Skip to main content
Formal Specification

Saudi Baloot Trick-Taking Card Sport — Formal Specification

Saudi Arabia · card · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: directGap-fills: 4
# Al-Baloot (البلوت) — AI-optimized specification ## General idea A Belote-derived partnership trick-taker whose defining structure is that **the two contracts are different games sharing a deck**. Choosing *Sun* (no-trump) or *Hakam* (trump) does not merely set a trump suit — it swaps the entire card-ranking table, so the Jack that is worth 20 and outranks everything under Hakam is worth 2 and outranks nothing under Sun. Bidding is therefore a decision about which valuation of your own hand you want to be true, taken while holding only five of your eventual eight cards. ## Formal model - **Abstraction taken:** None of substance. This is a card game with fully discrete state, so the formal model is the game — no physical skill is being proxied. The only reduction is that table talk, tempo, and the bluffing the entry emphasises culturally are outside the action space. - **Player/agent count:** Exactly 4, in 2 fixed partnerships seated alternately. - **Turn model:** Strictly sequential and clockwise, in three phases per deal — bidding, declarations, then eight tricks of four plays each. ## Formal rules R1. Deal (entry §1.1): 5 cards to each player, then 1 card face-up in the centre. R2. Bidding (entry §1.2): in clockwise order, players bid to buy the centre card, choosing contract `Sun` or `Hakam`. The buying partnership is the declaring side [gap-fill: the entry does not state what happens if all four players decline; a redeal by the next dealer is assumed, since the entry provides no pass-out resolution and play cannot otherwise proceed]. R3. Completion of the deal (entry §1.3): once the contract is bought, 3 further cards are dealt to each player, giving 8 cards in hand. R4. Card values, contract-dependent (entry §1.2) — this table *is* the contract: - `Sun`: Ace 11, Ten 10, King 4, Queen 3, Jack 2, Nine 0, Eight 0, Seven 0. Trick total 130, plus 10 for the last trick = 140. - `Hakam` (trump suit only): Jack 20, Nine 14, Ace 11, Ten 10, King 4, Queen 3, Eight 0, Seven 0. Trick total 162. Non-trump suits use the Sun values. R5. Declarations / *Mashari'* (entry §2), scored as Trump / Sun: `Sira` 3-card sequence 20/40 · `Khamsin` 4-card sequence 50/100 · `Mia` 5-card sequence or four Aces/Kings/Queens/Tens 100/200 · `Arba'miya` four Jacks in Sun 400 · `Baloot` King+Queen of trump held in hand 20. R6. Follow suit (entry §3.1): a player holding the led suit must play it. R7. Trump obligation, Hakam only (entry §3.2): a player unable to follow suit must play a trump if holding one, and must over-trump the highest trump already in the trick if able. With neither the led suit nor a trump, any card may be played. R8. Contract outcome (entry §4.1-4.2): if the buying side takes more points than the defenders, both sides record what they took. If the buying side fails to take the majority — *Khasara* — the defending side takes **all** trick points from the deal. R9. *Kaboot* (entry §4.3): a side taking all 8 tricks receives a bonus [gap-fill: the entry says "massive bonus points" without quantifying; the value is left as a named parameter rather than invented, because it materially changes optimal bidding]. R10. Match victory (entry §5): the first partnership to accumulate **152 game points** wins [gap-fill: the entry gives two scoring notations — a 152-point target, and a tournament notation in which a Sun deal is worth 26 and a Hakam deal 16 game points. These are alternative conventions rather than one scale; an implementation must fix one at match start. The 152 target is used here as the entry's primary statement]. ## Edge cases - Tie at the end: both partnerships crossing 152 in the same deal is unaddressed by the entry [gap-fill: the deal is awarded to the side with the higher total in that deal, with play continuing if those are also equal — the least invasive reading, since the entry supplies no tiebreak]. - A player with no legal move: impossible. R6 and R7 constrain *which* card may be played but never exclude every card — a player unable to follow suit and holding no trump may play any card in hand, and a player's hand is non-empty until the deal ends. - Repetition / cycles: deals repeat until R10 is satisfied, but no state repeats, because match scores are monotone non-decreasing across deals and cards are