Vote Chess

This page describes a game of correspondence chess between the users of two microblogging platforms in which moves are selected via a multiple-choice poll.

How it works

The following microblogging accounts host one or two games of online chess per year.

On both sides, moves are chosen by poll from among the best four moves from the current board position as computed by Stockfish 14. Stockfish will search as much of the game tree as is practical, using a 48GiB cache and six-piece endgame tablebases. The search stops after approximately 236 nodes, giving a search depth of up to roughly forty-five plies or more, so we're pretty sure the moves offered will be good ones. These four possible continuations are shuffled and presented on a diagram of the board with four colour-coded arrows. The winning move in the poll is chosen as the reply. If more than one option is tied to win the vote, then the first is chosen from among them.

Each move is decided in a seventy-two-hour poll. Soon after it concludes, the result is reflected in the current board position and a new poll is offered to the opposing team. A chess game takes about forty moves usually, so we're expecting this to last well into spring of next year.

If you're reading this, we would be pleased to have you join us. You can join (and leave) the game at any time, and you can vote for the next move without even following the account on the platform of your choice. However, it might help keep the game going if you agree to be reminded to review new moves.

Other details

I have an arbiter-like role in this game. Specifically, I will check for dead positions, insufficient material, stalemate, repetition of board positions, applicability of the fifty- and seventy-five-move rules, and so on. However, I'm not a strong enough chess player to offer insightful commentary. If you are reading this and you are a strong chess player who would like to add some colour by commentating on the moves, please get in touch. Otherwise, or maybe as well as, we will offer some computer analysis at the end of the game to understand the turning points.

Starting a game. If a game is not yet in progress, there should be an active poll to guage interest. Vote in this poll with your preferred time per ply (one day or three days, owing to limitations in Mastodon). Once around half a dozen people have indicated they want to play, the game will start and run at the chosen speed.

Running out of time happens when a poll runs out of time with no votes. In this case a second, identical poll is run for one week. If at the end of this time there are still no votes, the player loses, unless there's no possible way for their opponent to win, in which case the game is drawn.

The 'prize' for winning is that the profile picture on both accounts will be either a bird (if Twitter users win) or a woolly mammoth (if Mastodon users win).

Draws and resignations are handled as follows. Where fewer than four options are possible from the board position given, an additional option will be added to resign the game. If this option wins in the vote, the king will be virtually tipped, and victory for the opponent announced. I can't see a sensible and consisent way to support offering a draw without reducing the variety of choices available to the player to an unacceptably small set. Therefore, offering a draw is not supported in the current form of the game.

Pawn promotion occurs when a pawn reaches its eighth rank. If this happens, an additional poll is held to determine whether the pawn is promoted to a queen, rook, bishop or knight.

At six days per complete move, the speed of the game may seem ludicrously slow. However, if the response within the first day on both sides climbs to the extent where we think it can support the rest of the game continuing, we will be able to reduce the time for one ply to one day, with up to three complete moves per week.

The number of tweets and toots is minimised to avoid annoyance. Therefore, usually, once a poll concludes the next update will be from the opponent's reply. Owing to limitations in what constitutes a tweet or a toot, the board position is posted, and then a reply containing the poll.

Please don't use chess engines to help you select the next move. It kind of goes against the point of seeing which way the game goes. Technical limitations restrict the vote to a set of four often strong moves anyway, and all four are already chosen by a chess engine. It will be difficult to blunder!

Comparing the moves offered at the start with those which feature in chess opening theory, you may notice that many continuations appearing in the chess openings are sometimes not offered in vote chess. Instead, Stockfish might compute that better options exist, perhaps with benefits many moves ahead. Depending on how you view this, it might be a feature or a bug. A bug, because it is impossible to play out even common lines from centuries of knowledge of chess openings. A feature, because these new lines may lead to interesting game situations not commonly encountered. In any case, we could try to fix this, perhaps using a database of openings until an agreed point.


Thanks to the authors of LaTeX and its packages chessboard and skak, the image editor GIMP, and of course the mighty Stockfish combined with the Syzygy endgame tablebases, for making our games possible.

Any questions or suggestions? Just get in touch with me.