Play Mancala Online

Pick up all stones from one of your pits and sow them counter-clockwise. Land your last stone in your store for a bonus turn. The player with the most stones wins!

Created by Brian Hamilton

Player 1 (You)
0
Player 2
0

Your turn — pick a pit

How to play Mancala

Collect more of the 48 stones than your opponent. Your six pits run down the right-hand side into your store at the bottom.

  • Tap one of your pits to pick up all its stones, then drop them one at a time into each pit going round — down your side, through your store, and up your opponent's.
  • You never drop a stone into your opponent's store, only your own.
  • If your last stone lands in your own store, you take another turn straight away. A pit does this when its stone count equals its distance to your store.
  • If your last stone lands in an empty pit on your own side, you capture it along with every stone in the pit facing it — unless that pit is empty too, in which case nothing happens.
  • The game ends the moment either side has no stones left. Everything still on the other side goes to that player's store, so running dry while your opponent is holding stones can lose a won game.

Controls: tap any of your own pits that still has stones in it. Undo takes back your last move. Choose an opponent — one of three AI levels, or a second player — above the board.

Mancala strategy game — sow seeds and capture stones to win this classic board game

The free turn is a counting problem

Mancala looks like a game of scooping and scattering. It is really a game of counting one thing: how far each pit is from your store. Pick up the stones in a pit and drop them one at a time into each pit going round, and if the last stone falls in your own store, you go again.

The free turn, and the counting rule behind it The opening position, with four stones in each of the twelve pits and both stores empty. The board is drawn the way the game shows it, rotated so the long axis runs top to bottom: your six pits run down the right-hand side into your store at the bottom. The third pit down holds four stones and sits exactly four steps from your store, so sowing it drops the last stone into the store and earns another turn. It is the only opening move that does. In general a pit gives another turn when the number of stones in it equals its distance from your store.
Only the third pit down is exactly four steps from the store, so it is the one opening move that drops a stone in and earns another turn.

So a pit earns you another turn exactly when the number of stones in it equals its distance to your store. That single sentence is most of the tactics. From the opening position, where every pit holds four stones, only one pit is four steps from home — and it is the move almost every experienced player makes first.

Chains follow from this. Because a free turn lets you move again, and moving changes the counts, one well-chosen move can set up the next. Strong play in this game often looks like three or four moves in a row before the opponent touches the board at all.

Capturing is a trap you have to set

The other way to win stones in bulk is the capture. If your last stone lands in a pit on your own side that was empty, you take that stone and everything sitting in the pit facing it.

Capturing by landing in an empty pit on your own side A single stone sits in the top pit of your column and the pit below it is empty. Sowing that one stone lands it in the empty pit. Because that pit is on your own side and was empty, that stone together with all five stones in the pit facing it are captured - six stones in all - and go to your store, taking it from twelve to eighteen. On this rotated board a pit and the pit facing it sit side by side. If the facing pit had been empty there would be no capture at all.
The last stone lands in an empty pit on your own side, so it and the five facing it go to your store. On the rotated board, facing pits sit level with each other.

Both conditions matter, and both are easy to forget. Landing in an empty pit on your opponent's side does nothing. Landing in an empty pit on your own side when the facing pit is also empty does nothing either — you have simply parked a stone there.

That makes captures something you engineer rather than stumble into. Emptying one of your own pits is the first half of the trick; arriving back at it with exactly the right count, while the pit opposite is full, is the second. Watch for your opponent doing the same, and notice when one of their pits has been suspiciously empty for a while.

Running out of stones is a move

The game ends the moment either player has no stones left on their side — and everything still sitting on the other side goes straight into that player's store. This is the rule that decides more games than any other, and the one beginners discover the hard way.

