Turnover Chess

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
hgm
Posts: 27788
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Turnover Chess

Post by hgm »

I came across a very original game related to Chess (I would hesitate to call it a chess variant). The pieces there are composits of up to 3 concentric rings I(nner), M(iddle) and O(uter), which can be combined in 6 ways (I, M, O, IM, MO, IMO; I+O does not 'fit' and is illegal) to represent the 6 different kinds of Chess pieces. When it moves (as the piece it represents) only the outermost ring moves, though, so if they are not Pawns (O), Rooks (I) or Bishops (M) they would split up and leave something behind. It is also allowed to recombine rings that exactly fit on the target square, even with an enemy piece (which will then be 'turned over' to become a friend before it gets absorbed). Only if it doesn't fit it will be captured in the normal way (removing the old occupant), and you are not allowed to do that with friendly pieces. Kings (IMO, called Castles) move as a Pawn, though (leaving behind a Queen, IM). At the end of your turn you must at have at least one Castle that is not under attack (i.e. subject to capture or turnover); if there are no moves that would realize that you are checkmated.

PDF rule description at https://drive.google.com/drive/folders/ ... _nl2zg7Uqd

Any ideas for an engine? :D
nionita
Posts: 175
Joined: Fri Oct 22, 2010 9:47 pm
Location: Austria

Re: Turnover Chess

Post by nionita »

This kind of games would be ideal for a "zero" approach, as there is no human experience or intuition how the game strategy should look like. Then, especially if the game is interesting (i.e. the complexity not too low and not too high), it would be also interesting to see how quick a "dedicated" approach (with handcrafted eval) can beat the zero one (if at all).
User avatar
hgm
Posts: 27788
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: Turnover Chess

Post by hgm »

Indeed. This is even more true for 'Piece Chess' (Paco Shako), where I really have no idea what to strive for as an intermediate goal (as all material always stays on the board). In Turnover Chess you can at least decimate the number of rings the opponent controls. So my first attempt at a hand-crafted evaluation would be to assign a piece value to each of the 3 ring types. And then there would be a small positional penalty for combining a Pawn (O) with something else, because combining it with Queen (IM) makes a 'Castle' (which moves as Pawn, and is thus very weak), and combining it with Bishop (M) gives a Knight (which should also be weaker than B a separate Bishop + Pawn).
luciofilho
Posts: 2
Joined: Tue Nov 03, 2020 3:15 am
Full name: LUCIO J. PATROCINIO FILHO

Re: Turnover Chess

Post by luciofilho »

Thank you guys for your interest at this game I designed to let my vision to chess variants universe.
If you don't mind, I wonder if you guys can try Super C, it is a Turnover game variant, designed to solve this lack of complexity, only found on traditional chess, I'm talking about checkmate complexity.
On Turnover, Castles are like chess King, this concept is based on castles from the Middle Ages, so a King will be inside the last castle, but as castles were moving like pawns, this rule resulted in a loss of complexity on check and checkmate, letting as only option to try to block attack line and this issue was really an outstanding debt on that variant. After hundreds of matches, I finally realized that this is a serious issue and as a design solution I decided to change castles move rule, so now, when a castle is in check, it is as if it's challenged to reveal its King, then castle in check must move like King's move from traditional chess. This small change brought back chess checkmate mechanics, plus Turnover pieces combination complexity, forged an interesting balance of forces. I am quite satisfied at the moment.
I'm try to develop a online version by myself, at same time as I learn javascript and all that kind of things, so you can follow my development on: http://luciofilho.github.io

Thank you.
Madeleine Birchfield
Posts: 512
Joined: Tue Sep 29, 2020 4:29 pm
Location: Dublin, Ireland
Full name: Madeleine Birchfield

Re: Turnover Chess

Post by Madeleine Birchfield »

First figure out the board representation you wish to use for this game. Then easiest would be to use neural network methods, whether NNUE with alpha-beta search, or CNN with MCTS.
luciofilho
Posts: 2
Joined: Tue Nov 03, 2020 3:15 am
Full name: LUCIO J. PATROCINIO FILHO

Re: Turnover Chess

Post by luciofilho »

Hello, Madeleine.
I found your approach very interesting and scientific. I will learn these methods, even though I am far from reaching an AI implementation. I'm alone in this endeavor, so everything is very slow. Thank you.