speak. The mesh-first plan and the voiced ball described above are retained as disclosed variants.
The entire game runs on one device, offline, with no accounts. Names, pronunciations, pacing, both scoring arrows, start numbers, and celebration thresholds are local state; the complete data model of a thirty-player circle with recorded pronunciations is on the order of a megabyte. Only genuinely shared artifacts require a network — cross-circle discovery, records that travel between places, and the object's own biography.
This is not primarily an economy measure. It is the architectural expression of constraints the design already carries: no per-player statistics, no analytics on minors, chosen names never leaving the circle, and a private-mirror rule for any per-player derivation. Each of those independently says keep it on the device. That the result is also nearly free to operate at any scale, and that it works in a park with no signal, a school with a locked-down network, and a village on intermittent data, is a coincidence of goods rather than a trade-off.
Two engineering notes. The synchronized announcement requires device-to-device transport, not a server — the platform peer-to-peer facilities (Apple's MultipeerConnectivity; Android's Nearby Connections) are designed for exactly this. The genuine risks are cross-platform pairing, since any real circle is mixed, and clock alignment, since simultaneous playback across N devices reads as an echo rather than one voice beyond roughly twenty milliseconds of skew; a leader device scheduling playback at a negotiated offset is the standard remedy, with single-announcer fallback — never a deliberate stagger — if alignment resists. And the classroom deployment of §9.1 requires no mesh at all, so the hardest component is not on the critical path to the design's primary channel.
The object's job is smaller than it appears. Because the caller assigns identity (§5.1), an instrumented object need not report where it is or who holds it. Radio-signal-strength proximity cannot in any case discriminate between players standing one to three metres apart with bodies intervening; precise ranging radios exist but are confined to premium handsets, which would breach the design's zero-floor commitment. The object therefore needs an accelerometer and a clock, and uses its radio as a messenger rather than as a tape measure — reporting {object_id, touch, t} for the system to pair with the call it has just issued. The consequence is that instrumentation costs the price of a motion sensor rather than a positioning system, and that the object needs no speaker at all, since the mesh already speaks — deleting the most expensive and least durable component from an object designed to be kicked.
Two constraints follow that an implementer will not derive on their own. First, an object broadcasting a stable identifier is a tracking beacon, and children are the primary population; the identifier must rotate publicly while the object's durable identity is held cryptographically and locally, so that the biography of claim 10 survives without the object becoming a tag on a child. Second, and more subtly: once touch events exist and the system knows who was called, a per-player success table becomes trivially computable — the exact artifact §4 and §12 forbid. The guard must therefore be restated at the hardware layer: object telemetry aggregates to the circle, never to a player scoreboard. The general lesson is worth stating plainly for anyone instrumenting a sport: adding a sensor can silently re-open a guarantee that was previously enforced by the absence of data.
Stated as the shared spine of this sprint and applied here in its simplest form: competition and play may generate the story; only witnessed giving moves the currency. In sey:
An extractive implementation of this paper's mechanisms — an AI caller with engagement-optimized distribution, player-skill surveillance, streak leaderboards, and rated matchmaking — is entirely feasible and is the obvious product a games incumbent would build from the same parts. The claims below are published precisely so that the dignity-preserving configuration is prior art: the caller that never evaluates, the handicap that cannot humiliate, the record that has no names in it.
A sport's mechanisms are worthless if nobody plays. The design's adoption thesis is stated in one line for the record: the sport's path to scale is the utility path — it addresses a non-sport problem (strangers becoming friends; names being learned; nobody excluded), so institutions that adopt tools (schools, workplaces, festivals, parks departments) can carry it to places marketing never reaches. Its floor is very low (any ball, any space, a free app, rules learned in seconds because the call grammar is Simon Says' — one voice names who must act, everyone else waits — which every child already knows); its mixed-age property is pickleball's growth engine, engineered threefold; its streak is a native short-video format; and its ceiling — the acrobatic kick — is inherited from sepak takraw, whose brutal floor kept a spectacular sport regional and whose lesson (spectacle recruits, access retains) this design encodes as the paddle. The full ranked analysis is not published here; this paper registers only the wager and its instruments.
The predictions, pre-registered at the design layer (n = 0; every instrument awaits the first pilots):
The adoption analysis behind the thesis above (not published here) ranked schools fifth. On the analysis below they rank first, and the reason is a general one worth stating separately from this sport.
