Re: [computer-go] Paper presents results on proximity heuristic
> This was not tested in any formal way, but including the book does seem > to increase the chance that the program will open with E5 (which I > believe is the correct opening move on 9x9) ... Just a side note, as I've spent a lot of time studying high-level 9x9 games. I've seen strong players win against other strong players starting with any of the 6 plausible openings moves: 5,5; 5,4; 4,4; 5,3; 3,4; 3,3. When the pros switched form 5.5 to 6.5pt komi there were more 4,4 openings played and fewer 5,5 openings. But this does not mean that 4,4 is better than 5,5. Anyway, my point was, it would be dangerous to think your program is playing better because it starts choosing 5,5 more often. However, choosing one of those six moves more than a first move on the 1st or 2nd lines could probably be considered an improvement. Darren ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
On 8, Feb 2007, at 9:34 AM, Eric Boesch wrote: On 2/7/07, David Doshay <[EMAIL PROTECTED]> wrote: On 7, Feb 2007, at 1:35 PM, Don Dailey wrote: > Although suicide can occasionally be the best move in some > rule-sets, I think it weakens your program to include it, At best you are going to get a ko threat, so it requires a pretty sophisticated program to know how and when to use it. There are weird cases where suicide is more than just a ko threat. http://www.goban.demon.co.uk/go/bestiary/rule_challenge.html (Just picking nits.) nit effectively picked ... In light of these very odd situations, please change my comment above to read : ... so it requires a *very* sophisticated program to ... ;^) Cheers, David ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
On 2/7/07, David Doshay <[EMAIL PROTECTED]> wrote: On 7, Feb 2007, at 1:35 PM, Don Dailey wrote: > Although suicide can occasionally be the best move in some > rule-sets, I think it weakens your program to include it, At best you are going to get a ko threat, so it requires a pretty sophisticated program to know how and when to use it. There are weird cases where suicide is more than just a ko threat. http://www.goban.demon.co.uk/go/bestiary/rule_challenge.html (Just picking nits.) ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
On Feb 8, 2007, at 4:10 AM, terry mcintyre wrote: Thanks, Peter! I have a question or two regarding the opening book, based on a collection of 3000 9x9 games provided by Nici Schraudolph. Who played the games in this collecton - pros, strong amateurs, or go programs? I believe some of each, but mostly amateurs. Second, were any statistics on the number of game moves "in book" versus "out of book" collected? No. Lastly, it was assumed that the book moves were winning moves; was this hypothesis ever tested on a move-by-move basis, whether against GnuGo or itself? This was not tested in any formal way, but including the book does seem to increase the chance that the program will open with E5 (which I believe is the correct opening move on 9x9) and that it will occupy 3-3 points before doing anything else. Peter Drake Assistant Professor of Computer Science Lewis & Clark College http://www.lclark.edu/~drake/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
Thanks, Peter! I have a question or two regarding the opening book, based on a collection of 3000 9x9 games provided by Nici Schraudolph. Who played the games in this collecton - pros, strong amateurs, or go programs? Second, were any statistics on the number of game moves "in book" versus "out of book" collected? Lastly, it was assumed that the book moves were winning moves; was this hypothesis ever tested on a move-by-move basis, whether against GnuGo or itself? Many thanks for an informative paper! - Original Message From: Peter Drake <[EMAIL PROTECTED]> By the way, the paper was rejected on first submission, largely because we were just testing Orego against itself. We're now testing Orego against GNU Go and have a revised version: https://webdisk.lclark.edu/xythoswfs/webui/_xy-2352013_1-t_Gct7yJ5s%22 (Markus, could you change the link and title? This was DPSV07.) Peter Drake Assistant Professor of Computer Science Lewis & Clark College http://www.lclark.edu/~drake/ On Feb 7, 2007, at 12:36 PM, Chris Fant wrote: > In this paper, you say that you limit the number of moves to > BoardArea*2 during the playouts. For me, this barely increases the > playout rate and slightly reduces the strength (perhaps not > statistically significant). > > Your paper does not mention suicide. Prohibiting single-stone suicide > during playout gave me a nice increase in playout rate and strength. > > > On 11/28/06, Peter Drake <[EMAIL PROTECTED]> wrote: >> Here it is: >> >> https://webdisk.lclark.edu/xythoswfs/webui/_xy-2115826_1-t_OX34gnaB Bored stiff? Loosen up... Download and play hundreds of games for free on Yahoo! Games. http://games.yahoo.com/games/front___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
