| Kalah | |
|---|---|
|   | |
| Ranks | Two | 
| Sowing | Single lap | 
| Region | United States, United Kingdom | 
Kalah is a modern variation in the ancient Mancala family of games. The Kalah board was first patented and sold in the United States by William Julius Champion, Jr. in the 1950s. [1] [2] This game is sometimes also called "Kalahari", possibly by false etymology from the Kalahari Desert in Namibia.[ citation needed ]
For most of its variations, Kalah is a solved game with a first-player win if both players play perfect games. The pie rule can be used to balance the first-player's advantage.
The game provides a Kalah board and a number of seeds or counters. The board has 6 small pits, called houses, on each side; and a big pit, called an end zone or store, at each end. The object of the game is to capture more seeds than one's opponent.
It is possible for the game to end in a draw.
|   |   |   |   |   |   |   |   | 
|   |   |   |   |   |   | 
The player begins sowing from the highlighted house.
|   |   |   |   |   |   |   |   | 
|   |   |   |   |   |   | 
The last seed falls in the store, so the player receives an extra move.
|   |   |   |   |   |   |   |   | 
|   |   |   |   |   |   | 
The last seed falls in an empty house on the player's side. The player collects the highlighted seeds from both their own house and the opposite house of their opponent and will move them to the store.
Kalah was implemented on the PDP-1 in the early 1960s, [3] and was able to out-play experienced human players. [4] Since then, there have been myriad Kalah implementations for various systems including MS-DOS [5] and the Nokia 3310. [6]
 
 As mentioned above, if the last seed sown by a player lands in that player's store, the player gets an extra move. A clever player can take advantage of this rule to chain together many extra turns. Certain configurations of a row of the board can in this way be cleared in a single turn, that is, the player can capture all stones on their row, as depicted on the right. The longest possible such chain on a standard Kalah board of 6 pits lasts for 17 moves. On a general n-pit board, the patterns of seeds which can be cleared in a single turn in this way have been the object of mathematical study. [9] One can prove that, for all n, there exists one and only one pattern clearable in exactly n moves, or equivalently, one and only one clearable pattern consisting of exactly n seeds.
These patterns require arbitrarily long rows of pits and n increases. For example, it can be seen on the right that the unique 5-seed pattern requires only 3 pits, but the 17-seed pattern requires 6 pits. The relationship between the required number of pits and the number of seeds can be described in the following way. Let s(n) denote the minimum number of seeds which requires n pits to clear. Then where the symbol denotes asymptotic equivalence, that is, , or equivalently, . [9]
Mark Rawlings has written a computer program to extensively analyze both the "standard" version of Kalah and the "empty capture" version, which is the primary variant. The analysis was made possible by the creation of the largest endgame databases ever made for Kalah. They include the perfect play result of all 38,902,940,896 positions with 34 or fewer seeds. In 2015, for the first time ever, each of the initial moves for the standard version of Kalah(6,4) and Kalah(6,5) have been quantified: Kalah(6,4) is a proven win by 8 for the first player and Kalah(6,5) is a proven win by 10 for the first player. In addition, Kalah(6,6) with the standard rules has been proven to be at least a win by 4. Further analysis of Kalah(6,6) with the standard rules is ongoing.
For the "empty capture" version, Geoffrey Irving and Jeroen Donkers (2000) proved that Kalah(6,4) is a win by 10 for the first player with perfect play, and Kalah(6,5) is a win by 12 for the first player with perfect play. Anders Carstensen (2011) proved that Kalah(6,6) was a win for the first player. Mark Rawlings (2015) has extended these "empty capture" results by fully quantifying the initial moves for Kalah(6,4), Kalah(6,5), and Kalah(6,6). With searches totaling 106 days and over 55 trillion nodes, he has proven that Kalah(6,6) is a win by 2 for the first player with perfect play. This was a surprising result, given that the "4-seed" and "5-seed" variations are wins by 10 and 12, respectively. Kalah(6,6) is extremely deep and complex when compared to the 4-seed and 5-seed variations, which can now be solved in a fraction of a second and less than a minute, respectively.
