[computer-go] cgos 13x13 seems down
No game have started for half an hour, so it seems to have a problem. And if some other people want to play on 13x13 don't be afraid, my bot will be happy to play with yours. Tom -- Thomas LavergneEntia non sunt multiplicanda praeter necessitatem. (Guillaume d'Ockham) [EMAIL PROTECTED]http://oniros.org ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Kaori-Crazystone
On Thu, 2008-09-04 at 18:07 +0200, Rémi Coulom wrote: When the playouts evaluate a critical semeai the wrong way, then no supercomputer can help, even at long time control. Semeais require a better algorithm, because no computing power can search them out with a tree, and playouts have to be extremely intelligent in order to evaluate them correctly. Rémi Rémi, I have no doubt that is true and that ways should be found to deal with this. It will probably happen like in chess where at some point players learn how to play against machines and this will suppress the ratings for a while. I fully expect that to happen. This happened around the time that chess playing computers suddenly became a commodity item in the late 70's early 80's approximately. It was almost as if computers stopped improving for 2 or 3 years even though they really were getting better. I don't think this will have any affect on their scalability - it is just a reflection on how difficult go is to play for computers. There were also seemingly insurmountable problems in chess just like this where people said no amount of search will fix it but it didn't stop the very slow but gradual improvement. To this very day it's possible to find chess positions that computers play stupidly, but it's getting really rare. - Don ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] cgos 13x13 seems down
I will also run Valkyria on CGOS 13x13 over the weekend, (or long as things are stable). -- Magnus Persson Berlin, Germany ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] semeai
On Thu, 2008-09-04 at 18:07 +0200, Rémi Coulom wrote: When the playouts evaluate a critical semeai the wrong way, then no supercomputer can help, even at long time control. Semeais require a better algorithm, because no computing power can search them out with a tree, and playouts have to be extremely intelligent in order to evaluate them correctly. Has anyone tried implementing the ideas in Richard Hunter's Counting Liberties in a manner which can be used to guide the playouts? Analysis should be able to determine the outcome of a semeai, and use that information to protect its lead with minimal expenditure, or conversely to exploit the threat of resurrection to improve some other aspect of the board. By the middle game, humans will partition the board into areas with little or no interaction. Group A cannot be killed if the owner reacts appropriately to threats; group B cannot be defended, if the opponent plays correctly; group C is alive, if group D is killed. Humans know about nakade and ko, and exploit their knowledge to do efficient local tree search, and use that local knowledge to guide the global search for the best move. Even in the opening, humans have a sense whether a pair of stones have enough room to expand and make two eyes, or whether they're cramped for space. 19x19 differs from 9x9 in that it can be more readily partitioned; in smaller boards, everything often closely interacts with everything else. http://senseis.xmp.net/?CountingLibertiesAndWinningCapturingRaces Terry McIntyre [EMAIL PROTECTED] Go is very hard. The more I learn about it, the less I know. -Jie Li, 9 dan ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] semeai
On Sep 5, 2008, at 7:49 AM, terry mcintyre wrote: 19x19 differs from 9x9 in that it can be more readily partitioned; in smaller boards, everything often closely interacts with everything else. I have also found this to be true. I tried various ideas relating to finding the important areas of the board this summer, and they tend not to help much on 9x9. I hope to work on 13x13 in the near future, so I hope the 13x13 CGOS will still be up (and there will be some competition). Peter Drake http://www.lclark.edu/~drake/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Playout acceleration
Has anyone tried heavy, slow playouts near the tree and light, fast playouts near the end of the game? I'm calling this playout acceleration because it starts slow and speeds up. You could have many different playout weights/speeds in a single playout. It seems like a reasonable idea to me since the moves near the tree should be the most important. Thoughts? ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Playout acceleration
There is this: http://webdisk.lclark.edu/drake/publications/GAMEON-07-drake.pdf The idea of gradual acceleration is intriguing, although I shudder to think of setting the parameters for such a process. Peter Drake http://www.lclark.edu/~drake/ On Sep 5, 2008, at 10:43 AM, Michael Williams wrote: Has anyone tried heavy, slow playouts near the tree and light, fast playouts near the end of the game? I'm calling this playout acceleration because it starts slow and speeds up. You could have many different playout weights/speeds in a single playout. It seems like a reasonable idea to me since the moves near the tree should be the most important. Thoughts? ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: Rating systems (was Re: [computer-go] Kaori-Crazystone)
You would have to ask these questions of Paul. He is an extremely serious and careful person, so while I would find it hard to believe that every person had exactly the same rating down to 0.01, it must have been very close when the entire collection of AGA members was considered. I do not believe that Paul would have allowed the change in the previous rules if the data were not convincing. But I was not involved at the time, so he would know and I only know the story as it was told to me (and how it is being used as the basis for rating computer Go playing programs now). Cheers, David On 5, Sep 2008, at 1:46 AM, Robert Jasiek wrote: David Doshay wrote: Two separate rating tables were kept, one for handicap games and another for non-handicap games. Over time it turned out that the ratings for individuals converged Did they converge for each person individually or converge only for all persons on average? Did the convergence occur for all persons regardless of whether they played Black the more often in handicap games? Did fixed versus free handicaps make a difference? Would altering the handicap density per rank difference / komi system have made a difference? IOW, was the conclusion well justified or misinterpretation? -- robert jasiek ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] semeai
