Skip to main content
Formal Specification

Psitekatekisana (two-phrase riddle-guessing contest) — Formal Specification

South Africa · riddle-word · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: directGap-fills: 5
# Psitekatekisana — AI-optimized specification ## General idea A memorized call-and-response riddle contest with an elimination structure: a fixed riddle is put to players one at a time until someone supplies its fixed correct answer from memory. Unlike a logic-puzzle riddle game, success is pure recall — there is nothing to deduce in the moment. ## Formal model - **Abstraction taken:** the real riddle corpus (Junod records ~100 examples in Thonga with English glosses) is treated as an external content database rather than generated content, since the riddles are fixed cultural artifacts, not proceduraly generatable. - **Player/agent count:** 1 questioner, 2+ respondents (a 2-total-player version reduces to a simple challenge-response duel). - **Turn model:** sequential — the questioner challenges respondents one at a time with the *same* riddle until it is answered correctly. ## Formal rules R1. A riddle (first-phrase, second-phrase pair) is selected from the corpus by the questioner. R2. The questioner announces the fixed ritual opening ("Teka, teka, teka... heeee!") and then the riddle's first phrase, addressed to one respondent. R3. That respondent must immediately answer with the riddle's exact fixed second phrase. R4. If correct, the round ends; [gap-fill] this spec awards that respondent one point and, per the gap noted in the entry, makes them the new questioner for the next round (a reasonable default absent an attested rule). R5. If incorrect (including no answer, or a mismatched second phrase from a different riddle), the questioner declares them beaten ("Psi ku hlulile") and repeats the *same* riddle (R2-R3) against the next respondent in turn. R6. This repeats until some respondent answers correctly, or all respondents have been tried once per round — [gap-fill] if all respondents fail once, this spec has the questioner reveal the answer and move to a new riddle, since the source does not attest an all-fail outcome. ## Edge cases - Tie at end: not applicable to a single riddle-round (exactly one respondent answers correctly to end it, or R6's all-fail gap-fill triggers); a multi-round match could tie on points, [gap-fill] unresolved by the source. - A player with no legal move: a respondent can always attempt an answer (even a wrong guess); "no legal move" does not arise. - Repetition/cycles: bounded by R6's gap-filled all-respondents-once limit per riddle, preventing infinite repetition of a single unanswered riddle. - Simultaneous claims: not applicable, strictly one respondent challenged at a time. - Player elimination: a respondent who fails is not eliminated from the game, only from that specific riddle-round (attested: play simply moves to the next respondent). - Resource exhaustion: the riddle corpus itself is finite; [gap-fill] a real implementation needs a rule for what happens when all riddles have been used (e.g. reshuffle), not attested in the source. - Termination guarantee: each riddle-round is bounded (R6), and a full match is bounded by the finite riddle corpus with a reshuffle policy, so the game provably terminates given the R6 gap-fill. ## State representation Persistent state: match score per player (if scoring is adopted per R4's gap-fill); which riddles have been used this session. Per-round ephemeral state: the active riddle, which respondents have already failed on it this round. No hidden information — the riddle's first phrase is announced publicly; only the correct second phrase is privately "known" by whichever players have memorized it, which is itself the whole game. ## Action space - `pose_riddle(riddle_id, target_respondent)`: questioner's action, opens a challenge. - `answer(claimed_phrase)`: respondent's action, checked against the riddle's true fixed second phrase. ## Playability notes - Single-player-vs-AI: natural — a human respondent against an AI questioner cycling through a riddle database, or a human questioner testing an AI's "memorized" answer database. - Group games (n>4): natural fit, the elimination-per-riddle structure scales cleanly to any number of respondents. - Physical-skill games: none — purely a memory/knowledge game, though the source emphasizes real-time speed ("speaking very rapidly"), suggesting a timed-response element could be added as a faithful [gap-fill] enhancement. - What is lost: the ritual performative elements (rapid delivery, the fixed nonsense opening phrase, the social dynamic of group riddling) reduce to a simple database-lookup quiz mechanic in a straightforward digital adaptation. ## Implementation hints This maps very cleanly onto a trivia/quiz-game engine: fixed question-answer pairs, sequential-until-correct challenge structure. The one genuinely distinctive design element worth preserving is R5's "same riddle repeats against the next player" structure, rather than the more common quiz-game pattern of moving to a new question after any single failure.

Download as JSON

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

Download JSON
Cite this page

Psitekatekisana (two-phrase riddle-guessing contest) (South Africa). (n.d.). Games of Countries. https://gamesofcountries.com/games/za-south-africa/psitekatekisana

@misc{gamesofcountries-za-south-africa-psitekatekisana,
  title = {Psitekatekisana (two-phrase riddle-guessing contest) (South Africa)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/za-south-africa/psitekatekisana}
}