Rational Chess

849

This post is intended to be understandable even if you have never played chess, but will probably not interest you if you don't like games like chess.

TL,DR: The most fair version of chess would be Armageddon (if neither player can checkmate the other, black wins) with a "you cut I chose" rule for black (once black plays, the first player can decide to flip the board).

The Problem

Chess as it is played today has 2 main issues:

White has a slight advantage because they make the first move
Games tend to end in a draw (partly because the players tend to play the same safe openings and partly because it seems like the natural outcome of the game as the 2 players do less and less mistakes)

Players have created many variants in order to cope with these, but none of them rose to the same popularity as classical chess. All of these variants add rules that change the whole flow of the game. For example, this article describes a game of no castling chess (which is one of the least "invasive" variant of chess) where Gukesh lost because he played what would have been a very normal classical chess move.

This post demonstrates a variant of chess that has only decisive games, no advantage for a given player, and that should theoretically allow players to retain all the skills and intuitions they earned playing classical chess.

You cut I choose

There is already a solution for games which have a first player advantage situation: the "you cut I chose" rule. Player 1 cuts the cake and player 2 chooses which side they want to eat. In chess, that would translate to player 1 making the first move with white, and player 2 choosing whether they continue the game as black or white.

We don't use the "you cut I chose" rule in chess tournaments. I believe it is because making the game too fair would create more draws.

Armageddon

There already exists a simple way to remove draws from chess, which is to have black win if neither player can checkmate the other. This variant is called Armageddon.
As I said before, chess tends to have a lot of draws, so Armageddon advantages black a lot.
The variant is used as a tie breaker in some fast chess tournaments, and the organizers usually try to balance it by giving an arbitrary 1 minute time advantage to white. Yikes. To be fair, they also put in place the way more sensible approach where players bid some of their time at the start of the game in order to gain the black pieces, but that adds a lot of complexity.

Conclusion

In Armageddon, there are 2 parameters (time and win condition) that are different from classical chess and we could make it only one parameter different by combining it with the "you cut I chose rule".

In this case the second player (black) has an advantage, so the game would unfold like this: Player 1 plays with white, Player 2 plays with black, then Player 1 chooses whether to keep white or switch to black. Player 2 is incentivized to make a bad move to balance the advantage, and after that the whole game continues just like Armageddon chess.

In your proposal, my initial guess is that player 1 has an advantage among top players in slow time controls, because you can play such that player 2 has no way to balance the game with just one move? Like, I propose the following strategy for player 1. First, play b4 as white. Then if standard engine evaluations are to be trusted (even though they are ultimately computed wrt the wrong objective function given your change in what the utility function is in chess), the move that makes the position as bad as possible for black is g5, after which the engine gives +60 regular-chess-centipawns in favor of white. So say player 1's policy is to always pick black, whatever second move player 2 chooses. We've concluded that this way, the start position will be at worst 60 regular-chess-centipawns in favor of white. My guess is that among top players, player 1 will hold the ensuing chess game to at least a draw of the time, but I'm not sure. (I asked ChatGPT-5.2 to guess a winrate from 1. b4 g5 in your chess variant for white for 2750-rated players in a slow time control, and it said , but that feels way too low for me. My guess is that black could not hold (at least) a draw of the time even from the usual chess starting position. It corrected to after I pointed this out.) ^[1]

I think your game is probably indeed balanced as play gets worse though (so, at faster time controls and lower elo ratings).

If this is indeed an issue, you can probably fix this by just letting player 2 ^[2] choose the first move by white and the first response by black (you could also add in another move if this isn't sufficient to create a balanced position or to create more opening variety), and then letting player 1 pick a side. (This is like the swap opening protocol in gomoku.) We could also add either the option or requirement for player 1 to make another move or two after and let player 2 choose. (This is like the swap2 opening protocol in gomoku, which avoids the issue that you can extensively study both sides of an opening configuration you're going to place and so have an advantage from either side even if the position looks balanced to an engine. This might be not that much of an issue in chess if the configuration-placer only gets two moves, but it quickly becomes a huge problem if you allow the configuration-placer more moves.)

on another topic: If you want to preserve the relevance of skills and intuitions from regular chess, one issue with these proposals is that they completely upend opening theory. Like, people will be playing completely different openings. Basic opening principles probably largely survive, though even that isn't so obvious because white is now much less happy with a regular-chess-draw, and that probably reverberates to some impact on more sophisticated aspects of opening principles.

another point for fun: My guess is that this rule change does not call for new endgame skills/intuitions that much — like, much less than opening ones and midgame ones. One could perhaps find this surprising, because the different win condition is "most closely seen" in the endgame. I think what's going on here is this. The rule change probably meaningfully changes the distribution of endgame "start positions" reached, but I'm guessing that almost all the "endgame start positions" reached are already familiar to good chess players (it's more like a shifted distribution on the same support than a distribution supported on new weird positions). Furthermore, usually, from an "endgame start position", when two top players are facing each other, only at most two of the three usual chess outcomes (a win for white, a draw, a win for black) are available (like, would be assigned probability, let's say). Given this, with the outcome-merge you suggest, when you get to an "endgame start position", the game is either already nearly determined (if there is only one remaining plausible outcome or if there are two remaining plausible outcomes which have been merged) or equivalent to playing that endgame in regular chess (if there are two remaining plausible outcomes and those have not been merged). In other words: if the endgame matters, it's usually basically equivalent to a regular chess endgame.

one last point for fun: Technically, perfect play in your game (after the opening procedure) is compatible with perfect play in usual chess — from the perspective of optimal play, these two games are kinda the same. Like, if you win all the winning positions and draw all the drawing positions, you're playing optimally in both games at once. Despite this, good human play in your game is quite different from good human play in usual chess (especially among top players). It's a nice example of the phenomenon that the cardinalization of the assignment of value to outcomes matters a lot in practice even in games of perfect information (where technically only the ranking of outcomes matters for determining perfect play).

According to code written by ChatGPT-5.2 which I haven't checked, b4 is the min-over-white-moves max-over-black-moves white standard engine evaluation first move, but I think it used a kinda bad engine so there might be something better. (I manually checked in the lichess evaluator that there isn't a worse response to 1. b4 than g5 though.) ↩︎
Really we should call this "player 1" in this proposal in isolation, but I'll say "player 2" to stick with the property that player 1 gets to choose a side from your proposal. ↩︎

Why can't player 1 just make a really bad move, then switch with player 2 no matter what he plays? That seems like it would give player 1 (who then becomes player 2) a huge advantage.

oh I just hadn't thought of that! I am very embarrassed, this seems like a very unfair strategy indeed.

That explains why I couldn't find any example of "you cut I choose" where the first player was choosing. Anything I can think of to counter this strategy would just make the game more complicated

I suppose you could wait to start the timer until after Player 1 chooses. The first move and response plus choosing shouldn't take very long since it can probably be fully precomputed.