What happens when one side runs out of stones Your six pits are all empty while your store holds twenty stones and your opponent's holds eighteen, with ten stones still sitting in their pits. Because one side has no stones left, the game ends immediately, and all ten remaining stones are swept into your opponent's store. What looked like a two-stone lead becomes a twenty-eight to twenty defeat, with no further moves played.
One side is empty, so play stops and everything left opposite is swept up. A 20-18 lead becomes a 28-20 defeat with no further moves.

It means emptying your own side is not a neutral act, and a lead is not safe. A player two stones ahead can lose by eight without their opponent making another move. Late in the game, the question stops being "which move gains most" and becomes "who runs out first, and how much is sitting on the other side when they do".

The practical version: when you are ahead and the board is emptying, keep stones on your own side. When you are behind, pile them up — if your opponent runs dry first, every stone you are holding is yours.

Why the board is turned on its side

A real mancala board is long and thin, which is exactly the wrong shape for a phone. This version rotates it 90°: your six pits run down the right-hand side into your store at the bottom, and your opponent's run down the left into theirs at the top. Sowing goes down your side, through your store, and back up theirs.

One useful side effect: a pit and the pit facing it — the one you capture from — end up level with each other, side by side. On a traditional horizontal board they sit above and below, which is much harder to read at a glance.

What was verified on this page

Kalah has a conservation property as clean as any game on this site: there are 48 stones at the start, and there are 48 stones for ever afterwards. Sowing moves them, capturing moves them, the end-of-game sweep moves them — but nothing may create or destroy one. Checking that after every move tests sowing, the skip of the opponent's store, the capture and the sweep all at once.

It held across 26,129 individual sowings and 2,000 end-of-game sweeps. Sowing itself was compared against an independent implementation written from the rules on 69,799 moves and matched exactly every time, including which moves earn a free turn (9,542 of them) and how many stones each capture takes (6,229 capturing moves). All 400 complete games played to a proper finish, with every stone ending in a store.

One real defect was found in the AI, and fixed. When the search reached a position where the game was already over, it scored it with its running estimate instead of the actual result — and that estimate valued the stones about to be swept up at 0.3 each rather than 1. A finished position the computer wins 28–20 was being rated +1.0 instead of +8, so the search would happily prefer a line that looked tidier and lost.

A fix nobody can measure is not worth shipping, so both versions were built from the same page and played against each other. The corrected engine chooses a different move in 8.8% of positions overall and 17.9% of endgames — exactly where you would expect it to matter — and it wins the match 70–40 with 10 draws, over 120 games from 60 different openings with each side going first.

The corrected engine against the original, both at the Hard setting, playing the game on this page.
Measurement Result
Positions where the two engines differ8.8%
… in the endgame (12 stones or fewer left)17.9%
Head-to-head result, 120 games70–40–10
Time to choose a move (median / worst)46ms / 134ms

One thing that could not be established. This version of the game is small enough that it has been solved outright by researchers, who found a first-player win. An exhaustive search was run here to confirm it independently, and it did not finish: after 84.7 million positions examined and 15 million stored, in 200 seconds, it was still going. That is reported as what it is — an unfinished search, which proves nothing either way. The claim that the first player wins is left to the people who did complete it.

Older than the rules it is played by

Mancala is not one game but a family of hundreds, played across Africa and Asia for well over a thousand years — boards cut into stone have been found at sites in Ethiopia and Eritrea, and the family is plausibly far older than any surviving board. Oware, Bao, Congkak and Sungka are all cousins with genuinely different rules.

The version here is Kalah, which is much younger than it looks: it was designed and marketed in the United States in 1940 by William Julius Champion Jr. The free-turn rule and the capture-the-facing-pit rule are his additions, and they are what make the counting so sharp. It is the version almost everyone in the English-speaking world means by "mancala", and the one nearly every computer version implements.

More Puzzle & Strategy Games

If you enjoy Mancala, try these other free games on our site:

Dots & Boxes

Draw lines between dots. Complete boxes, steal chains, outscore your opponent.

Reversi

Flip your opponent’s discs. Control the board. Classic strategy.

Game Over!