Skip to main content
Formal Specification

I Drop, I Drop a Handkerchief — Formal Specification

Bulgaria · chase-tag · Confidence: medium

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 3
# Puskam puskam karpichka — AI-optimized specification ## General idea A spatial circular pursuit game with hidden item placement and detection latency. A circling runner agent chooses an optimal seat behind which to drop a token; seated agents perform blind periodic checks. Once detected, the round transitions into a deterministic 1D cyclic pursuit race to claim the single vacant seat. ## Formal model - **Abstraction taken:** The continuous circular perimeter is modeled as a 1D cyclic discrete array of M=N−1 seat nodes (indices 0..M−1). Spatial movement is discretized into node-per-tick steps. - **Player/agent count:** N agents (1 Runner R, M Seated agents S0..SM−1, M≥3). - **Turn model:** Real-time ticked simulation (e.g. Δt=0.5 sec per movement tick). ## Formal rules R1. Setup: M seated agents occupy discrete circle nodes i∈{0,…,M−1} arranged in clockwise order. Runner R begins at node position posR=0, holding token K. R2. Circling phase: - At each tick t, runner advances position posR=(posR+1)(modM). - Runner may execute action `drop_token()` at current node posR=j. When executed, token K is placed at node j, and drop_tick=t. R3. Detection model: - Seated agent Sj cannot observe the runner's visual state. - [gap-fill] At each tick t≥drop_tick, seated agent Sj executes a blind tactile check action with detection probability Pdetect=min⁡(1.0,β⋅(t−drop_tick)), where β=0.4. - [gap-fill] If posR reaches (j+M)(modM) (a full lap) without detection, Sj is penalized as "burned" (forfeit counter increments) and Sj immediately detects K. R4. Pursuit phase: - Upon detection at tick tdet, agent Sj vacates seat j, picks up K, and starts at node j. - Both R and Sj race clockwise toward target seat node j. - Runner distance to target: dR=(j−posR(tdet))(modM). If dR==0, dR=M. - Chaser distance to target: dC=M (must travel full circumference to catch R from behind, or dtag=(posR(t)−posC(t))(modM)). R5. Pursuit resolution: - Both agents have sprint speeds vR,vC (nodes per tick). - [gap-fill] At each pursuit tick: - If posC==posR, Chaser tags Runner → Runner fails; role assignment unchanged for next round. - If posR==j and posC≠posR, Runner claims seat → Runner succeeds; Sj becomes the new Runner for next round. R6. Round reset: Game state resets to R1 with the determined Runner and rotated seating. A session ends after Kmax rounds; score is defined by rounds spent as a successful seated player. ## Edge cases - Tie at the end condition: If session ends on round Kmax with tied scores, the agent with fewer total penalty points (fewer times "burned") wins. - A player with no legal move: Not applicable — circling, checking, and sprinting actions are always available within their respective phases. - Repetition / cycles: Bounded: The runner is strictly forced to drop the token within M ticks (or drop is auto-triggered), and sprint resolution terminates in at most ⌈M/min⁡(vR,vC)⌉ ticks. - Simultaneous claims on the same resource or position: Handled explicitly in R5: If Chaser and Runner reach seat j on the exact same tick, tag resolution takes precedence (posC==posR). - Player elimination mid-round; turn order repair: If a seated agent disconnects, M decrements by 1; seat nodes re-index to 0..M−2. - Resource exhaustion: Exactly 1 handkerchief token exists in the system; state transitions are conservative. - Termination guarantee: Each round provably terminates in finite ticks (finite drop window + finite pursuit window); Kmax bound terminates the session. ## State representation - Persistent state: Roster of agents, cumulative scores/penalties, session round counter k∈[1,Kmax]. - Per-tick ephemeral state: Phase ∈{Circling,Pursuit}, posR∈[0,M−1], token dropped flag, token drop node j, token drop tick tdrop, posC∈[0,M−1]. - Observable information: Seated agents observe only their private tactile probe results; Runner observes all spatial positions. In pursuit phase, all positions are public. ## Action space - `circle_step()`: Runner advances 1 node clockwise. - `drop_token()`: Precondition: Phase == Circling, token in hand. - `tactile_check()`: Precondition: Phase == Circling, executed by seated agents. - `sprint_step()`: Precondition: Phase == Pursuit, executed by R and Sj. ## Playability notes - Single-player-vs-AI: Highly playable — Human plays as runner (deciding when to drop) or as seated agent (timing checks and executing the chase). - Group games (n>4): Reduces to 3 agents (N=3, 1 runner, 2 seated) with identical mechanics. - Physical-skill games: Sprint speed is modeled via an agent attribute v∈[1.0,2.0] nodes/tick; reaction latency is governed by the stochastic check model. - What is lost: The psychological tension of physical breathing sounds, subtle eye cues, and footstep audio is abstracted into the probabilistic detection formula. ## Implementation hints Low complexity. State is a 1D integer ring. The critical design parameter is tuning the detection probability slope β against sprint speeds (vR,vC) so that dropping behind a distant agent provides a viable strategic advantage. Suggested first milestone: implement the 1D discrete ring pursuit resolution with deterministic detection.

Download as JSON

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

Download JSON
Cite this page

I Drop, I Drop a Handkerchief (Bulgaria). (n.d.). Games of Countries. https://gamesofcountries.com/games/bg-bulgaria/puskam-karpichka

@misc{gamesofcountries-bg-bulgaria-puskam-karpichka,
  title = {I Drop, I Drop a Handkerchief (Bulgaria)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/bg-bulgaria/puskam-karpichka}
}