Formal Specification
Yoruba Traditional Singing Hide-and-Seek — Formal Specification
Nigeria · hide-seek · Confidence: low
← Back to game pageImplementation Suitability
Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 2
# Bójú Bójú — AI-optimized specification
## General idea
The Yoruba hide-and-seek game, structurally distinguished from its regional relatives by
**where the round stops**: it ends at the *first* tag, not when all hiders are found. The
seeker is not clearing a board — they are hunting for exactly one person to hand the role to.
The chant that opens the round is also a functional timer, since its fixed call-and-response
length defines precisely how long the hiders have to scatter.
## Formal model
- **Abstraction taken:** Concealment, visual search, and sprinting are collapsed into
stochastic checks: a spot probability while the seeker searches, and a race outcome once a
hider is named. The entry describes these qualitatively with no numeric model. The chant
(entry §2), by contrast, is treated as an exact fixed-length hiding window, since its text
is fully specified.
- **Player/agent count:** 1 Seeker (*Olóró*), K Hiders, K in [3, 14] (entry: 4-15 children
total).
- **Turn model:** A sequential chant phase, then synchronous ticks during the search.
## Formal rules
R1. Setup (entry §1): a home base — a tree or wall, *Ilé* — is designated.
R2. Chant phase (entry §2): the Seeker covers their eyes at the base and performs the fixed
call-and-response with the group: *Bójú bójú ó!* / *Olóró ń bọ̀!* / *Ẹ pamọ́ pamọ́!* /
*Ṣé kí n wá?* answered *Wá!* This exchange is the hiding window and has fixed length.
R3. Hiding (entry §3): during R2, all hiders scatter and conceal themselves. A hider not
concealed when the chant ends is `exposed` and subject to immediate spotting
[gap-fill: the entry does not state a consequence for failing to hide in time; treating
such a hider as simply easier to spot, rather than automatically caught, is the minimal
reading].
R4. Search (entry §4): each tick, the Seeker submits `search(direction)`. Any concealed
hider within range resolves a `spot_check`; on success the Seeker calls that hider's
name aloud, which is a required part of the mechanic, not flavour.
R5. Race (entry §4): once named, the hider and the Seeker race — the Seeker to `tag()` the
hider, the hider to `touch_base()`. This resolves a stochastic race outcome. A hider who
reaches base first is **safe** for the round.
R6. Round end and role transition (entry §5): the **first** hider tagged becomes the Seeker
for the next round, and the round ends at that moment. Hiders still concealed are never
found, which is a normal outcome rather than a loose end.
R7. Match structure: none. The entry describes no score, no target, and no final round
[gap-fill: the number of rounds played is a social decision the entry does not encode].
## Edge cases
- Tie at the end: not applicable — R6 ends the round at the first tag, a single event with a
single subject, so no two hiders can be "first tagged" and there is no score to tie.
- A player with no legal move: not applicable — the Seeker always has `search`; a concealed
hider always has `stay_hidden` or `dash_for_base`; a named hider always has `touch_base`.
- Repetition / cycles: the Seeker role rotates indefinitely across rounds (R6), and within a
round the Seeker may search fruitlessly for many ticks. Both are expected and structural.
- Simultaneous claims on one position: the base admits any number of hiders simultaneously —
the entry sets no capacity and safety is not exclusive, so arriving together is harmless.
The only genuinely contested event is R5's race, which resolves to a single winner.
- Elimination mid-round and turn-order repair: no player is ever eliminated. A tagged hider
becomes the next Seeker (R6) and a safe hider simply stops being at risk; both remain in
the game. The search phase is synchronous, so there is no turn order to repair.
- Resource exhaustion: not applicable — the game uses no equipment at all (entry: "No
equipment") and nothing is consumed. Hiding places are reusable across rounds.
- Termination guarantee: **not guaranteed at match level, though each round ends with
probability 1.** A round terminates almost surely: the play area is bounded, each tick
carries a fixed positive spot probability for any concealed hider, and each resulting race
has a fixed positive chance of ending in a tag — so a first tag occurs with probability 1,
satisfying R6. But the entry defines **no match end whatsoever** (R7): no score, no target
number of rounds, no win condition. This is continuous social play, and
`termination_guaranteed: false` records that honestly rather than inventing a stopping
rule the tradition does not have.
## State representation
- Persistent (session) state: `seeker_id`, `round_number`.
- Per-round state: `hider_status[h]` ∈ {hiding, concealed, named, safe, tagged},
`chant_ticks_remaining` (during R2 only).
- Per-tick ephemeral state: `positions` (Seeker, and any hider currently racing),
`named_hider`.
- Observable information: **partially observable, and the asymmetry is the game.** The
Seeker cannot observe concealed hiders' positions until a `spot_check` succeeds; hiders
can observe the Seeker's position, which is what lets them judge a dash for base. An
engine granting the Seeker full visibility has removed the game.
## Action space
- `chant(line)`: the Seeker during R2; the fixed call-and-response sequence.
- `search(direction)`: the Seeker; precondition: chant complete.
- `tag(hider)`: the Seeker; precondition: the hider is `named` and within reach.
- `stay_hidden()` / `dash_for_base()`: any concealed hider; always legal after R2.
- `touch_base()`: a `named` or dashing hider at the base; confers `safe` (R5).
## Playability notes
- Single-player-vs-AI: natural on either side. A human hider judging when to break for base
against an AI seeker's search pattern is the game's core decision; a human seeker
searching for AI hiders is a clean partially-observable search problem.
- Reducing K > 4 to 2 agents: workable but thin. With one hider, R6's "first tagged" becomes
simply "tagged", and the social texture of a group scattering to a shared chant — much of
the point — disappears entirely.
- Physical-skill games: concealment, visual search, and sprint speed are the entire physical
contest and are collapsed into two probability checks (R4, R5). No genuine line-of-sight
geometry is modelled, though it could be, and would improve fidelity considerably.
- What is lost: the chant itself, which is the game's name and its heart — the call-and-
response cadence teaches Yoruba vocal rhythm, and the entry situates the whole game as an
evening moonlight activity (*Erē Òṣùpá*). A spot-probability captures none of that.
## Implementation hints
Model line of sight explicitly if the environment permits it: this is one of the few games in
the corpus where a real geometric visibility model would meaningfully outperform a flat spot
probability, because hiding-place quality is the hiders' only strategic variable. Keep the
chant as a fixed-length timer rather than a formality — it sets the hiders' budget and
therefore how far they can get. The representational trap is R6: continuing the round until
all hiders are found turns this into a different game (its Armenian and Azerbaijani relatives
work that way) and roughly doubles round length. Suggested first milestone: the chant timer,
a flat spot check, and the base race, verifying that rounds end at a first tag and that a
meaningful fraction of hiders end the round undiscovered.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Yoruba Traditional Singing Hide-and-Seek (Nigeria). (n.d.). Games of Countries. https://gamesofcountries.com/games/ng-nigeria/boju-boju
@misc{gamesofcountries-ng-nigeria-boju-boju,
title = {Yoruba Traditional Singing Hide-and-Seek (Nigeria)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/ng-nigeria/boju-boju}
}