reshuffled each deal. No cycle is possible. - Simultaneous claims on one position: not applicable — play is strictly sequential and exactly one card occupies each position in a trick. The single contested resource, the centre card in bidding, is resolved by clockwise bid order (R2). - Elimination mid-round and turn-order repair: not applicable — no player or partnership is ever eliminated, and the clockwise order over 4 seats is invariant for the whole match. There is nothing to repair. - Resource exhaustion: structural and expected — each player's 8 cards are exhausted after 8 tricks, which is precisely the deal's terminal condition. The deck is fully reconstituted and reshuffled for the next deal, so exhaustion never blocks progress. - Termination guarantee: **guaranteed.** Every deal distributes a fixed positive quantity of trick points between the two partnerships — 140 under Sun, 162 under Hakam (R4) — and R8 assigns all of it to one side or splits it, never discarding it. Match scores are therefore monotone non-decreasing with a strictly positive increase to the combined total each deal, so the 152-point threshold (R10) is crossed by some partnership within a bounded number of deals. Declarations (R5) and Kaboot (R9) only accelerate this. ## State representation - Persistent (match) state: `match_score[team]` (monotone non-decreasing), `dealer` (rotates). - Per-deal state: `contract` ∈ {Sun, Hakam}, `trump_suit` (Hakam only), `declaring_team`, `hands[player]` (8 cards, shrinking), `declarations[player]`, `tricks_won[team]`, `deal_points[team]`. - Per-trick ephemeral state: `led_suit`, `cards_played` (ordered), `trick_leader`. - Observable information: **partially observable, and this is the game's core.** Each player sees only their own hand plus all cards played so far; the face-up centre card is public during bidding. Declarations reveal partial information about a hand deliberately. An engine must not leak the full deal to any agent. ## Action space - `bid(contract)` / `pass()`: the player to act during bidding; precondition: bidding phase. - `declare(combination)`: any player holding a qualifying combination (R5); precondition: declaration phase of the deal. - `play_card(card)`: the player to act; precondition: the card is legal under R6 (follow suit) and R7 (trump and over-trump obligations under Hakam). ## Playability notes - Single-player-vs-AI: natural, and one of the better fits in this corpus. Partnership play means a human needs exactly one AI partner and two AI opponents, and the hidden-hand structure gives an AI genuine inference work rather than a scripted routine. - Reducing n > 4 to 2 agents: **not possible without changing the game.** Partnership is not decoration here — declarations inform a partner, and *Khasara* punishes a partnership jointly. A two-player reduction produces a different game, not a smaller one. - Physical-skill games: not applicable. Nothing physical is being modelled, which is why `fidelity_class: direct` and `engine_suitability: direct` are appropriate — this game needs no proxy at all. - What is lost: the bluffing, table talk, and reading of a partner's tempo that the entry singles out as the game's cultural substance, and the rapid mental arithmetic of tracking points across two different valuation tables under time pressure. ## Implementation hints The state is small but the information structure is the whole problem: this is an imperfect-information game requiring determinization or belief tracking, not minimax over a known position. Branching factor is at most 8 during trick play and typically far lower once R6/R7 bind. The two representational traps are both in R4 — hardcoding a single card ranking, which silently breaks one of the two contracts, and applying Hakam's inflated values to non-trump suits, which corrupts every deal's point total. Suggested first milestone: implement Hakam only, with declarations disabled, and verify that a deal's trick points sum to exactly 162 before adding Sun, declarations, and Kaboot.

Download as JSON

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

Download JSON
Cite this page

Saudi Baloot Trick-Taking Card Sport (Saudi Arabia). (n.d.). Games of Countries. https://gamesofcountries.com/games/sa-saudi-arabia/al-baloot

@misc{gamesofcountries-sa-saudi-arabia-al-baloot,
  title = {Saudi Baloot Trick-Taking Card Sport (Saudi Arabia)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/sa-saudi-arabia/al-baloot}
}