Skip to main content
Formal Specification

Handkerchief Whip — Formal Specification

Uzbekistan · chase-tag · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: discretizedGap-fills: 5
# Durra urish — AI-optimized specification ## General idea Durra urish is an asymmetric circular chasing and vigilance game where seated players form a ring facing inward with heads down, while a carrier walks around the outside perimeter to secretly drop a knotted cloth scarf behind one player. If the chosen player fails to detect the scarf before the carrier completes a full lap, they are penalized with a whipping run around the circle; if they detect it in time, a high-stakes sprint ensues to see whether the carrier can steal the vacant seat before being tagged. Structurally distinct from simple Duck-Duck-Goose, detection relies on backward blind hand-checking and penalty laps rather than immediate vocal naming. ## Formal model - **Abstraction taken:** The physical circular perimeter is discretized into a ring graph of N seat nodes V={0,1,…,N−1} with modulo arithmetic adjacency. Time is discretized into sequential ticks. Seated agents have a backward-check action each tick to detect if the scarf has been dropped behind their seat node. - **Player/agent count:** 1 Carrier agent (C), N Seated agents (S0,…,SN−1), where N≥4 (formal default: N=6). - **Turn model:** Simultaneous discrete tick-based commitments across all active agents. ## Formal rules R1. **Initial Setup:** - Ring nodes 0…N−1 arranged in clockwise circle. - Agents S0…SN−1 assigned to seat positions 0…N−1. - Carrier agent C starts at exterior position 0 holding `cloth_held = True`. - `dropped_at_node = None`. R2. **Patrol & Drop Phase:** - Each tick, carrier C advances clockwise: pos(C)←(pos(C)+1)modN. - While `cloth_held == True`, carrier C may execute `drop_cloth()` at their current node k, setting `dropped_at_node = k` and `cloth_held = False`. - Each tick, seated agent Si at node i chooses between: - `feel_behind()`: Checks whether `dropped_at_node == i`. (Has a 1-tick cooldown or attentional cost [gap-fill]). - `idle()`: Remains passive. R3. **Detection and Chase Resolution:** - **Case A — Timely Detection (Sprint to Vacant Seat):** - If agent Sk executes `feel_behind()` while `dropped_at_node == k`, Sk snatches the cloth and becomes active runner (R), vacating seat node k. - Both C and R sprint clockwise around the ring graph: - Carrier speed: 1 node/tick. Runner speed: 1 node/tick [gap-fill: or runner has 1.2x speed advantage based on sprint attribute]. - If C reaches vacated node k before R catches C, C takes the seat (C→Sk), and R becomes the new Carrier for the next round. - If R enters the same node as C before C sits, R executes `tag_carrier()`; C remains the Carrier and resets with cloth at node k. - **Case B — Undetected Full Lap (Kuyish Penalty):** - If carrier C returns to `dropped_at_node == k` without Sk detecting the cloth, C executes `penalize_sleeper(S_k)`. - Sk must immediately perform a 1-lap penalty circuit pursued by C. At the conclusion of the lap, Sk re-occupies seat k, and C initiates a new drop patrol with the cloth. R4. **Round and Match Structure:** - A full match consists of M=10 round cycles [gap-fill]. - Points: An agent scores +2 points for successfully stealing a seat as carrier, +1 point for catching a carrier as runner, and -1 point if penalized for failing to detect a drop. Highest score wins. ## Edge cases - **Tie at the end condition:** If multiple agents have equal points after M rounds, the agent who suffered fewer penalty laps wins [gap-fill]. - **A player with no legal move:** A seated player can always choose `idle()` or `feel_behind()`; a running agent must advance along the ring. - **Repetition / cycles:** Bounded by the maximum tick limit per round (Tmax=30 ticks per drop) and the M-round cap. - **Simultaneous claims on the same resource or position:** If carrier C and runner R enter the target seat node on the exact same tick, the carrier is awarded the seat if they entered without being tagged on the preceding tick [gap-fill]. - **Player elimination mid-round:** Not applicable; all N+1 agents participate across all rounds with roles shifting between Seated, Carrier, and Runner. - **Resource exhaustion:** Ring nodes and cloth object are invariant and never depleted. - **Termination guarantee:** Each round terminates within at most 30 ticks; the complete match terminates in ≤300 ticks. ## State representation - **Persistent state:** - Ring size N. - Current round index r∈{1…M}. - Agent role map: Carrier ID, Seated IDs {0…N−1}. - Scores: integer vector of length N+1. - **Per-tick ephemeral state:** - Carrier position pos(C)∈{0…N−1}. - Runner position pos(R)∈{0…N−1}∪{None}. - Scarf status: `held_by_carrier`, `on_ground_at(k)`, or `held_by_runner`. - Vacant seat node index. - **Observable information:** - Seated agents observe only their own node and whether their own `feel_behind()` action succeeded (carrier position is hidden until detected or penalty triggered). - Carrier observes all node locations and whether a runner has stood up. ## Action space - `move_clockwise()`: Advance 1 node along the perimeter. Precondition: active carrier or chasing runner. - `drop_cloth()`: Deposit cloth at current node. Precondition: carrier at node k, `cloth_held == True`. - `feel_behind()`: Check ground for cloth. Precondition: seated agent. - `idle()`: Pass tick. Always legal for seated agents. - `claim_seat()`: Sit down in vacant node. Precondition: carrier entering vacant node. - `tag_carrier()`: Strike carrier with cloth. Precondition: runner occupying same node as carrier. ## Playability notes - **Single-player-vs-AI:** Natural. Human can play as Carrier trying to deceive heuristic AI seated agents with varied drop timing, or as a Seated player managing attention intervals to detect drops. - **Group games reduction:** Reduces to minimum viable ring of N=4 seated agents + 1 carrier (5 total agents) without loss of bluffing or sprint dynamics. - **Physical-skill games:** Physical sprinting speed and vigilance are modeled via discrete movement speeds and stochastic detection windows. - **What is lost:** Physical tactile cloth vibration, peripheral hearing cues, and the chaotic physical joy of laughing children running in a yard. ## Implementation hints - Implement ring positions strictly using modulo arithmetic: next_node=(curr+1)modN. - Enforce strict information hiding: seated agents must not have access to carrier's coordinates in their observation tensor. - First milestone: Implement 5-agent ring with deterministic drop at node 2 and test sprint resolution to seat vacancy.

Download as JSON

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

Download JSON
Cite this page

Handkerchief Whip (Uzbekistan). (n.d.). Games of Countries. https://gamesofcountries.com/games/uz-uzbekistan/durra-urish

@misc{gamesofcountries-uz-uzbekistan-durra-urish,
  title = {Handkerchief Whip (Uzbekistan)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/uz-uzbekistan/durra-urish}
}