A product that claims to ride an existing behaviour must name a behaviour a target adult actually performed, unprompted, in the last twelve months — and the host must also match the product's commitment level and be in good standing with whoever must permit it. Run that test across this sport's candidate channels and exactly one host passes all three parts. Park circles are a new habit: nobody currently goes to a park to play a called circle game, so that channel must be budgeted as habit formation. The classroom is different. Teachers commonly face the same two problems in the first week of every school year — learn thirty names, and include every child — and many already run a name game to solve them, unprompted and annually. Elementary morning-greeting routines are a standing institutional ritual. Recess is daily. The commitment shape matches almost exactly: ten to fifteen minutes, free, no equipment, teacher-led, repeatable. And the standing is positive rather than merely tolerated: name learning, inclusion, and anti-bullying are things administrators actively want, which is the inverse of the polarity most unfamiliar objects carry into a school.
The reframe this forces is the practical conclusion of the whole paper. The behaviour this sport attaches to is not sport at all; it is the name-learning icebreaker — performed at school openings, orientations, retreats, onboardings, receptions, and newcomer groups, widely, and widely disliked. Its incumbent competitors are the seated round of self-introductions and the disclosure game. Against those, a called circle wins on every axis that matters: twenty exposures to each name instead of one, movement instead of standing, nothing to disclose about oneself, and no performance in front of strangers. The sport ships as an icebreaker and is discovered as a sport — and Friendship mode at long lead time (§3.1, §4.1) is that icebreaker, arrived at from the sport's own logic rather than designed for the channel.
Rosterless classroom mode. The classroom deployment is specified as single-device: one teacher's device speaks; no child device participates; the roster is entered once by the teacher and reused; no accounts exist for minors, no analytics are collected, nothing is uploaded, and the roster is local and erasable. This is not compliance decoration — a roster of children's names and voices is precisely the artifact that fails an institutional privacy review, and the permitting authority and the privacy authority are the same office. It also happens to remove the mesh, the network, the accounts, and the instrumented object from the primary channel: a teacher, a phone, and a soft sack.
The pronunciation commitment (§4, claim 9) is satisfied in this mode without recording anything. At long lead the called child raises a hand and says their own name, and the circle repeats it; the scaffold is itself the pronunciation-capture mechanism, performed by a human. The circle converges on the child's own delivery rather than a synthesizer's approximation — which is the correct outcome in the venue where mispronunciation is actually inflicted, and which requires no biometric capture to achieve.
Two further pre-registrations. The design's existential uncertainty (§11) is whether being told where to pass is delightful or officious, and only play answers it. The classroom supplies an unusually clean and cheap instrument:
P-S5 costs one teacher, one class, one sack, and two weeks, and its readout cannot be spun: unsupervised children do not organize a game out of politeness. Free (§5.2) supplies the control arm — a second class playing free play with the same object and the same counter. The comparison isolates the caller itself, which is the single variable the entire design rests on and the one no amount of design reasoning can settle.
The economics follow the architecture. Because §7.1 puts the game on the device, the marginal cost of a player is approximately zero, and only the shared layer — cross-circle discovery, travelling records, organizer tooling — has a server behind it. The free tier is therefore the entire sport, permanently, and revenue must come from somewhere that is not the player.
The specification is place sponsorship: a patron funds the shared layer for a geofenced public place — a school, a park, a neighbourhood — and every circle that plays there receives it. The individual player never pays, at any tier, ever.
The scoping test is non-excludability, and it is the whole rule:
A place is sponsorable if and only if the sponsor cannot exclude anyone from it.
Not does the sponsor benefit — they always do, and should — but could the sponsor shut the gate. A private residence fails: the door is controlled, the good is excludable, and the payment is a purchase wearing a patron's coat. A neighbourhood, park, school, plaza, or temple courtyard passes: the good is non-excludable, so the payment remains a gift even when the sponsor's own family plays there.
Four constraints are specified with it, and the first is the one an implementer will get wrong:
The resulting circuit is worth naming because it is the reason this revenue model is not a compromise with the mission: a patron funds a place, the circles that play there thank the patron, and the thanks funds the patron's next gift. The payment mechanism is itself an instance of the thing the institution builds.
Current form. Sponsorship as now specified works at city scale rather than per place. Every city carries the same fixed number of sponsor chairs (twelve), no sponsor may hold more than a third of them, and chairs are priced on a coarse four-rung ladder; when a city is oversubscribed, chairs are allocated by sealed bids cleared at the highest rejected tier, with admission at the clearing tier by lot from a public randomness beacon, so that anyone can recompute the allocation. A sponsor receives acknowledgment — a name and dates, oldest first, in that city's permanent record of those who kept it — and thanks routed into forward-spendable gift capacity; there is no control, no data, no exclusivity, and no bidding on child-facing places, which are funded by invitation outside the market. The same design treats sponsorship as a way to seed local play, not as the sport's revenue: the sport and its apps are free. The place-scoped model above is retained as a disclosed variant.