The endgame databases created by Mark Rawlings were loaded into RAM during program initialization (takes 17 minutes to load). So the program could run on a computer with 32GB of RAM, the 30-seed and 33-seed databases were not loaded.
Endgame database counts: seeds position count cumulative count ------------------------------------------- 2-25 1,851,010,435 1,851,010,435 26 854,652,330 2,705,662,765 27 1,202,919,536 3,908,582,301 28 1,675,581,372 5,584,163,673 29 2,311,244,928 7,895,408,601 30 3,158,812,704 11,054,221,305 31 4,279,807,392 15,334,028,697 32 5,751,132,555 21,085,161,252 33 7,668,335,248 28,753,496,500 34 10,149,444,396 38,902,940,896 -------------------------------------------
For the following sections, bins are numbered as shown, with play in a counter-clockwise direction. South moves from bins 1 through 6 and North moves from bins 8 through 13. Bin 14 is North's store and bin 7 is South's store.
<--- North ------------------------ 13 12 11 10 9 8 14 7 1 2 3 4 5 6 ------------------------ South --->
Starting position with 4 seeds in each bin:
<--- North ------------------------ 4 4 4 4 4 4 0 0 4 4 4 4 4 4 ------------------------ South --->
The following tables show the results of each of the 10 possible first player moves (assumes South moves first) for both the standard rules and for the "empty capture" variant. Note that there are 10 possible first moves, since moves from bin 3 result in a "move-again." Search depth continued until the game ended.
Standard Rules: move result perfect play continuation ------------------------------------------------------- 1 lose by 14 10 13 3 9 13 12 1 13 11 5 13 2 lose by 10 10 13 5 9 13 8 4 10 13 8 5 3-1 lose by 6 10 11 2 13 1 12 1 13 9 4 12 3-2 tie 10 13 5 9 13 8 3 11 1 13 10 3-4 win by 2 10 9 13 2 1 12 3 5 8 12 13 3-5 win by 4 9 10 2 5 12 1 2 11 2 13 5 3-6 win by 8 9 8 2 12 6 5 11 6 1 6 5 4 lose by 2 10 12 2 4 13 1 5 9 13 12 13 5 lose by 8 10 9 11 2 5 10 1 8 4 12 5 6 win by 4 9 12 2 6 1 11 4 10 6 5 13 -------------------------------------------------------
"Empty Capture" Variant: move result perfect play continuation ------------------------------------------------------- 1 lose by 14 10 13 4 9 13 11 2 13 8 13 10 2 lose by 8 10 13 5 9 13 8 4 10 13 9 5 3-1 lose by 8 10 11 4 9 12 2 10 5 11 12 9 3-2 lose by 2 10 13 5 9 13 8 3 11 5 13 10 3-4 win by 2 10 9 13 2 1 12 3 5 8 12 13 3-5 win by 4 9 11 2 4 8 12 5 13 5 11 4 3-6 win by 10 9 8 4 11 6 2 6 4 9 5 13 4 lose by 2 10 12 2 5 9 8 12 9 4 10 11 5 lose by 6 10 9 11 4 8 13 5 6 4 12 6 6 win by 4 9 12 2 6 1 11 4 10 6 5 13 -------------------------------------------------------
Starting position with 5 seeds in each bin:
<--- North ------------------------ 5 5 5 5 5 5 0 0 5 5 5 5 5 5 ------------------------ South --->
The following tables show the results of each of the 10 possible first player moves (assumes South moves first) for both the standard rules and for the "empty capture" variant. Note that there are 10 possible first moves, since moves from bin 2 result in a "move-again." Search depth continued until the game ended.
