Two player chess

The rules of chess, as this board plays them

Chess is a game for two on an eight by eight grid of sixty-four squares. Thirty-two pieces stand on it at the start, sixteen a side: a king, a queen, two rooks, two bishops, two knights and eight pawns. White moves first and the two of you alternate from there. Rooks travel along ranks and files, bishops along diagonals, the queen does both, and the king takes one step in any direction. Knights jump in an L and are the only piece that passes over anything in the way. Pawns push one square forward, two from their starting rank, and capture one square diagonally instead. A piece captures by landing on the enemy square. Attack the enemy king so that no legal move gets him out of it and the game is over.

Castling, en passant and promotion are all here. Castling is generated with all of its conditions checked: the king and that rook must both be untouched, the squares between them empty, and the king may not be in check, pass through an attacked square, or land on one. Capture the rook on its home square and the right to castle on that side goes with it. En passant is here properly: a pawn that has just jumped two squares can be taken by an enemy pawn beside it as though it had gone only one, and that window shuts after a single reply. A pawn reaching the last rank promotes to whichever of queen, rook, bishop or knight you pick, so an underpromotion is available when that is the move that mates.

Every candidate is generated, then played out and discarded if it would leave your own king in check, so an illegal move is never one of the squares on offer. Select a piece and the board shows you where it may go. When your king is attacked, his square is marked too.

Every way a game ends on this table

Checkmate is one of the two ways a side wins. The other is resignation, which here is an action rather than a way of leaving: the position records who resigned and the other seat takes the game.

Three things end a game as a draw. Stalemate, where the side to move is not in check and has nothing legal to play. The fifty-move rule, counted as a hundred half-moves without a capture and without a pawn moving. And insufficient material, which this table reads as king against king, king and one bishop or one knight against a bare king, a bishop each when both stand on squares of the same colour, and, going a step past the official rule, a knight each. With a knight each a mate can still be constructed if the losing side helps, so an arbiter would not call that position dead. This board does. The test counts the pieces left standing and stops there.

Two familiar draws are not implemented. There is no threefold repetition, so shuffling the same position three times over settles nothing, and a perpetual check runs on until the fifty-move count catches it. There is no draw offer either, because the only two things a seat can send are a move and a resignation, so there is no button to offer one with. A game here finishes when the position finishes it or when somebody gives it up.

Three ways to start a game of chess for two

Playing chess online here starts with one click, which sets out a table and hands you a link. You keep White, whoever opens the link sits down as Black, and nobody registers anything to do it. The two of you can be in different countries; the link is the whole arrangement. The table is listed nowhere and its address is a short random string, so that link is the only route to it.

Reload the page in the middle of a game and your seat comes back, because the tab is holding the token that proves the seat is yours, and a dropped connection is recovered the same way. Close the tab and the token goes with it.

Two people in one room can manage with a single device instead. Choose one-screen mode on the same board, and the laptop or the phone goes across after every move. The board does not flip between moves. White stays at the bottom for both of you, so the squares are where you left them when the device comes back.

With nobody to pass it to, the third option seats the computer opposite you. Colour goes with the seat rather than with a choice, so the first seat is White in all three modes. You take White, the computer takes Black, and it starts thinking the moment your move lands.

What same device and offline actually mean here

The middle option is two people who want to play chess on one computer, or on one phone, with no account made and nobody joining from the internet. One-screen chess and the game against the computer both run the rules inside your browser. The file that decides whether your bishop may go there is loaded into the page, so no server is judging anything and no move leaves the device. The computer’s search runs in a background thread in the same browser, which is why the position stays responsive while it thinks.

None of that is a download. This is a web page, so it has to arrive before you can use it, and nothing here installs itself and keeps working once the network is gone. There is no app to fetch from a store, no account to make first, and no cost. If offline to you means no opponent on the internet and nothing to install, this does the job. If it means chess on a train with the signal gone, it does not.

The computer opponent, described exactly

It is a minimax search with alpha-beta pruning, three plies deep: its move, your reply, its answer to that. At the end of that it counts what is standing. A pawn scores 100, a knight 320, a bishop 330, a rook 500 and a queen 900, and every piece also gets a small bonus or penalty for the square it occupies, so it likes pawns in the centre and a king tucked away in the opening. It tries captures first, biggest catch by smallest attacker, because that is what makes the pruning cheap.

It has no opening book, so its first move comes out of the same search as its fortieth. It has no endgame tables. The search stops dead at the third ply with no extension for an unfinished exchange, so a capture sequence that resolves on the fourth is invisible to it. It is the same strength every time, with no difficulty level to choose and no rating to climb. And it is deterministic: give it the same position twice and you get the same answer twice, so an opening that worked once will work again until you deviate.

It sits on every board here as a test instrument before it is a feature, since somebody has to lose to a board before anyone else is offered it. The owner’s own words for that: "I have play tested every game that ships."

Chess strategy that works against this opponent

The habits a beginner is taught first are the ones this machine handles worst. Control the centre with your pawns, because its evaluation hands out its largest positional bonuses in the middle files and it will contest them while leaving a flank alone. Castle early, since it scores a king that has reached the corner and it will go after one still standing in the middle. And learn to look one exchange further than it can: its search stops after three plies, so a trade that only turns in your favour on the fourth sits past its horizon. Control of the centre, an early castle and one more move of lookahead: two friends learning together will find the same three ideas doing most of the work against each other.