The circle games. The direct ancestors are the Southeast Asian circle keep-alive family: the Khmer sey, played at dusk in courtyards and parks across Cambodia; Myanmar's chinlone, a six-person circle game explicitly organized around beauty and collective flow rather than opposition — the world's great existing proof that non-adversarial ball sport can sustain a culture for centuries; China's jianzi and the global hacky-sack/footbag family. The sport is a descendant of this lineage and says so in its name, sey; what the lineage lacks — and what kept it local — is an equalization layer (the skill floor of foot-only play is high) and any addressing mechanism (circles of strangers stay strangers). Sepak takraw contributes the net form and the spectacular ceiling, and the cautionary lesson of a magnificent sport whose difficulty floor confined it to its region.
The designed prosocial sports. Sport design as deliberate social engineering has a distinguished small lineage: korfball (1902), which mandated mixed-sex teams by rule; tchoukball (1970), designed by Hermann Brandt from an explicit medical-humanist brief to eliminate aggression — its no-interference rule is the closest ancestor of the cannot-target-weakness property, though it removes contact rather than targeting; kin-ball (1986), designed for total participation with its all-touch service rule — the nearest prior art to equal distribution, achieved by static rule rather than adaptive caller; Special Olympics Unified Sports and the walking-football/walking-netball movement, which achieve mixed-ability play by restricting the able rather than equalizing the field. The Ultimate frisbee community's spirit-of-the-game — self-officiating as ethos — is the officiating seat's spiritual ancestor; sey completes what spirit-of-the-game gestures at by making the neutral judge structural.
The handicap systems. Golf's handicap (recorded as a practice from 1687 and formalized from the 1890s) is the great proof that start-line equalization can carry a mixed-ability sport for more than a century; go's fixed-stone handicaps and chess odds-giving are older still. All are lattice-coarse in practice, none that the authors have found treats coarseness as a dignity mechanism (no census has been run on this point), and none operates in a team net sport. Videogame dynamic-difficulty adjustment and matchmaking ratings are the anti-examples engaged in §6: fairness delivered as either hidden mid-stream mercy or public skill measurement, the two failure modes the start-line design refuses.
The caller's ancestors. Simon Says supplies the call grammar — a named authority directing named players — as a children's commons centuries old; square-dance calling is the adult form: a caller directing the figures of a cooperative physical game in real time, and the closest single ancestor of the caller seat (what square dance does with figures, sey does with receivers). Neither selects whom adaptively; neither equalizes; both prove the pleasure of called play.
The name games. Names are already called in circles. In the soccer coaching drill known as the Passing Name Game, the passer calls the name of the teammate they have chosen to pass to; in the Name Toss icebreaker, players call a name and toss, and the round closes once everyone has received. Both are recorded in the census of the companion paper Circulation Sports, and they are the nearest prior art to claims 1, 3 and 9. The difference is who calls: in both, the passer names their own choice, so selection — and with it targeting and favouritism — stays in human hands, and the Name Toss buys full coverage with an order that repeats. Neither announces receivers from outside the play, and neither carries a round across sessions.
Name-pronunciation tools. Recording a name's pronunciation in its owner's voice is established practice: dedicated services (NameCoach, NameDrop, SayMyName), university directories, and professional-profile features (LinkedIn; Microsoft 365 profile cards) all offer it, and a genre of consumer apps helps a user remember other people's names and pronunciations. What claim 21 adds to that practice is the owner's approval act, the non-retention of which capture path was used, and the transfer of the owner's authority to a different articulator at call time.
The draw beneath the caller is not new, and this paper claims none of it. Every leg of the arithmetic §4.2 and §4.4 describe was public before this sport used it: selection that is seeded, published and recomputable by anyone (Eastlake's RFC 2777 and RFC 3797, 2000 and 2004); the equal-count bag (rotating savings associations that order the pot by lot, formalised as the random ROSCA by Besley, Coate and Loury in 1993; permuted-block randomisation in clinical trials, whose literature already records the bag's cost that §4.4(a) states — the last turns of a block are predictable from the first; the Tetris seven-piece bag, from 2001); the keyed per-span permutation that v2 uses (Ethereum's beacon-chain committee shuffle; Chainlink's Off-Chain Reporting 3.0); and v2's boundary swap itself, in a Rust shuffle-bag crate published in April 2025. The census that established this, and the generalisation of the draw beyond play — to recurring benefits, to a lot form, and to a sealed regime for an operator whose own gifts must stay anonymous — is the companion paper Decided by No One, which names this sport as the primitive's origin. What claims 3, 17 and 23 place in the public domain is the draw's use as a sport's caller, and the constraints that use imposes: named receivers, no conditioning on evaluated play, count-only coherence, round continuity, and no per-player record.
