From: darse@cs.UAlberta.CA (Darse Billings) Subject: Lines of Action: Standardized Rules Date: 6 Jun 2000 09:58:35 GMT Newsgroups: rec.games.abstract Summary: colours, draws, etc This is a repost of a reply to the loa-programmers e-mail group. Since it deals with standardization of LOA rules, it seems more appropriate to discuss it here. It also includes an interesting game to play over... BTW, there is now a Java applet available to play YL and Mona on the web. They are quite strong, even at one second per move. Connect via the U of A GAMES group page at: http://www.cs.ualberta.ca/~games/index.html or via our Lines of Action page at: http://www.cs.ualberta.ca/~darse/LOA/ - Darse. -- From: Darse Billings To: loa-programmers@egroups.com Date: Tue, 6 Jun 2000 01:47:12 -0600 Subject: Re: [loa-programmers] Rules > From: "Mark Winands" > I think it's wise to determine some rules Or know what the standards already are... :) > Rule 1: > Which side moves first? > a) White > b) Black > In some programs white starts, but in some other programs black > starts. > Carl von Blixen favours white, but Dave Dyer favours black. I think > it's better that white begins because that's the same as in checkers, > chess and draughts. Black moves first in the large majority of games, including checkers. Chess and draughts are pretty much the only notable exceptions. Some games deliberately leave the colours and first move unspecified. Hex, for example, is often red and blue, but it can be any two colours, and there is no pre-set colour for moving first. More important is to standardize the first player using the pieces at the top and bottom of the board. > Rule 2 > If both players reach a winning position then > a) the game is a draw (Von Blixen & Mind Sport Olympiad). > b) the player moving wins (Dyer & second edition of a gamut of games). > I use b, but why not a? We use the draw rule for simultaneous connections, simply because we knew that would be the rule used at the MSO. I was under the impression that Soucie's original rule was a win for the player moving, and that Sid Sackson got it *wrong* in "Gamut of Games". However, I may have gotten this from a thread on rec.games.abstract four or five years ago, and the information may not be reliable. I believe NOST declares a win for the moving player. From a logic standpoint, there are arguments in favour of either rule. The situation is rare, but probably less rare in games between two strong programs. After 378 games in the current match between Mona and YL, there have been 20 draws: 19 by repetition, and one by simultaneous connection. There was at least one other training game that ended in a draw by simultaneous connection, and it would have played a role in the outcome of searches many more times than that. Here is the most recent simul draw. I've looked at it briefly, and it does indeed look like both players are forced to take it. Since neither side had much of a positional advantage, awarding a win to the player making the final capture seems a bit arbitrary (since the final capture could have gone either way, in principle). YL vs Mona (3 minutes per move, 2-ply fixed openings, Game 190) 1. b1-d3 h5-f3 2. g1-e3 a6:d3 3. c1-f4 a2-c4 4. d1:f3 a7-c7 5. e8-e5 h3-f5 6. f3-e4 h4-f6 7. e3-d4 f5-d7 8. f1-f5 h7-e7 9. f8:h6 a3-c5 10. c8-b7 f6-d6 11. b7-d5 h2-h4 12. f5:d7 h4:h6 13. e1-f2 h6-f6 14. f2:c5 d3-b5 15. e4:e7 a4:d7 16. b8-e8 a5-b4 17. g8-g7 b4-b6 18. f4:f6== Draw by simultaneous connection +-----------------+ 8 | . . . B B . . . | 7 | . . W W B . B . | 6 | . W . W . B . . | 5 | . W B B B . . . | 4 | . . W B . . . . | 3 | . . . . . . . . | 2 | . . . . . . . . | 1 | . . . . . . . . | +-----------------+ a b c d e f g h [After playing over the game with Mona, it now appears that White may have held the upper hand going into the final stage of the game, and that Black was lucky to escape. In light of that, awarding the game to Black seems even less reasonable -drb]. I think the game might be enriched with the simul draw rule, and it doesn't have much negative impact if 95% of draws are by repetition anyway (either forced or semi-forced). > Rule 3 > If a player can't move then > a) he has to pass. > b) he has lost the game. > I use a, because I think passing in LOA is disadvantageous. > I've never seen such a situation, but I'm convinced it can occur > in a real game. This is known. The player must pass. > Rule 4 > What to do with repetitions? > a) The player about to repeat a position for the third time must > vary the move or lose. > b) If a player repeats a position for the third time, it's a draw. > c) Something else. > I use b. A draw by repetition is both simple and logically sound. In essence, the players have agreed that neither side can make progress safely. With non-monotonic piece-movement games (like chess and LOA), the underlying mathematical structure of the domain is a graph, not a tree. Repetitions (cycles) are completely natural. Incidentally, one could eliminate draws entirely, by using something like the super-ko rule in Go (you are simply forbidden to repeat the position). However, this isn't very elegant, and I don't think there is anything inherently bad about an occasional game being drawn. LOA is in no danger of suffering "draw death", like checkers and chess. If you want no draws whatsoever, play Hex. :) > Rule 5 > A draw can be agreed any time (Mind Sport Olympiad). > a) Why not? I think it's pretty cool > b) No way. If you play a game of LOA, there has to be a winner. Yngvi and I talked about this yesterday. He gave a very convincing argument that the players should be able to agree to any valid result the game allows. If we exclude all draws, then the players cannot agree to a draw. If draws are possible, then two players who are in agreement can easily arrange the desired outcome. Suppose an arbiter insists that two chess grandmasters play at least 20 moves. Two agreeing players could demonstrate their contempt by having one side capture all the opponent's pieces, and then having the bare King capture all of his pieces in return. Satisfied, sir? :) > What is the average game-length (in ply)? > I think it's 40 ply (half moves). I could compute an exact average for Mona vs YL, but I won't. It's pretty close to 40 ply, and *rarely* more than 60, playing to mate. - Darse. [I may compute the average once the match is finished. -drb] -- "There's no sense in searching for perfection when you're making successful mistakes" -The Tragically Hip.