Skip to main content
Formal Specification

Dog and Bone (Handkerchief Snatch) — Formal Specification

Cyprus · chase-tag · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 3
# Siyllo tziai kokkalo — AI-optimized specification ## General idea A competitive symmetric 1v1 dash-and-capture duel embedded within a team scoring match. Two agents sprint from opposite baselines to a center point containing a single token; the agent who seizes the token must sprint back across their baseline while avoiding a tag by the other agent. ## Formal model - **Abstraction taken:** 1D linear track of length 2L (e.g. grid x∈[−L,L], with center token at x=0, Team A baseline at x=−L, and Team B baseline at x=L). - **Player/agent count:** Team A (A1,…,AN), Team B (B1,…,BN), 1 active agent per team per bout. - **Turn model:** Simultaneous tick-based simulation per duel round. Both active agents submit actions each tick until resolution. ## Formal rules R1. **Setup & Calling:** Scores start at SA=0,SB=0. In each bout, the engine samples an index k∈{1..N} uniformly at random [gap-fill]. Active agents are Ak starting at xA=−L and Bk starting at xB=L. Token is at xT=0, TokenState=CENTER. R2. **Movement & Speed:** Each tick, both active agents simultaneously select an action: `dash_forward()`, `dash_backward()`, `feint()`, `grab_token()`, or `tag()`. Movement advances position by velocity v∈{0,1,2} cells per tick. R3. **Token Grab Resolution:** - An agent may execute `grab_token()` only if their position satisfies |x|≤1 and TokenState==CENTER. - If agent Ak executes `grab_token()` while Bk does not, TokenState←CARRIED_BY_A. - If both agents attempt `grab_token()` simultaneously on the exact same tick, the grab succeeds with 50/50 probability, or a tie-break is awarded based on prior feint score [gap-fill]. R4. **Tag Resolution:** - If agent Bk executes `tag()` on agent Ak while |xA−xB|≤1: - If TokenState==CARRIED_BY_A, the tag succeeds. Bout ends: Team B gains 1 point (SB←SB+1). - If TokenState==CENTER (pre-grab tag), a foul is committed. Team A gains 1 point (SA←SA+1). R5. **Scoring on Safe Return:** - If agent Ak holds the token and reaches xA≤−L without being tagged, Team A gains 1 point (SA←SA+1). - If agent Bk holds the token and reaches xB≥L without being tagged, Team B gains 1 point (SB←SB+1). R6. **Bout Termination & Match Win:** - After a point is scored, active agents reset to baselines. - The match terminates when max⁡(SA,SB)≥Starget (e.g. Starget=5). The team with higher score wins. R7. [gap-fill] If a standoff at center (x=0) exceeds Tmax⁡=60 ticks without either agent grabbing the token, both agents are warned; if stalled for another 10 ticks, the bout is declared void and a new number is called. ## Edge cases - **Tie at the end condition:** Not applicable — points increment by 1 each bout; first team to reach Starget triggers immediate match termination. - **A player with no legal move:** All agents have valid movement actions `dash_forward()` or `dash_backward()` at all coordinates. - **Repetition / cycles:** Standoff loops at x=0 are bounded by Tmax⁡ in R7. - **Simultaneous claims on the same resource or position:** Handled by simultaneous grab resolution in R3. - **Player elimination mid-round:** In standard scoring mode, players are not eliminated. In elimination variant, the losing agent is removed from the active roster; if team size reaches 0, the opposing team wins. - **Resource exhaustion:** Token cannot be destroyed; score bounds are finite (Starget). - **Termination guarantee:** Each bout terminates either on score, tag, foul, or Tmax⁡ timeout; since Starget<∞, the match provably terminates. ## State representation - **Persistent state:** Team scores SA,SB, active roster lists, target score Starget. - **Per-bout state:** Active agent IDs (Ak,Bk), positions (xA,xB)∈[−L,L]2, token state TokenState∈{CENTER,CARRIED_BY_A,CARRIED_BY_B}, bout tick counter t∈[0..Tmax⁡]. - **Observable information:** Fully observable to both agents (both positions, velocities, and token possession state are public knowledge). ## Action space - `dash_forward()`: Move towards opponent baseline (or towards center). - `dash_backward()`: Move towards own baseline. - `feint()`: Stationary bluff action that introduces state hesitation in opponent reaction model. - `grab_token()`: Seize token (valid at |x|≤1). - `tag()`: Attempt to touch adjacent opponent (valid at |xA−xB|≤1). ## Playability notes - **Single-player-vs-AI:** Excellent 1v1 reaction game. Human agent must balance the risk of grabbing the token against the speed of the approaching AI chaser. - **Group games (n>4):** Team match naturally maps to sequential 1v1 bouts; full team composition adds strategic role selection if agents have heterogeneous speed/feint stats. - **Physical-skill games:** Physical acceleration and bluffing are abstracted into discrete movement ticks and reaction-delay windows. - **What is lost:** Subtle physical eye contact and micro-feints are collapsed into discrete choice trees. ## Implementation hints Model as a discrete Markov Decision Process on a 1D grid of 21 cells (−10 to +10). The game-theoretic core is a game of chicken at x=0: grabbing first grants the prize but leaves the agent vulnerable to an immediate tag on the next tick. First milestone: implement symmetric agents on the 1D grid with deterministic collision resolution.

Download as JSON

Download the full game corpus entry for this game as structured JSON.

Download JSON
Cite this page

Dog and Bone (Handkerchief Snatch) (Cyprus). (n.d.). Games of Countries. https://gamesofcountries.com/games/cy-cyprus/siyllo-tziai-kokkalo

@misc{gamesofcountries-cy-cyprus-siyllo-tziai-kokkalo,
  title = {Dog and Bone (Handkerchief Snatch) (Cyprus)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/cy-cyprus/siyllo-tziai-kokkalo}
}