The instrumented-sport patent field. Smart balls (Adidas miCoach, Wilson X and successors), computer-vision officiating (Hawk-Eye, VAR, automated ball-strike systems), and AI sports-coaching applications (HomeCourt and the computer-vision training genre) constitute an active commercial and patent field moving fast in precisely the opposite direction from this paper: toward finer individual performance measurement. The field is engaged respectfully as the reason this publication exists now: an AI-directed pass-calling mechanism with adaptive distribution is an obvious near-future patent claim in extractive form (engagement-optimized, stat-generating), and the present publication places the mechanism class — and specifically its dignity-preserving configuration — in the public record as prior art.
The third-place literature. Oldenburg's The Great Good Place and Putnam's Bowling Alone frame the vanishing social infrastructure the sport addresses; the recent public-health literature on loneliness (including the 2023 U.S. Surgeon General's advisory) supplies the epidemiological register. Bowling leagues were Putnam's emblem of what was lost; a self-assembling park circle that teaches names is offered, without irony, as a candidate replacement part.
The name-and-belonging literature, and the side of the relation it sits on. The claim that names carry belonging is not this paper's invention, and the existing evidence sits almost entirely on the instructor's side. Cooper et al. (2017) found that students in a high-enrollment biology course who perceived that an instructor knew their names reported better attitudes toward the course — the construct is perceived knowing, narrower still than knowing. Kohli and Solórzano (2012) establish the negative case in the identity register, documenting name mispronunciation in K–12 classrooms as a racial microaggression with durable effects on how students relate to their own language and family. Walton and Cohen (2011) is the outcome-side anchor and not a name study at all: a brief social-belonging intervention produced academic and health effects measurable three years later, which is the reason to believe belonging is movable by small structured means in the first place. Goodenow's (1993) Psychological Sense of School Membership scale is the validated instrument P-S7 and P-S8 specify rather than inventing one.
What that literature does not contain is the relation this sport operates on. All of the above run teacher → student. Peer-side work establishes that perceived peer inclusion predicts belonging, but not that a name intervention produces it. The authors are aware of no controlled study of student-to-student full-coverage name knowledge — every student able to name, and correctly pronounce, every classmate — which is the completeness condition equal-yet-random produces as a by-product, and which the practitioner literature on icebreakers and name games asserts freely without testing. That gap is P-S7's subject. It is stated here as a gap, not as a discovery: it is the kind of question an education researcher is far better equipped to answer than a sport designer, and this paper would rather be corrected than cited.
This section is deliberately free of lineage, resonance, and metaphor.
The following mechanisms and architectures, individually and in combination, are published as prior art and dedicated to the public domain:
A coincidence the design did not arrange. The Khmer សី is pronounced close to the English say, which lets sey circle read, to an English ear, as a circle of sayings — and the sport turns out to be exactly that, since a completed pass is a name said and answered. The two words are unrelated: Khmer and English do not share this root, and the reading is a gift of one language rather than a fact about the other. It is recorded because it is pleasing and because the paper would be poorer for pretending not to have noticed it, and for no other reason. Strike this paragraph and every mechanism above stands untouched.
The lineage of this design is a courtyard at dusk. Across Cambodia, in the hour when the heat breaks, people stand in circles and keep a small woven ball in the air — no teams, no referee, no score that anyone writes down, only the shared project of not letting it drop. The game survived what Cambodia's twentieth century did to everything else that was gentle. This paper's whole mechanism is, in one reading, an attempt to give that circle three things the modern world took from its conditions of possibility: a fair way for the unequal to stand in it together, a voice to welcome strangers into it by name, and a record that remembers the circle rather than sorting its members. The sport is named for that game, and the name performs its own thesis — sey (សី), the inherited circle; say, the spoken name; a children's call-game hiding inside both. A note on the script, since the pun is a claim about a language and should be checkable in it: សី is sey, and the institution's rabbit mascot elsewhere is ទន្សាយ, tonsay. The two are unrelated words that merely share a syllable once romanized, which is worth stating plainly: in Khmer the resemblance is visible as coincidence, and only in Latin letters can it be mistaken for an etymology. The pun is offered as a pun. What the caller calls, in the end, is the oldest social technology there is: you, here, now — the ball is coming to you.
