Stash is a strategy game.

It's simple on the surface, and easy to learn.

But it contains complexity.

It is a 'simultaneous action' game.

Players choose their own moves in secret, simultaneously.

Then the moves play out at the same time.

The best way to play is to predict what your opponent will do, and play a superior counter.

There are two ways to win a game of Stash.

Be the first player to put 20k in your hand.

Or, eliminate the other player.

A player richer than you can be shot.

If the player is not richer than you, your shot will miss.

In the case of a tie, the larger pile wins.

As long as players have equal amounts, play continues indefinitely.

DOUBLE any pile or add to $0

HALVE the amount in any pile

SPLIT all the money on the table between the players

MOVE any pile of money to anywhere else

SWAP any two piles in the game

SHOOT your opponent, which eliminates him, and wins the game. For a shot to hit, the player's opponent must be richer after all other actions have resolved. If the opponent is not richer after resolution, the shot is missed.

Players get one shot per game only.

Moves play out at the same time, but there is a resolution order for different actions.

The buttons are laid out in resolution order.

I created this game with a view to learning more about Game Theory.

I'm interested in dilemma games, where players must co-operate with, and betray, one another to win.

I'm also interested in simultaneous action games.

I'm a complete beginner game theorist. My goal is just to make a fun, balanced, playable, simultaneous game.

No dominant strategy for Stash has yet been discovered.

Training an A.I. player for simultaneous action is different than it is for turn-based games like chess.

Typically, A.I. has better information about other players' moves.

In simultaneous action games, A.I. has to make its best guess about what the opponent will do in the same turn.

The thinking is more approximate than decisive.

To train the A.I. player, I gave it a little bit of search depth (it 'thinks' 3 moves ahead), and then it played millions of rounds against itself, so that it could learn to improve.

This is quite basic, and there are lots of methods out there for creating and improving A.I. players in this type of game.

If anyone out there has suggestions for how to create a better or different A.I. stash player, please get in touch. I would love to learn more.

Thank you for visiting, and I hope you enjoy playing this game.