5

Consider a chess variant in which normal chess rules apply, with one exception. The player that is able to bring a piece or a pawn onto the opponent’s back rank, which would be the 8th rank for White and the 1st rank for Black, wins immediately. The move has to be legal and it ends the game, even if the opponent could mate on the next move.

On Liches some chess variants like King of the Hill, Atomic, and Three-Check can be analyzed. Can I easily adjust Stockfish or another chess program such that this new rule applies? Would White have a big advantage in this game? Or can we even show that White has a winning strategy?

6
  • 1
    I created an engine that plays a related variant. With no kings, the task is to move a pawn to the final rank. You can play against the engine apronus.com/chess/puzzle/… Aug 25 '20 at 9:06
  • Sounds like a fun variation. Simply tweaking Stockfish's evaluation function so that it evaluates such positions as being mates would probably lead to decent play, but deeper changes to the evaluation function would probably be needed for optimal play. Deep learning approaches would be easier to modify. Aug 24 '21 at 19:03
  • So if you capture all of your opponents pawns, is the worst you can do a draw or can they still win with checkmate? Can you promote a pawn if in check?
    – user70889
    Dec 15 '21 at 16:12
  • @MichalRyszardWojcik: Could you also program the variant only with pawns? That's a standard fun game in kid training. Dec 15 '21 at 18:07
  • @HaukeReddmann apronus.com/chess/puzzle/… -- Pawn Wars. Promotion wins. Having no moves loses. No draw. On my editor you can set up any position with or without kings and create a link to share. Dec 17 '21 at 14:11
3

Revised stockfish and added this goal to win the game.

  • Created 22 start opening mostly 2 plies others are 4 plies. Done carefully to minimize white advantage in the opening.
[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 c5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 e5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 e5 2.Nf3 Nc6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 e6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 Nf6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 Nc6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 Nc6 2.Nc3 e5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.e4 d5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.d4 d5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.d4 d5 2.Bf4 Bf5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.d4 Nf6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.c4 e5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.c4 c5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.c4 Nf6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.c4 d6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.c4 g6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.c4 Nc6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.Nf3 d5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.Nf3 Nf6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.Nf3 c5 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.Nf3 e6 *

[Event "?"]
[Site "?"]
[Date "?"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]

1.Nf3 d5 2.d4 Bf5 *

  • Run a match at TC 3m+1s

Result

Score of sfbackrank_e2 vs sfbackrank_e1: 8 - 13 - 1 [0.386]
...      sfbackrank_e2 playing White: 5 - 5 - 1  [0.500] 11
...      sfbackrank_e2 playing Black: 3 - 8 - 0  [0.273] 11
...      White vs Black: 13 - 8 - 1  [0.614] 22
Elo difference: -80.4 +/- 156.7, LOS: 13.8 %, DrawRatio: 4.5 %
22 of 22 games finished.

sfbackrank_e2 and sfbackrank_e1 are the same engine. The white side won.

White vs Black: 13 - 8 - 1  [0.614] 22

Sample games

https://lichess.org/Ab5lGE1w
https://lichess.org/4VdHGQkm
https://lichess.org/b8xxzvjZ
https://lichess.org/IVpjazCV
https://lichess.org/gwwYdwCA
https://lichess.org/QdiEcpyK
https://lichess.org/GmCdzkz8
https://lichess.org/cDR66Vuq
https://lichess.org/5Fc6G9MC

2
  • I checked several of the example games and don't see any actual wins. Are you using some kind of adjudication rule? If so, what is it?
    – Allure
    yesterday
  • When 2 engines agreed that the last 2 successive moves have scores of 900 cp or more in favor of one engine then game is adjudicated. I use cutechess-gui.
    – ferdy
    yesterday
-3

Well, since White has the first move, then White would always be trying to put a piece on the back rank, while Black is trying to catch up. So Black can eventually trade off all the pieces, and Black can keep opposition.

However, I'm not sure if this is a legitimate strategy. We would have to solve until we find a forcing winning line or until all possible positions are analyzed before we know the actual result.

1
  • 3
    I would rather guess that White has a winning strategy, but probably only exhaustive search will give a final answer. I think, chances to solve this are better than to solve "king of the hill"
    – Peter
    Aug 25 '20 at 9:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.