Coined names used in this paper and the standard terms an examiner would search for them.
| Term used here | Standard technical term |
|---|---|
| sey (the sport); B-Sey℠ app; B-Sey™ ball | circle keep-up ball game (footbag / shuttlecock-kicking family), with net and cooperative variants; its software and ball |
| the caller; the caller mechanism | automated turn assignment: receiver selection announced by spoken name before each play (automated pass calling) |
| equal-yet-random distribution | shuffle-bag scheduling: a fresh random permutation per round, giving equal turn counts per round in random order |
| determined caller | deterministic pseudorandom sequence from a declared seed and roster; publicly verifiable, recomputable random selection |
| count-only circle coherence | multi-device consistency by exchanging only a shared sequence index (call counter) |
| round continuity | persistence of the round-robin index across sessions |
| closer effect; boundary swap | position bias at the boundary between consecutive shuffles; no-immediate-repeat constraint implemented by swap rather than rotation |
| relay call (Relay format) | two-stage announcement: private audio prompt to one rotating human announcer, who speaks the name aloud |
| Call format; Free format | synchronized device text-to-speech announcement; free play with automated counting and no receiver selection |
| synced-circle mesh | clock-synchronized simultaneous audio playback across participants' mobile devices over peer-to-peer links |
| address constraint; uniform calling voice | rule that a participant is never called in their own recorded voice; one speech-synthesis voice for all names |
| pronunciation-authority transfer | owner-approved name-pronunciation recording or respelling used to drive another speaker or a speech synthesizer |
| cross-recording chain | cyclic assignment in which each participant records one other participant's name |
| cannot-target-weakness property | receiver designation on both sides of a net game, removing opponent selection |
| dyadic scoring unit | pass-completion scoring credited to no individual |
| dual-arrow ruleset | ascending cooperative streak count; descending (countdown-to-zero) competitive score |
| start-line handicap; quantization-for-dignity | handicap by unequal starting score, quantized to multiples of three |
| last-gift rule | finishing rule requiring the final point to be won, not conceded by an opponent's error |
| call-lead-time ladder; scaffold fading | difficulty set by the interval between announcement and ball arrival, with a raised-hand identification cue at the longest interval |
| method as declared handicap | self-declared striking method (hand, paddle, foot) as a pre-play handicap |
| circle-scoped adaptation | group-level, never per-player, adaptive difficulty |
| implied point; evidence tiering | score inferred from the call, optionally validated by k-of-N microphone transient detection or an accelerometer in the ball; records labelled by validation tier |
| acoustic inversion | names announced aloud, score shown visually |
| collective-only chronicle | group- and equipment-level records with no individual performance statistics |
| rosterless classroom mode | single-device, account-free, local-only deployment for minors |
| place sponsorship; sponsorable-place test | geofenced patronage of a shared service, admissible only for non-excludable public places |
| host-agnostic turn-assignment layer | the turn-assignment mechanism applied to any group activity |
| determined and judged seats; seat-migration bar | separation of recomputable algorithmic functions from perceptual judgment functions in automated officiating |
| play/currency wall | rule that game outcomes never create, move or gate monetary value |
0a. Carnegie, D. (1936). How to Win Friends and Influence People. Simon & Schuster. (The premise's popular statement — cited at lens tier and explicitly inverted in §1: offered there as a technique for advantage, kept here only for the recognition.) 0b. Moray, N. (1959). "Attention in Dichotic Listening: Affective Cues and the Influence of Instructions." Quarterly Journal of Experimental Psychology 11(1). (The own-name effect in the unattended channel.) 0c. Wood, N., & Cowan, N. (1995). "The Cocktail Party Phenomenon Revisited." JEP: Learning, Memory, and Cognition 21(1). (Replication and bounding of own-name capture; the narrower finding §11 distinguishes from the paper's broader premise.)
need-compiled-questlines). (The mints-nothing class; claims division per the scope note.)@333eco/sey 1.1.0 (2026). The caller sequence — a cross-platform contract, §8, with conformance vectors; and @333eco/primitives, The called draw — a recomputable contract, Parts I–II. CC0 packages. (The normative text of both caller versions; §4.4 summarises and they govern.)— End of paper —
Marks referenced: HeartBank®, Miss Aquarius℠, B-Sey™, B-Sey℠, B-Match℠, B-Grace℠, Proof of Coordinate℠. Document License: CC0 1.0 Universal. The authors and HeartBank® will not seek patent on this specification or any portion thereof. This document constitutes a defensive publication establishing prior art as of its date.