Arimaa Forum - Print Page


    
      
        Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> General Discussion >> Rules Question
        
(Message started by: jdb on Nov 17^th, 2004, 12:49pm)

Title: Rules Question
Post by jdb on Nov 17^th, 2004, 12:49pm

I have a question about the rules regarding three fold repetition. Is the side to move considered? That is, if the same configuration of pieces occurs on the board, which gold and silver to move, are they considered the same position or different positions?

Title: Re: Rules Question
Post by MrBrain on Nov 17^th, 2004, 1:57pm

If it's not defined yet, I would recommend it be the same as chess, which would mean that the same side would have to have the move for it to be a truly equivalent position.

Title: Re: Rules Question
Post by Fritzlein on Nov 17^th, 2004, 6:34pm

I asked Omar before, and he said that the side to move is not considered. The position with Gold to move is the same as the position with Silver to move.

I think that I would be more comfortable if it were the same as chess instead, but the practical effect is very small. For the difference to come into play there would have to be an issue of the stronger side wishing to purposely "lose a tempo", which happens routinely in chess, but doesn't seem relevant in Arimaa.

If Omar wanted to change the rule to match that of chess, the most important benefit would be less confusion about the rule. Otherwise it hardly matters.

Title: Re: Rules Question
Post by MrBrain on Nov 18^th, 2004, 9:46am

I don't think how it happens or how likely it is to happen should be important here. To me, a same position of pieces, but with the other player to move, is a different "position". If you evaluated the score (based on a scoring function) of these two "positions" you would get a different answer. The positions are inherently different, despite the fact that the pieces are in the same positions.

Title: Re: Rules Question
Post by RonWeasley on Nov 18^th, 2004, 11:26am

Look at the end of game 9646 between bot_Loc and bot_Arimaazon. The repeated position was enforced by bot_Loc. It was willing to lose lots of tempos to try and frame the horse. Imagine if it was bot_Loc's turn. It would frame the horse and continue. Too bad bot Arimaazon wasn't smart enough to take full advantage of this.

I think this example supports MrBrain's point. I agree that the rule should be like chess for the same reasons. Further, as play improves, especially for the bots, we could see many more metastable positions in the opening and in the endgame.

It shouldn't be that hard. Just add "mover" to the state definition. Even Longbottom could do it.

Title: Re: Rules Question
Post by Fritzlein on Nov 18^th, 2004, 2:11pm

Game 9646 isn't the best example, because that was three-fold repetition with the same player to move each time, so it doesn' t matter which repetition rule is in place.

On the other hand, I agree that if a different player has the move, it's a different position, even when all the pieces are in the same place. I think I'll cast my vote for Omar to change this rule, because the same configuration of pieces with a different player to move isn't the same thing, and shouldn't count as repetition.

Although the practical difference is miniscule, I can imagine a situation where a certain configuration of pieces has occurred twice with me to move, but I couldn't win from that position. Later I could bring about the same configuration of pieces with my opponent to move (the is the relevant form of losing a tempo), and I could win from that position, but instead I am given a loss because I "repeated" the position. That's not the right rule, even if it never happens in an actual game.

Since both rules prevent infinite loops of moves equally well, it makes sense to have to better of the two.

Title: Re: Rules Question
Post by omar on Nov 21^st, 2004, 9:44pm

Actually the only repitions that I encounted when I was working on the rules was the ABA and ABCDA type. I never encountered the ABCA type repition and so had not really thought about it. But now that I think about it the ABCA repition is really ABCA' because even though the position looks the same its a different situation. I agree that the side to move should be considered in the repitition rule. So I will change it effective March 1st 2005. This is just so that the rules are not being changed in the middle of Arimaa season when we have tournaments going on.

Title: Re: Rules Question
Post by omar on May 29^th, 2005, 7:54am

Sorry Im quite a bit late on this change, but I just updated the server to consider the side to move in the repition rule.

If anyone notices any problems with the server please notify me throught the 'Contact' link. Thanks.

Title: Re: Rules Question
Post by 99of9 on May 29^th, 2005, 8:28am

You might have to notify Don and David of this change - I believe both have coded the original rule into their bots.

Title: Re: Rules Question
Post by Fritzlein on May 29^th, 2005, 9:39am

Thanks for making this change, Omar. Even though it has never been relevant as far as I know, it still makes more theoretical sense. I will update Wikipedia to reflect the change.

As for Don's bots not knowing the rule, they didn't seem to grasp the previous version of the rule either, as Arimaazilla et. al. regularly lost by repetition of position even before. Also, difference between the rules will probably never arise in 100,000 games. But I understand the desire to have the code know the rules theoretically 100% correctly, so it would be good to let Don and David know if they start actively developing their bots again.

Title: Re: Rules Question
Post by 99of9 on May 29^th, 2005, 5:04pm

Arimaazilla was made before Don started asking questions about the rule. I think it was for Occam_2004 (which is no longer available anyway) and the sample bot.