Formal Specification
King and Thief — Formal Specification
Uzbekistan · guessing-hidden · Confidence: medium
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: directGap-fills: 4
# Podshoh-o'g'ri — AI-optimized specification
## General idea
Podshoh-o'g'ri is a fast-paced four-player hidden-role social deduction and bluffing game. Four secret roles—King (*Podshoh*), Vizier (*Vazir*), Guard (*Askar*), and Thief (*O'g'ri*)—are dealt to players; upon the public revelation of the King and Vizier, the Vizier is commanded by the King to identify the hidden Thief between the two remaining unrevealed players. If the Vizier deduces correctly, the Thief is punished; if the Vizier accuses the Guard, the Vizier is condemned for royal incompetence and punished instead.
## Formal model
- **Abstraction taken:** None required; the traditional game already operates as a discrete, finite hidden-role state machine with distinct information sets.
- **Player/agent count:** Exactly 4 agents ().
- **Turn model:** Sequential state machine phases: Deal Royal Declarations Accusation Verification & Score Update.
## Formal rules
R1. **Role Deck & Scoring Baseline:**
- Role set: .
- Point values: , , , .
- Target match score: points [gap-fill: default target 5000 points].
- Cumulative score vector initialized: .
R2. **Dealing & Secret Observation:**
- At the start of each round, a uniform random permutation of is dealt to agents .
- Each agent privately observes their own assigned role .
R3. **Public Announcement Phase:**
- The agent holding publicly reveals their role.
- The agent holding publicly reveals their role.
- The remaining two agents () remain unrevealed; the game engine identifies them as the Candidate Set , where .
R4. **Accusation & Bluffing Phase:**
- The King issues an accusation command to the Vizier.
- [gap-fill] Both candidates submit a public message/signal (plea of innocence or bluff).
- The Vizier chooses an accusation target: .
R5. **Verification & Score Allocation:**
- **Case 1 — Correct Accusation ():**
- Vizier succeeds.
- .
- .
- .
- .
- **Case 2 — Mistaken Accusation ():**
- Vizier fails.
- .
- (loses round points).
- .
- (steals Vizier's bounty).
R6. **Match Termination:**
- Rounds repeat until at least one player reaches . The player with the highest cumulative score wins.
- [gap-fill] Round cap: If after rounds no player has reached , highest score wins.
## Edge cases
- **Tie at the end condition:** If multiple players exceed on the same round with equal scores, additional single rounds are played until the tie is broken [gap-fill].
- **A player with no legal move:** Not applicable; role reveal is mandatory, and Vizier must select between candidates and .
- **Repetition / cycles:** Scores increase strictly monotonically (at least total points distributed per round); guaranteed termination.
- **Simultaneous claims on the same resource or position:** Not applicable; role drawing is an exact 1-to-1 permutation without replacement.
- **Player elimination mid-round:** Not applicable; all 4 players participate in every round.
- **Resource exhaustion:** The 4-card role deck is reshuffled identically each round.
- **Termination guarantee:** Guaranteed by strictly positive score distribution each round () reaching within finite rounds.
## State representation
- **Persistent state:**
- Target score .
- Score array .
- Round counter .
- **Per-round ephemeral state:**
- Permutation vector .
- Identified King ID, Vizier ID.
- Candidate set .
- Accusation choice .
- **Observable information:**
- Agent observes only , public King/Vizier identity, public candidate IDs , public chat messages, and public scores. Roles of are hidden until post-accusation reveal.
## Action space
- `reveal_role()`: Mandatory action executed by King and Vizier holders in Phase 3.
- `broadcast_signal(msg)`: Bluffing / statement emitted by candidates .
- `accuse(target_id)`: Executed by Vizier. Precondition: `target_id` .
## Playability notes
- **Single-player-vs-AI:** Natural. Human can play any role in the 4-agent circle against AI opponents implementing Bayesian deduction, deceptive speech policies, and expressive bluffing.
- **Group games reduction:** The 4-player structure is mathematically minimal and canonical for this game.
- **Physical-skill games:** Not applicable (pure discrete deduction).
- **What is lost:** Micro-expression reading, eye contact twitch detection, nervous laughter, and the physical comedy of forehead finger-snaps (*chertmak*).
## Implementation hints
- Implement the round lifecycle as a clean 4-step state machine: `DEAL -> REVEAL_ROYALS -> ACCUSE -> SCORE`.
- Ensure AI Vizier maintains a belief distribution updated by previous round deception tendencies.
- First milestone: Implement the deterministic 4-player dealing and accusation scoring logic with random guessing to verify point accumulation and win-threshold termination.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
King and Thief (Uzbekistan). (n.d.). Games of Countries. https://gamesofcountries.com/games/uz-uzbekistan/podshoh-ogri
@misc{gamesofcountries-uz-uzbekistan-podshoh-ogri,
title = {King and Thief (Uzbekistan)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/uz-uzbekistan/podshoh-ogri}
}