From: Bob Hearn [EMAIL PROTECTED] On Sep 5, 2008, at 10:34 AM, Robert Jasiek wrote: terry mcintyre wrote: Has anyone tried implementing the ideas in Richard Hunter's Counting Liberties ... So although Hunter's study is a good fundament from which to start working (in expert systems or for becoming a stronger player), one should not expect complete generality prematurely. But conceivably, it is not necessary to completely analyze semeai statically, merely to produce some better heuristics so that the playouts do a better job with semeai, correct? Better heuristics would certainly help, but my hope is that local search trees can be precomputed. It's ok for a top level analysis to try all sorts of goofy moves, looking for the one tesuji which denies eyespace to a group, but if all these variations are tried millions of times, it greatly increases the size of the search tree. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] semeai
Bob Hearn wrote: But conceivably, it is not necessary to completely analyze semeai statically, merely to produce some better heuristics so that the playouts do a better job with semeai, correct? In principle, right. I'd guess, things would be different during middle game in an expert system and during late MC playouts. Regardless, nothing needs to be static; if you use good heuristics, you can have everything dynamic:) Needless to say, it boils down to speed. So if an MC program calculates countless playouts, a too sophisticated static evaluation does more harm than good. Currently for MC playouts it seems though that the programs should get nakade right first before bothering about semeais. Therefore currently I think that static semeai evaluation is more useful for classical expert system programs. -- robert jasiek ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Playout acceleration
There might be no need for heavier playouts to be slower. Sometimes they are even faster. (maybe it was in case of Crazy Stone?) On Fri, Sep 5, 2008 at 18:43, Michael Williams [EMAIL PROTECTED] wrote: Has anyone tried heavy, slow playouts near the tree and light, fast playouts near the end of the game? I'm calling this playout acceleration because it starts slow and speeds up. You could have many different playout weights/speeds in a single playout. It seems like a reasonable idea to me since the moves near the tree should be the most important. Thoughts? ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] cgos 13x13 seems down
On Fri, 5 Sep 2008, Magnus Persson wrote: I will also run Valkyria on CGOS 13x13 over the weekend, (or long as things are stable). One anchor would be nice. Christoph ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Playout acceleration
Yes, I tried heavy playouts for N plys, then switching to light.?It didn't really speed things up all that much but it weakened my bot quite a bit, on a per playout basis, resulting in a clear net loss. ? I tried ladder reading for the first N plys, then no ladder reading. The results were muddled. Currently, I use slightly more stochastic playout rules for the first 7 plys of the playouts, than for the remainder. It gives me a small, but statistically significant, strength increase against Gnugo. - Dave Hillis -Original Message- From: Michael Williams [EMAIL PROTECTED] To: computer-go computer-go@computer-go.org Sent: Fri, 5 Sep 2008 1:43 pm Subject: [computer-go] Playout acceleration Has anyone tried heavy, slow playouts near the tree and light, fast playouts near the end of the game? I'm calling this playout acceleration because it starts slow and speeds up. You could have many different playout weights/speeds in a single playout. It seems like a reasonable idea to me since the moves near the tree should be the most important. Thoughts?? ? ___? computer-go mailing list? [EMAIL PROTECTED] http://www.computer-go.org/mailman/listinfo/computer-go/? ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] MoGo v.s. Kim rematch
MoGo and Myungwan Kim will hold an exhibition rematch at the Cotsen Open on Saturday September 20. The exhibition will start at about 5pm Pacific Daylight time. As probably known by all on this list, MoGo won the last game, held at the US Go Congress in Portland Oregon, when it was given a 9 stone handicap and played with a one hour time limit. At this time the expected handicap will be 7, and it is not clear if there will be one game or two. It is not known at this time how many cores MoGo will be running on. Mr Kim has asked for MoGo to be given 90 minutes because he saw how much the increase in time from 15 minutes to one hour increased its playing strength. Mr. Kim has also asked that there be only one or at most 2 blitz games at the start. The MoGo team also wants to have some 9x9 games, but Mr Kim does not feel familiar enough with 9x9 to play those games, but he is searching for an alternate strong player. MoGo has some new features for 9x9, and the team is anxious to see the newest code in action. Cheers, David ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] MoGo v.s. Kim rematch
Great news! Look forward to seeing it happen. I hope Mogo has some great hardware. - Don On Fri, 2008-09-05 at 15:54 -0700, David Doshay wrote: MoGo and Myungwan Kim will hold an exhibition rematch at the Cotsen Open on Saturday September 20. The exhibition will start at about 5pm Pacific Daylight time. As probably known by all on this list, MoGo won the last game, held at the US Go Congress in Portland Oregon, when it was given a 9 stone handicap and played with a one hour time limit. At this time the expected handicap will be 7, and it is not clear if there will be one game or two. It is not known at this time how many cores MoGo will be running on. Mr Kim has asked for MoGo to be given 90 minutes because he saw how much the increase in time from 15 minutes to one hour increased its playing strength. Mr. Kim has also asked that there be only one or at most 2 blitz games at the start. The MoGo team also wants to have some 9x9 games, but Mr Kim does not feel familiar enough with 9x9 to play those games, but he is searching for an alternate strong player. MoGo has some new features for 9x9, and the team is anxious to see the newest code in action. Cheers, David ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] cgos game archive
Cgos game archives have been updated so we now have all of August's games. - Don ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] cgos 13x13 seems down
Christoph Birk: [EMAIL PROTECTED]: On Fri, 5 Sep 2008, Magnus Persson wrote: I will also run Valkyria on CGOS 13x13 over the weekend, (or long as things are stable). One anchor would be nice. OK, I will run Gnugo-3.7.10-a4 soon. Hideki Christoph ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ -- [EMAIL PROTECTED] (Kato) ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/