Formal Specification
Cape Verdean Traditional Draughts — Formal Specification
Cabo Verde · board-abstract · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: directGap-fills: 3
# Xadrez Crioulo — AI-optimized specification
> **Reconstructed source.** The entry carries a provenance warning: its rules were written from
> general knowledge, not read from the two works it cites. Carreira (1983) and Lopes (1957) are
> real books, but the entry does not evidence that either describes this game.
## General idea
International-style draughts on an 8x8 board, with the two rules that define the Portuguese and
Brazilian branch of the family: **flying kings** and the ***Lei da Maioria*** — the law of
quantity, under which a player facing several capture sequences **must** take the one capturing the
most pieces. That second rule is what gives the game its character. Captures are not merely
compulsory but *determined*: in many positions the player has no choice at all, and the skill lies
in constructing positions where the opponent's forced maximum capture is the move you want them to
make. Sacrificing into a long forced sequence is the game's signature tactic, and an engine that
implements capture as optional has removed it entirely.
## Formal model
- **Abstraction taken:** None. Discrete, deterministic, perfect information, finite action set.
- **Player/agent count:** Exactly 2, 12 pieces each.
- **Turn model:** Strictly sequential; one move per turn, with a full capture sequence resolved as
a single move.
## Formal rules
R1. Setup (entry §1): played on the **dark squares** of an 8x8 board with **12 pieces** per player
— 32 playable squares, so 8 are vacant at the start.
R2. Man movement (entry §2): a regular piece moves **1 square diagonally forward**.
R3. Capture (entry §3): jumping over an **adjacent** opponent piece to the vacant square beyond is
**mandatory**. Captures may be made **forward and backward**, by men as well as kings — the
backward capture is a real asymmetry with R2's forward-only movement and is easy to miss.
R4. *Lei da Maioria* (entry §4): where multiple capture paths exist, the player **must** choose the
path capturing the **greatest number** of pieces. This makes capture selection a constrained
maximisation rather than a free choice [gap-fill: the entry does not state a tiebreak among
equally long maximal sequences, nor whether king captures outrank man captures when counts tie
— a rule some codifications of this family do include. Free choice among equal maxima is
assumed].
R5. Promotion (entry §5): reaching the **back rank** crowns a piece as a King (*Dama*), which
**flies** — moving any distance along a clear diagonal [gap-fill: the entry does not say
whether a man that merely passes over the back rank mid-capture promotes, or must end its move
there. Ending the move there is assumed, which is the standard resolution and the one that
keeps R4's sequences well-defined].
R6. King capture: kings capture along diagonals, consistent with R3's mandatory rule and R5's
flying movement [gap-fill: the entry does not specify the landing square after a flying
capture — whether the king may stop on any vacant square beyond the captured piece or only the
one immediately beyond. Free landing is assumed, being the standard companion to flying kings
and the reading that makes R4's maximisation meaningful for kings].
R7. Victory (entry §6): capture **all** opponent pieces, **or block** the opponent from having any
legal move.
## Edge cases
- Tie at the end: **the entry defines no draw at all**, which is its most significant omission.
King-versus-king endgames in this family are frequently drawn in practice and there is no
repetition rule, no move counter, and no material-based draw here. Both R7 conditions are
one-sided, so no *tie* can be declared — but many positions are objectively drawn and the entry
provides no way to end them.
- A player with no legal move: **explicitly a loss** under R7. Note that R3's mandatory capture
interacts with this: a player whose only pieces are blocked has no move and loses, and a player
with a legal capture available has no choice but to make it.
- Repetition / cycles: **unresolved, and the most likely failure mode in practice.** Two flying
kings can manoeuvre indefinitely without capturing. With no repetition rule and no move limit,
such a game never ends. The standard 40-move or 25-move draw conventions of codified draughts are
**not** in this entry and are not imported.
- Simultaneous claims on one position: not applicable — strictly sequential turns. A full capture
sequence resolves entirely within one turn before the opponent acts, so no square is ever
contested.
- Elimination mid-round and turn-order repair: pieces are eliminated, players are not. Two-player
alternation is invariant and needs no repair.
- Resource exhaustion: pieces are strictly monotone non-increasing — captures remove them
permanently and nothing restores any. Promotion (R5) converts a piece rather than adding one, so
material never grows. This monotonicity is what makes *capture* lines terminate, and it is
exactly what non-capture king manoeuvring escapes.
- Termination guarantee: **not guaranteed.** Any line containing infinitely many captures is
impossible, since at most 24 pieces can ever be removed, so capture-rich games terminate quickly.
But R2 and R5 movement is unbounded and nothing compels a capture when none is available, so
capture-free king endgames can continue forever. With no repetition or move-limit rule stated,
`termination_guaranteed: false` reflects the entry rather than the game as actually played. An
implementation must add a draw rule and label it as its own addition.
## State representation
- Persistent (game) state: `board[32]` ∈ {empty, man_p1, king_p1, man_p2, king_p2} over dark
squares only, `current_player`. Piece counts are derivable and monotone non-increasing.
- Per-turn ephemeral state: `capture_sequence` (the ordered path under evaluation), `captured_set`
(pieces removed by the sequence in progress, which must not be re-jumped), `max_capture_count`
for R4's comparison.
- Observable information: fully observable, perfect information. No hidden state, no randomness.
## Action space
- `move(from, to)`: the active player; precondition: no capture is available anywhere (R3 makes
capture mandatory), and the destination is legal for the piece type (R2 for men, R5 for kings).
- `capture_sequence(path)`: the active player; precondition: a capture exists, and `path` is among
the sequences capturing the **maximum** number of pieces (R4). In many positions this set is a
singleton and the player has no decision at all.
## Playability notes
- Single-player-vs-AI: excellent and unusually cheap to search, precisely because R3 and R4 prune
so aggressively — a forced maximum capture is a branching factor of one, and long forced
sequences let search reach far deeper than the nominal depth. The absence of a draw rule (see
above) is the practical problem: without one, self-play games will hang in king endgames.
- Reducing n > 4 to 2 agents: not applicable; inherently two-player.
- Physical-skill games: not applicable — nothing physical is modelled.
- What is lost: the setting the entry describes, which is the whole social point — concrete tables
in the *praças* of Mindelo and Praia, played daily and in public alongside Ouril. Also lost is
the blitz variant the entry records, where fast clocks change the game considerably.
## Implementation hints
Implement R4 as a **two-pass move generator**: enumerate every capture sequence to its full depth,
find the maximum length, and return only the sequences achieving it. Doing this lazily or
greedily — following the first capture found and extending it — produces legal-looking moves that
violate the *Lei da Maioria* and will silently play a different game. Track `captured_set` during
sequence generation so a piece is not jumped twice within one sequence, and remove captured pieces
only at the end of the sequence, since intermediate removal changes which further jumps appear
available. R3's backward capture for men is the single most commonly omitted rule in this family
and must be handled separately from R2's forward-only movement. Finally, a draw rule is
**required** for a usable engine and is **not** in the entry: implement one, label it explicitly as
an addition, and do not present it as part of the tradition. Suggested first milestone: the
32-square board with man movement and mandatory backward-and-forward capture, then the maximal-
sequence generator verified against hand-constructed positions with two capture paths of different
lengths, then flying kings.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Cape Verdean Traditional Draughts (Cabo Verde). (n.d.). Games of Countries. https://gamesofcountries.com/games/cv-cabo-verde/xadrez-crioulo
@misc{gamesofcountries-cv-cabo-verde-xadrez-crioulo,
title = {Cape Verdean Traditional Draughts (Cabo Verde)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/cv-cabo-verde/xadrez-crioulo}
}