Standard Rules: move result perfect play continuation ------------------------------------------------------- 1 lose by 10 9 11 4 8 13 2 9 6 3 11 13 2-1 lose by 4 9 10 2 12 1 11 3 12 8 11 1 2-3 win by 10 10 1 6 9 5 13 6 2 8 4 13 2-4 win by 10 8 11 1 6 9 2 13 11 4 12 6 2-5 win by 8 8 10 1 6 9 5 13 12 2 13 11 2-6 tie 8 11 1 6 3 11 6 5 12 6 8 3 win by 2 9 8 12 1 4 11 2 12 10 4 3 4 win by 2 8 11 1 5 12 3 10 5 2 11 6 5 win by 2 8 12 1 4 9 2 12 4 9 3 11 6 tie 8 12 1 6 4 10 6 2 11 4 3 -------------------------------------------------------
"Empty Capture" Variant: move result perfect play continuation ------------------------------------------------------- 1 lose by 10 9 12 6 8 12 11 2 8 6 5 12 2-1 lose by 6 9 10 2 12 4 8 9 3 10 11 3 2-3 win by 12 8 10 1 6 10 5 13 9 6 4 11 2-4 win by 8 8 9 1 6 11 4 13 10 4 13 9 2-5 win by 8 8 10 1 6 9 5 13 12 3 13 6 2-6 lose by 2 8 11 1 6 5 9 6 3 11 12 5 3 win by 2 9 8 12 1 4 11 2 10 4 5 10 4 tie 8 11 1 5 12 3 9 5 2 11 3 5 tie 8 10 1 4 12 5 11 2 9 4 13 6 tie 8 12 1 6 4 9 6 2 12 6 5 -------------------------------------------------------
Starting position with 6 seeds in each bin:
<--- North ------------------------ 6 6 6 6 6 6 0 0 6 6 6 6 6 6 ------------------------ South --->
The following tables show the results of each of the 10 possible first player moves (assumes South moves first) for the "empty capture" variant and the current status of the results for the standard variation. Note that there are 10 possible first moves, since moves from bin 1 result in a "move-again." Search depth for the "empty capture" variant continued until the game ended.
"Standard" Variant: move result ------------------------------------------------------- 1-2 proven win, by at least 2 1-3 proven win, by at least 4 1-4 1-5 1-6 proven loss, by at least 2 2 trending towards a win 3 4 5 6 proven loss, by at least 2 ------------------------------------------------------- The remaining moves (1-4, 1-5, 3, 4, and 5) are probable ties based on very deep searches, however, the result has not yet been proven.
"Empty Capture" Variant: move result perfect play continuation ------------------------------------------------------- 1-2 win by 2 10 3 12 4 8 6 10 11 6 3... 1-3 win by 2 11 1 8 2 10 6 8 3 11 5... 1-4 tie 10 3 12 5 10 3 9 1 12 3... 1-5 tie 9 4 8 3 10 2 10 4 1 9... 1-6 tie 10 4 9 6 3 11 6 8 2 10... 2 win by 2 12 4 10 1 12 8 1 11 3 9... 3 tie 10 5 12 4 11 1 12 8 4 3... 4 tie 10 3 11 1 9 5 11 2 10 8... 5 tie 10 3 11 4 12 2 11 4 10 5... 6 loss by 2 10 3 8 6 4 13 1 10 13 8... -------------------------------------------------------
A breakdown of the 55+ trillion nodes searched to solve the "empty capture" variant of Kalah(6,6):
move time (sec) nodes searched ---------------------------------------- 1-2 305,791 2,214,209,715,560 1-3 403,744 2,872,262,354,066 1-4 401,349 2,335,350,353,288 1-5 317,795 1,886,991,523,192 1-6 392,923 2,313,607,567,702 2 1,692,886 9,910,945,999,186 3 1,296,141 7,398,319,653,760 4 1,411,091 9,623,816,064,478 5 1,607,514 9,318,824,643,697 6 1,354,845 7,824,794,014,305 ---------------------------------------- total 9,184,079 55,699,121,889,234
{{cite magazine}}:  CS1 maint: url-status (link)