By the way, the paper was rejected on first submission, largely because we were just testing Orego against itself. We're now testing Orego against GNU Go and have a revised version: https://webdisk.lclark.edu/xythoswfs/webui/_xy-2352013_1-t_Gct7yJ5s%22 (Markus, could you change the link and title? This was DPSV07.) Peter Drake Assistant Professor of Computer Science Lewis & Clark College http://www.lclark.edu/~drake/ On Feb 7, 2007, at 12:36 PM, Chris Fant wrote: In this paper, you say that you limit the number of moves to BoardArea*2 during the playouts. For me, this barely increases the playout rate and slightly reduces the strength (perhaps not statistically significant). Your paper does not mention suicide. Prohibiting single-stone suicide during playout gave me a nice increase in playout rate and strength. On 11/28/06, Peter Drake <[EMAIL PROTECTED]> wrote: Here it is: https://webdisk.lclark.edu/xythoswfs/webui/_xy-2115826_1-t_OX34gnaB Peter Drake Assistant Professor of Computer Science Lewis & Clark College 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 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] Paper presents results on proximity heuristic
On 7, Feb 2007, at 1:35 PM, Don Dailey wrote: Although suicide can occasionally be the best move in some rule-sets, I think it weakens your program to include it, At best you are going to get a ko threat, so it requires a pretty sophisticated program to know how and when to use it. To put all of those considerations in the MC playout will slow things down, and right now it seems better to get better statistics via more playouts. Cheers, David ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
Okay, thanks for the feedback. I mentioned that I was allowing multi-stone suicide a couple days ago but no one said anything. It seems little more complicated to check for than single-stone suicide when only tracking pseudo-liberties. But I will get it in there and see what kind of improvement happens and how much it affects the speed. On 2/7/07, Don Dailey <[EMAIL PROTECTED]> wrote: You should never play suicide, whether multiple or single stone in the play-out portion of the search - ESPECIALLY when it's not legal anyway in the rule-set you are using. Although suicide can occasionally be the best move in some rule-sets, I think it weakens your program to include it, even in the UCT tree portion of an MC search. If the rule-set allows suicide, I believe it's best to just let your program accept suicide moves from the opponent. - Don On Wed, 2007-02-07 at 13:23 -0800, Christoph Birk wrote: > On Wed, 7 Feb 2007, Chris Fant wrote: > > Your paper does not mention suicide. Prohibiting single-stone suicide > > during playout gave me a nice increase in playout rate and strength. > > Does you program allow multiple-stone suicide during playout? > myCtest does NOT allow it; and neither does GenericMC_ by Don. > > Christoph > > ___ > 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 mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
You should never play suicide, whether multiple or single stone in the play-out portion of the search - ESPECIALLY when it's not legal anyway in the rule-set you are using. Although suicide can occasionally be the best move in some rule-sets, I think it weakens your program to include it, even in the UCT tree portion of an MC search. If the rule-set allows suicide, I believe it's best to just let your program accept suicide moves from the opponent. - Don On Wed, 2007-02-07 at 13:23 -0800, Christoph Birk wrote: > On Wed, 7 Feb 2007, Chris Fant wrote: > > Your paper does not mention suicide. Prohibiting single-stone suicide > > during playout gave me a nice increase in playout rate and strength. > > Does you program allow multiple-stone suicide during playout? > myCtest does NOT allow it; and neither does GenericMC_ by Don. > > Christoph > > ___ > 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] Paper presents results on proximity heuristic
On Wed, 7 Feb 2007, Chris Fant wrote: Your paper does not mention suicide. Prohibiting single-stone suicide during playout gave me a nice increase in playout rate and strength. Does you program allow multiple-stone suicide during playout? myCtest does NOT allow it; and neither does GenericMC_ by Don. Christoph ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Paper presents results on proximity heuristic
In this paper, you say that you limit the number of moves to BoardArea*2 during the playouts. For me, this barely increases the playout rate and slightly reduces the strength (perhaps not statistically significant). Your paper does not mention suicide. Prohibiting single-stone suicide during playout gave me a nice increase in playout rate and strength. On 11/28/06, Peter Drake <[EMAIL PROTECTED]> wrote: Here it is: https://webdisk.lclark.edu/xythoswfs/webui/_xy-2115826_1-t_OX34gnaB Peter Drake Assistant Professor of Computer Science Lewis & Clark College 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 mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/