Balancing Horde Chess

Throughout all of this testing, so far I have failed to find bugs in either SF's horde move-ordering or evaluation functions. So this improves our confidence in the test results.

@ubdip Are those percentages the white winning percentage? 60% is pretty high, but not too bad! I'm also interested in scores from "Advanced Wing Horde" with the side pawns moved up instead. It would be:

(rnbqkbnr/pppppppp/8/PPP2PPP/PPPPPPPP/PPPPPPPP/PPPPPPPP/1PPPPPP1 w kq - 0 1)

Would you like to run that analysis as well?

Now that I think about it, there are 16 candidate positions with the same number of pawns as Standard Horde. Each of the four mirrored files {a, b, c, d} could have a pawn missing from the back rank and fifth rank. That gives a lot of options to work with.

@ChessWhiz It's white's total score including draws, not the winning percentage, but since there are very few draws, it's almost the same. Considering that some variants have a similar or even stronger color bias, e.g. atomic, three-check, and crazyhouse, 60%-40% should be ok, I think. Anyway, the main question is how human players will perform with these setups.

Since we have the tools to analyze other starting setups, I think it would be good to investigate further to see if there's another reasonable obviously-more-balanced setup. 60% is still pretty steep. (Atomic has a strong starting-move bias, and it's not even 60% win rate, even at 2400+ ratings).

@ChessWhiz Well, in principle any setup can be analyzed in that way and might give interesting results, but it takes time and computing resources, so there is a trade-off.

You seem to be comparing results of human atomic games with horde computer games. If you compare Stockfish's results for these variants, you get a different result, e.g. for atomic I get about 66% (score, winning percentage is of course lower since there are more draws in atomic chess), for 3check even about 69%.

I think that the setup should mainly be balanced for humans, not just for engines. These measurements were just a way to find an objectively more or less equal setup, but the results of human games will tell in which direction we might have to adjust and by about how much.

Only changing starting configuration of horde still will be forced win for one side - white or black.

Therefore, to make this game fair I suggest to add 1 rule: white pawns have ability to move 1 square backwards.

With this rule horde chess will be balanced, and will end in a draw if both players play perfect. It will change the game-play significantly. Games will become longer and more strategical, and will require to take some risk if one want to play for a win.

For any strong horde player it is well-known fact that best strategies for black is to put pawns side into zugzwang, waiting for white to weaken their position, and finally to collect all weak pawns. White can do nothing to prevent this, because they are forced to move forward. With new rule horde side could play defensive style, maneuvering their forces back and forth, like in a real war battle.

"With this rule I guarantee horde chess will be perfectly balanced." Impressive! :-)

@FischyVishy @Stubenfisch @svenos
The 22nd is coming closer, so I would like to know whether you can give an update on the upcoming match. Will you use the originally proposed setup with two additional pawns or one of the suggested alternatives? And have you decided on the match time and the time control?

Hello @ubdip

svenos and Stubenfisch have agreed to postpone the match till August 26. They will play with the two extra pawns, but I don't have any problem if they want to follow any of the other brilliant solutions posted ITT.

I will record a video with live comments on the match. Let's get excited! Also, I will soon create a new team for the purpose of discussing ways in which we can improve Horde.