What this table does not have

No clock. Nothing here counts your time, so there is no blitz, and a game by link can sit half-finished while the two of you get on with something else. The table itself is the one limit: a room with no move in it for forty-eight hours is cleared, and the game goes with it. No ratings, no accounts, no ladder, no analysis afterwards, no engine to tell you where it went wrong, no takebacks.

The sites that do keep clocks and ratings have been at it for the better part of twenty years, with somebody at your exact strength awake somewhere at three in the morning and a machine that will explain your loss to you. What that costs is a sign-up, a rating you can lose, and an opponent who is a stranger.

What is here instead is one person you know, a set that opens in a browser, and nothing that turns the result into a number you carry around. It is chess with a friend.

It left India as chaturanga and reached Europe without its queen

The oldest ancestor anybody can point at is chaturanga, in India by the sixth or seventh century. The name means four divisions of an army: infantry, cavalry, elephants and chariots, which are the pawn, the knight, the bishop and the rook still standing on your back rank. It was set out on the ashtapada, an unchequered eight by eight grid that already existed for a race game.

It went to Persia as chatrang and became shatranj after the Arab conquest, since Arabic had no ch or ng to spell it with. Persian players called out shah when the king was attacked and shah mat when he could not get away, and English kept both: check, and checkmate. Carried west by traders and armies, chess was across Europe by about the year 1000.

What arrived was slower than what you are about to play. The piece beside the king was the vizier, and it moved one square diagonally. The elephant that became the bishop jumped two squares and could reach only an eighth of the squares on the board. Europe renamed the vizier a queen without changing what she could do, and she stayed a one step piece for roughly five hundred years.

Then, somewhere between about 1475 and 1500 in the kingdom of Valencia, the queen and the bishop were given the reach they have now. The earliest document showing it is a poem in Catalan, Scachs d’amor, written in Valencia in 1475, in which a match runs as an argument between Love and Reason. Italians called the new version alla rabiosa, chess with the madwoman. What had been a long grind of one-step pieces could now be lost in a dozen moves, and promoting a pawn was suddenly worth playing for. The game has barely moved since, and the last question still open, what a stalemate was worth, took until the nineteenth century.

Stalemate is a draw, and agreeing that took a thousand years

One line of our code calls a stalemate a draw, and it stands on a thousand years of disagreement about what that was worth. In chaturanga, stalemating the other side lost you the game. Shatranj reversed it and made the same move a win, and that was still the practice in Spain in the early fifteenth century. Lucena, around 1497, counted it as half a victory, worth half the stake in a match for money, and Spain was still doing that in 1600. In England between about 1600 and 1800 it was a loss for the side that delivered it. H. J. R. Murray traced it to Russian practice and called it illogical in print. A 1614 English handbook defended the rule on the ground that it disturbed the proper course of the game. Medieval France had a version where the stalemated side merely forfeited its turn. Thirteenth century Italy already counted it a draw, and that is the reading that spread, reaching England only in the nineteenth century through Jacob Sarratt.

So the draw you get here is a nineteenth century settlement rather than something chess always knew.

The game chess put away in the north

Chess did not arrive in an empty room. Northern Europe already had tafl, a family of siege games where one side is outnumbered two to one and runs a king for the edge of the board, and by the twelfth century chess had taken its place across Scandinavia. Hnefatafl is on this site as well, and the two boards sit differently in our hands. We can tell you how our chess works because chess never stopped being written down. We can only tell you how our hnefatafl was reconstructed, because nobody wrote its rules down before chess replaced it.

The handover left objects behind. A tafl board of eleven squares by eleven, cut in the twelfth century, came out of the ground near Trondheim. The Lewis chessmen, carved from walrus ivory and whale teeth between about 1150 and 1200, are usually thought to have been made in that same town, and were found in 1831 on a Hebridean shore. The hoard is counted at ninety-four objects. Seventy-eight of them are chessmen, fourteen are plain tablemen from the older family of board games, and one is a belt buckle. That comes to ninety-three, and what the ninety-fourth object was is not in the breakdown anybody publishes. Traces of red stain on some of the carvings suggest the two sides were red and white rather than black and white. Whether one workshop cut both the new game and the old one is not something anybody can say from that evidence.

Why a collection of old boards keeps a chess set at all

Chess belongs to nobody. Its rules cannot be owned, its grid is drawn in every country on earth, and that is the qualification for a shelf whose purpose is to put every public-domain board game where two people can reach it. Some of what stands on this shelf arrived because its rules were free, not because anyone had heard of it. Chess arrived on both counts, and it is the one game most people walking in already know how to play.

It also settles an argument the rest of the collection keeps having. Our hnefatafl is a reconstruction, our Royal Game of Ur rests on a description of play rather than a rulebook, and our Surakarta had to decide for itself when a stalled position is over. Chess needs none of that. The moves are known, the moves are agreed, and each rule stated on this page can be read off the same file the server runs when the two of you play online by link.