From Versteeg@voeding.tno.nl Tue Apr 27 06:25:08 1999 Received: from frontier.tno.nl (frontier.tno.nl [134.221.1.2]) by clinch.math.niu.edu (8.9.1a/8.9.1) with ESMTP id GAA00776 for ; Tue, 27 Apr 1999 06:25:07 -0500 (CDT) From: Versteeg@voeding.tno.nl Received: from ntexch1.voeding.tno.nl (NTEXCH1.VOEDING.TNO.NL [134.221.116.19]) by frontier.tno.nl (8.8.8+Sun/8.8.8) with ESMTP id NAA23169 for ; Tue, 27 Apr 1999 13:24:36 +0200 (MET DST) Received: by NTEXCH1.VOEDING.TNO.NL with Internet Mail Service (5.0.1460.8) id ; Tue, 27 Apr 1999 13:24:36 +0200 Message-ID: <10F8FCD13C51D111ABED0000F822DCDF01466403@NTEXCH1.VOEDING.TNO.NL> To: rusin@vesuvius.math.niu.edu Subject: RUMMIKUB in rec.games.abstract Date: Tue, 27 Apr 1999 13:24:33 +0200 MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.0.1460.8) Content-Type: text/plain Status: RO Hello, Let me introduce myself, I'm Andre Versteeg, student Artificial Intelligence at the VU (free university, amsterdam). Recently I became interested in how to let a computer play rummikub. I looked at your homepage to get some info about the person I'm addressing. I saw your rummikub question in rec.games.abstract (1998/01/06). I hope you still remember it: It is still saved on dejanews, see: http://www.dejanews.com/[ST_rn=ps]/dnquery.xp?ST=PS&QRY=rummikub&defaultOp=A ND&DBS=1&format=terse&showsort=score&maxhits=25&LNG=ALL&subjects=&groups=rec .games.abstract&authors=&fromdate=&todate= >Actually, a player is given two disjoint sets A and B of the pieces, with >A known to be partitioned already; the real question is to decide which >subsets S of B make A union S partitionable -- basically the larger S >is, the better.) Have you developed a method for a rummikub computer player based on this? If so, does it produce good results (smart computer moves)? I have downloaded two prograns today: rummi22.zip and rummikub.zip. Both are < 100k and probably DOS programs. Do you know any newer programs? Do you have any idea what methods they use? What are the standard rules? Are they called 'sambra' or 'sarba'? I have a theory about calculating an optimal move in rummikub, you might want to discuss it. I'm no mathematics wizard, my interests are program design/implementation, especially for AI and games with AI. Hoping to hear from you, Greetings, Andre Versteeg (alverste@cs.vu.nl / versteeg@voeding.tno.nl) From rusin@math.niu.edu Wed May 5 17:26:32 1999 Received: from vesuvius.math.niu.edu (vesuvius.math.niu.edu [131.156.3.93]) by clinch.math.niu.edu (8.9.1a/8.9.1) with ESMTP id RAA12403; Wed, 5 May 1999 17:26:31 -0500 (CDT) From: Dave Rusin Received: (from rusin@localhost) by vesuvius.math.niu.edu (8.8.8/8.8.5) id RAA16534; Wed, 5 May 1999 17:26:31 -0500 (CDT) Date: Wed, 5 May 1999 17:26:31 -0500 (CDT) Message-Id: <199905052226.RAA16534@vesuvius.math.niu.edu> To: Versteeg@voeding.tno.nl Subject: Re: RUMMIKUB in rec.games.abstract Cc: rusin@math.niu.edu Status: R Sorry about not responding sooner. >Have you developed a method for a rummikub computer player based on this? No. This is one game for which I have not sat down to write a simulation. Two issues are (a) It is clear that there are only finitely many possible ways to partition a set of tiles, so we may examine all of them to decide which partitions are legitimate according to rummikub rules. But it's NOT clear (to me) that there is an efficient way to do this; is there an algorithm which can analyze N tiles in a time which is polynomial in N? (b) Once one has decided what all the possible moves are, if none of those moves results in winning the game, how does one decide which move is best? Obviously in general we wish to play as many tiles as possible but this is not always true. So what's a good scoring function? >Do you know any newer programs? no >Do you have any idea what methods they use? no >What are the standard rules? Are they called 'sambra' or 'sarba'? I don't know those words; but then, I'm not at all an expert. My daughter plays one of the DOS games a lot at home. The rules do not quite match the ones in the (physical) RummiKub game we also own. This computer game insists that your first play be of runs and sets made completely with _tiles from your own hand_. It counts a blank tile as 25 points in this case so since a 30-point play is what is necessary to enter the game, it is very easy to start if you have any blanks! Also this computer game allows you to move the tiles around in any way you like when it is your turn (after you have entered with a 30-point play); the physical game we have forbids some of these rearrangements (e.g. it only allows you to replace a blank tile already on the board with the corresponding tile _taken from your own hand_) >I have a theory about calculating an optimal move in rummikub, If you would like to send it to me I will try to take a look at it when I have the chance. Good luck with your game! dave From wstomv@win.tue.nl Fri Jan 17 13:30:50 2003 Received: from kweetal.tue.nl (kweetal.tue.nl [131.155.2.7]) by sanford.math.niu.edu (8.12.5/8.12.5) with ESMTP id h0HJUmRh012601 for ; Fri, 17 Jan 2003 13:30:49 -0600 (CST) Received: from wsinpa16.win.tue.nl (wsinpa16.win.tue.nl [131.155.70.14]) by kweetal.tue.nl (8.12.3/8.12.3) with ESMTP id h0HJV0Wq37856852; Fri, 17 Jan 2003 20:31:00 +0100 (MET) Received: (from wstomv@localhost) by wsinpa16.win.tue.nl (8.12.6/8.12.3) id h0HJUrt7014425; Fri, 17 Jan 2003 20:30:53 +0100 (MET) Date: Fri, 17 Jan 2003 20:30:53 +0100 From: Tom Verhoeff To: rusin@math.niu.edu Subject: Rummikub rules correction (?) Message-ID: <20030117193053.GA14420@win.tue.nl> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.3.25i Status: O Content-Length: 1395 Dave, While searching the web for good algorithms to decide on "divisibility" in Rummikub, I hit your page There are two corrections I have: You write about the three subsets: "3. Each person's "rack pile" R_i (tile markings invisible to player i only)" Of course, "invisible" must be "visible". More serious is that I miss a restriction on the first (nonempty) transfer by a player from R_i to T . You write "Until a player may pass a set S with numerical sum 30 or more from R_i to T, that player must not pass anything" However, if I recall correctly, the first such set that any players passes to T must also satisfy the additional requirement that S itself be divisible. On the first move, a player may not "use" any of the tiles in T . By the way, I have not found any references to algorithms for deciding about divisibility. Time to look into this myself. Just in case it interests you: I have a web site with an optimal solitaire Yahtzee player: Best regards, Tom Verhoeff -- E-MAIL: T.Verhoeff @ TUE.NL | Fac. of Math. & Computing Science PHONE: +31 40 247 41 25 | Eindhoven University of Technology FAX: +31 40 245 17 33 | PO Box 513, NL-5600 MB Eindhoven http://www.win.tue.nl/~wstomv/ | The Netherlands