7

Is there a strong Losers chess variant engine?

I had a look at stockfish, but it doesn't support Losers chess, seemingly.

(to be used on a Linux platform)

Rewan Demontay
  • 16,942
  • 4
  • 65
  • 109
Déjà vu
  • 605
  • 4
  • 13
  • 1
    I think it does. Not the offical version. – ABCD Dec 28 '17 at 06:21
  • 1
    On lichess you can also start an analysis session with a specialised Stockfish for each chess variant. Here's the link for Losers chess (there known as Antichess): https://lichess.org/analysis/antichess – Ellie Dec 28 '17 at 15:50
  • The version of Stockfish at lichess.org supports Anti-chess aka Suicide chess (no check / checkmate, give away all pieces). Technically not the same variant as Losers chess (give away all pieces, but includes check / checkmate). – Rick G Dec 29 '17 at 02:47

2 Answers2

8

https://github.com/ddugovic/Stockfish

has what you want. Not an offical version, but should be stable for production.

enter image description here

ABCD
  • 22,419
  • 2
  • 43
  • 82
  • 2
    This version of Stockfish supports both Suicide chess (no check / checkmate, give away all pieces) as well as Loser chess (give away all pieces, but includes check / checkmate). – Rick G Dec 29 '17 at 02:46
  • @RingØ What do you mean by endgame against king only? In losers chess the game immediately ends if one side only has the bare king, so are you talking about losers or antichess? If Stockfish is really bad at playing certain endgames, please let me know or create an issue at https://github.com/ddugovic/Stockfish/issues, then we will try to investigate. – Fabian Fichter Jan 04 '18 at 23:15
  • @FabianFichter Thanks for contributing. Indeed, my phrasing is confusing. It's actually King with 1 (or 2) pawns only, blocked by 1 (or 2) SF pawns, SF having more material, still (typically the required material to win that kind of game: R + B for instance). In this case, SF repeats configurations endlessly (while at depth 28). I was (also) interested in programming a simple algo, but that endgame part seems to require a specific laborious implementation. – Déjà vu Jan 05 '18 at 01:58
  • @RingØ For losers chess we have not implemented any specific endgame evaluation functions, partly because I simply do not have much experience with losers chess (in contrast to giveaway/suicide chess). I have just tested positions like 8/8/8/1p3k2/1P6/5KR1/8/4B3 w - - 0 1, and SF finds "mate" quite quickly at depth 22. Could you please provide a sample position where SF has difficulties winning it? And if you know some basic ideas how to win such endgames, that would also be nice to know. To me it seems like sacrificing a piece, blocking the pawn and promoting your own pawn is the idea. – Fabian Fichter Jan 05 '18 at 09:59
  • @FabianFichter Well, I tried various positions at depth 28, and SF is strikingly strong! Could only get a draw from `7K/p1p5/P1P5/8/8/8/8/7k b` or `1b5K/p1p5/P1P5/8/8/8/8/7k b` (how the `b` went to `b8` is another story). Tried many configs and depths at the time and possibly the settings were different, so removed my previous wrong comment. SF does a very good job in Losers endgames, actually, and that's amazing if it's not even specifically programmed for that. – Déjà vu Jan 05 '18 at 14:07
  • 1
    Ok, thanks, that is nice to hear. If you still find something fishy, please let us know. – Fabian Fichter Jan 05 '18 at 17:00
4

Sjeng supports losers chess too since release 9.

Jacques Gaudin
  • 381
  • 2
  • 11