Skip to contents

Computes, by exhaustive alpha-beta search, the maximum card points the declarer can guarantee from a fully specified position where every hand is known (a "double-dummy" analysis). This is exact perfect-information play: it assumes both defenders cooperate optimally against the declarer.

Usage

sk_solve_dd(
  hands,
  game_type,
  declarer,
  leader = declarer,
  max_tricks = 8,
  call = rlang::caller_env()
)

Arguments

hands

A length-3 list of sk_card vectors (or character vectors of card codes): the remaining cards of seats 1, 2, and 3. Every seat must hold the same number of cards, and no card may be held twice.

game_type

A single game-type string, one of sk_game_types().

declarer

Integer in 1:3: the declaring seat.

leader

Integer in 1:3: the seat to lead the next trick. Defaults to declarer.

max_tricks

Maximum number of cards per hand the solver will accept, as a guard against accidentally launching an intractable search. Deeper positions raise an error rather than running. At the default of 8 a solve stays well under a second on ordinary hardware; each further trick multiplies the work several times over, and a whole ten-trick deal (max_tricks = 10) can take tens of seconds.

call

The calling environment, used for error reporting.

Value

A single integer: the maximum card points (0-120) the declarer can guarantee against best defence.

Details

The returned total counts every card point in the tricks the declarer wins, including the points on the declarer's own cards. It does not include skat points, which are added afterwards by sk_score_game().

The search is exponential in the number of remaining cards, tempered by alpha-beta pruning, move ordering and a transposition table. It is practical for endgames; a full ten-trick deal can still take a very long time, which is what max_tricks guards against. The heavy lifting runs in compiled C++, and the search polls for user interrupts so a long solve can be cancelled.

See also

sk_play_game() to play a deal out with heuristic players, and sk_score_game() to turn a point total into a scored result.

Examples

# A two-card-each endgame in a grand game.
hands <- list(c("CJ", "SA"), c("HA", "H9"), c("DA", "D9"))
sk_solve_dd(hands, game_type = "grand", declarer = 1)
#> [1] 35

# The declarer holds both top jacks and sweeps the table: 46 points, which
# includes the two card points on the declarer's own jacks.
sweep <- list(c("CJ", "SJ"), c("HA", "DA"), c("H10", "D10"))
sk_solve_dd(sweep, game_type = "grand", declarer = 1)
#> [1] 46

# A whole ten-trick deal is refused by default; raise the guard to solve it.
# \donttest{
deal <- sk_deal(seed = 1)
sk_solve_dd(deal$hands, game_type = "grand", declarer = 1, max_tricks = 10)
#> [1] 73
# }