Formal Specification
Holding-Breath Tag — Formal Specification
Thailand · chase-tag · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 5
# Tee Jub — AI-optimized specification
## General idea
A traditional Thai team territory-raiding and breath-holding tagging sport related to *kabaddi*. Two teams occupy opposing halves of a divided court; alternating raiders invade enemy territory while maintaining a continuous vocal drone ("Tee-e-e...") on a single breath, attempting to tag defenders or rescue captured teammates and return across the centerline safely before their breath exhausts or they are tackled and restrained.
## Formal model
- **Abstraction taken:** The continuous court is discretized into a 2D grid of size `12 x 8` cells, bisected by a centerline at `x = 6` separating Team A's territory (`x in [0..5]`) and Team B's territory (`x in [6..11]`). A team's Prison zone is located at its respective back row (`x = 0` for Team A, `x = 11` for Team B). The continuous vocal breath is formalized as a discrete "Breath Counter" `B = 15` ticks per raid.
- **Player/agent count:** 2 teams of `K` players each (`K in [4, 8]`).
- **Turn model:** Alternating team raids; within each raid, simultaneous movement ticks between the active raider and the defending players.
## Formal rules
R1. Setup: Court grid is `12 x 8`. Team A starts with `K` active players in `x in [1..5]`. Team B starts with `K` active players in `x in [6..10]`. Team rosters maintain active and prisoner lists.
R2. [gap-fill] Raid Initiation and Breath Economy:
- Teams strictly alternate sending one active player as the Raider (*Fai Ruk*).
- The raider crosses the centerline (`x = 6`), initializing `breath_ticks = 15`.
- Each raid tick, `breath_ticks` decrements by 1. If `breath_ticks` reaches `0` while the raider is still in enemy territory, the raider is captured as a prisoner.
R3. [gap-fill] Movement and Tagging Mechanics (per tick):
- Raider chooses: `Move(dx, dy)` where `dx, dy in {-1, 0, 1}` (speed 1 cell/tick).
- All defending active players simultaneously choose: `Move(dx, dy)` or `Hold()`.
- **Tagging Defenders:** If the raider moves into an orthogonal or diagonal neighbor cell of an active defender, the raider may execute `Tag(defender_id)`. The tagged defender is marked as `touched = true`.
- **Rescuing Prisoners:** If the raider reaches an adjacent cell to their team's imprisoned members at the back line and executes `Touch_Prisoner()`, all touched prisoners are marked as `rescued = true`.
R4. [gap-fill] Tackle and Return Resolution:
- **Successful Return:** If the raider returns across the centerline into their home territory (`x` on home side) while `breath_ticks > 0`:
- All defenders marked `touched` are captured and sent to the raider's prison zone.
- All teammates marked `rescued` are restored to active play.
- **Raider Tackle / Capture:** If two or more defenders surround the raider (occupying adjacent cells) and execute `Tackle()`, the raider is immobilized. If the defenders maintain containment until `breath_ticks == 0`, the raider is captured and sent to the defender's prison zone; any `touched` defenders revert to safe active status.
R5. [gap-fill] Raid Progression and Inning Cap:
- Each raid ends with either a successful return, a breath-out capture, or a tackle.
- The next raid begins with the opposing team selecting a raider from their active roster in round-robin sequence.
- [gap-fill] Victory Condition: A team wins immediately if it captures all opposing players (a full sweep / *Kwad Mueang*).
- Hard Raid Cap: The match concludes after 20 total raids (10 per team). The team with the strictly greater number of active players wins.
## Edge cases
- Tie at the end condition: If after 20 raids both teams have equal active players, a sudden-death 1-raid tie-breaker (Golden Raid) decides the winner.
- A player with no legal move: Not applicable; active raider and defenders always have valid directional grid movements.
- Repetition / cycles: Bounded by the 15-tick breath timer per raid and the 20-raid hard match cap.
- Simultaneous claims on the same resource or position: When a raider touches the centerline on the same tick a defender initiates a tackle, crossing the boundary has legal priority and resolves as a successful return.
- Player elimination mid-round; turn order repair: Captured players are transferred to the opponent's prison list; team raiding order skips imprisoned members, rotating sequentially through remaining active members. Rescued players are appended to the active roster queue.
- Resource exhaustion: Active player rosters are bounded within `[0, 2K]`; the 20-raid match budget strictly decrements to zero.
- Termination guarantee: why the game provably ends: Each raid lasts at most 15 ticks, and the match is capped at 20 raids, guaranteeing total match duration cannot exceed 300 discrete ticks.
## State representation
- Persistent state:
- `active_players_A`: list of player IDs
- `prisoners_A`: list of player IDs held in Team B prison
- `active_players_B`: list of player IDs
- `prisoners_B`: list of player IDs held in Team A prison
- `raid_counter`: integer in `[1, 20]`
- `current_attacking_team`: `Team_A | Team_B`
- Per-turn ephemeral state:
- `raider_id`: player ID
- `raider_pos`: `(x, y)` coordinate
- `defender_positions`: dictionary of player ID to `(x, y)` coordinate
- `breath_ticks`: integer in `[0, 15]`
- `tagged_this_raid`: set of player IDs
- `rescued_this_raid`: set of player IDs
- Observable information: Fully observable public state (court grid coordinates, breath ticks remaining, active rosters, and prison populations are known to all agents).
## Action space
- Raider:
- `move(dx, dy)`: Step in 2D grid direction.
- `tag_defender(defender_id)`: Tag adjacent defending player.
- `rescue_prisoner(prisoner_id)`: Touch adjacent imprisoned teammate.
- `cross_home()`: Cross centerline to complete raid.
- Defender:
- `move(dx, dy)`: Evade or close in on raider.
- `tackle()`: Join adjacent tackle containment.
## Playability notes
- Single-player-vs-AI: Highly tactical team-sports game; human can play as raider (juking and route-finding against AI zone defense) or defensive captain (organizing containments).
- Group games (n>4): Natively designed for 4v4 to 8v8 team play; scales smoothly across court dimensions.
- Physical-skill games: Lung capacity and continuous vocalization are modeled via the discrete `breath_ticks` pool; physical tackling is formalized via multi-agent adjacency containment.
- What is lost in the adaptation: The sustained physical strain of holding one's breath while chanting *"Tee-e-e-e..."*, the sweat and grappling on dirt ground, and visceral high-speed tackles.
## Implementation hints
- 2D grid pathfinding and containment checks (e.g. check if raider is surrounded by >= 2 defenders with no open exit vectors).
- Maintain clean lists for active vs imprisoned players per team.
- Suggested first milestone: Implement the 12x8 grid with 15-tick breath countdown and test raider return vs tackle-capture resolution.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Holding-Breath Tag (Thailand). (n.d.). Games of Countries. https://gamesofcountries.com/games/th-thailand/tee-jub
@misc{gamesofcountries-th-thailand-tee-jub,
title = {Holding-Breath Tag (Thailand)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/th-thailand/tee-jub}
}