Re: [Fwd: Re: [Bug-gnubg] Installation problem - 0.16 instead the wanted 0.15]
* Hardy Hübener [EMAIL PROTECTED] [070704 07:47]: Thanks Christian, my main confusion is, that I did a absolutly NEW, FRESH installation of GNUBG (supposed to be 0.15): Deleted the older versions before, installed the file gnubg-0.15-stable-20061119-setup.exe, which I got from the download section of the GNUBG website, from the link that was pointing to the 0.15 stable version. And the programm that is running after that installation is version 0.16!!??!!?? So your downgrade hints don't help me much :-( The download archive for the 0.15 version (on the windows download site) seems to contain 0.16 elements! Are you sure about this? How big is the binary? And do you see MT support when you open Settings - options -other? Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
* Achim Mueller [EMAIL PROTECTED] [070704 08:15]: Frank Berger made an announcement at gammonu and bgonline.org that someone made a simulation of 1000 25p matches with gnubg, snowie bgblitz and jellyfish. Results will be published today. I'm not sure whether he described correctly, how the gnubg setup setup was, but it might be that gnubg has used zadeh met as default. For this is the worst met of all - I estimated at least 2% worse than snowie met - my suggestion is to use either g11 or mec26 as default met. Is it enough to change this line in matchequity.c? pmd-mi.szFileName = BAD_CAST strdup ( met/zadeh.xml ); Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
On 7/4/07, Achim Mueller [EMAIL PROTECTED] wrote: Frank Berger made an announcement at gammonu and bgonline.org that someone made a simulation of 1000 25p matches with gnubg, snowie bgblitz and jellyfish. Results will be published today. I'm not sure whether he described correctly, how the gnubg setup setup was, but it might be that gnubg has used zadeh met as default. For this is the worst met of all - I estimated at least 2% worse than snowie met - my suggestion is to use either g11 or mec26 as default met. While I totally agree the default should be at least mec26, It is hard to believe 2%. where did you get this figure? Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
On 7/4/07, Achim Mueller [EMAIL PROTECTED] wrote: * Achim Mueller [EMAIL PROTECTED] [070704 08:15]: Is it enough to change this line in matchequity.c? pmd-mi.szFileName = BAD_CAST strdup ( met/zadeh.xml ); The default is set in gnubg.c somewhere. I changed it to g11 at some point, but I cannot remember if it was done for 0.15 as well and if somebody has the zadeh saved in their gnubgautorc that takes priority. Christian. ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
I recently installed gnubg (latest windows binaries) on a virgin windows system. Default was Zadeh. On 7/4/07 12:58 AM, Christian Anthon [EMAIL PROTECTED] wrote: On 7/4/07, Achim Mueller [EMAIL PROTECTED] wrote: * Achim Mueller [EMAIL PROTECTED] [070704 08:15]: Is it enough to change this line in matchequity.c? pmd-mi.szFileName = BAD_CAST strdup ( met/zadeh.xml ); The default is set in gnubg.c somewhere. I changed it to g11 at some point, but I cannot remember if it was done for 0.15 as well and if somebody has the zadeh saved in their gnubgautorc that takes priority. Christian. ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
* Joseph Heled [EMAIL PROTECTED] [070704 08:57]: On 7/4/07, Achim Mueller [EMAIL PROTECTED] wrote: Frank Berger made an announcement at gammonu and bgonline.org that someone made a simulation of 1000 25p matches with gnubg, snowie bgblitz and jellyfish. Results will be published today. I'm not sure whether he described correctly, how the gnubg setup setup was, but it might be that gnubg has used zadeh met as default. For this is the worst met of all - I estimated at least 2% worse than snowie met - my suggestion is to use either g11 or mec26 as default met. While I totally agree the default should be at least mec26, It is hard to believe 2%. where did you get this figure? I made a long simulation of snowie.met vs woolsey met 10,000 5p matches, it was 51.7% - 48.3%. A smaller simulation of snowie.met vs. zadeh.met (1,000 matches) showed 52.4% - 47.6%. I agree that this sample isn't big enough, but it was big enough for not trusting zadeh.met anymore. Didn't you also make a _very_ long simulation comparing mets? Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
* Christian Anthon [EMAIL PROTECTED] [070704 09:10]: On 7/4/07, Achim Mueller [EMAIL PROTECTED] wrote: * Achim Mueller [EMAIL PROTECTED] [070704 08:15]: Is it enough to change this line in matchequity.c? pmd-mi.szFileName = BAD_CAST strdup ( met/zadeh.xml ); The default is set in gnubg.c somewhere. I changed it to g11 at some point, but I cannot remember if it was done for 0.15 as well and if somebody has the zadeh saved in their gnubgautorc that takes priority. Christian. In gnubg.c it's: fprintf( pf, set matchequitytable \%s\\n, miCurrent.szFileName ); so it should be changed in matchequity.c, shoudn't it? Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Fwd: Re: [Bug-gnubg] Installation problem - 0.16 instead the wanted 0.15]
Thanks Christian, my main confusion is, that I did a absolutly NEW, FRESH installation of GNUBG (supposed to be 0.15): Deleted the older versions before, installed the file gnubg-0.15-stable-20061119-setup.exe, which I got from the download section of the GNUBG website, from the link that was pointing to the 0.15 stable version. And the programm that is running after that installation is version 0.16!!??!!?? So your downgrade hints don't help me much :-( The download archive for the 0.15 version (on the windows download site) seems to contain 0.16 elements! Hi, I think that may be my fault. I've recompiled the old 0.15 and updated it on gnubg.org, you may try ro download it again and check. In any case, it was just a minor issue (naming of th version was 0.16 instead of 0.15): no 0.16 elements were ni. MaX. ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
In 0.16: met = g_build_filename(PKGDATADIR, met, g11.xml, NULL); In 0.15 InitMatchEquity ( met/zadeh.xml, szDataDirectory ); Christian. ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Re: MT/SMP bug disappeared??!
Hi Jon, I did a bit of investigation. I think the problem is in the multi thread loop. The number of games isn't always correct (off by 1) and that hits both the average and the std err in the beginning of a rollout, but only the std. err. is affected in the long run because it is calculated recursively. Christian. ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Re: MT/SMP bug disappeared??!
* Jonathan Kinsey [EMAIL PROTECTED] [070704 10:11]: Are you running this on linux or windows? Linux (SuSE-10.1). Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
RE: [Bug-gnubg] new windows build
Hi Oystein, can you please make a new windows build before I go to Monte Carlo. I planned to burn a DVD including the newest stuff in case people will ask. Hi Achim, I've built a new distrib, it's available on www.gnubg.org. Code snapshot is latest available (20070703). Multi-thread exes are also available to download (at least 1 important bug in them, cube rollout issue). MaX. The multi-thread exe: gnubg crashes after a rollout Settings - Options - Other: Eval Threads 2 Start match - my turn to move - Ctrl + H - roll out n moves A window pops up: ** ERROR **: file bearoff.c: line 1702 (GetDistCompressed): assertion Failed: (FALSE) aborting... The same happens after analyzing and doing a rollout of n moves later. With Eval Threads 0 : The program stopped working I used a clean Windows Vista on Intel T2080 (Dual Pentium), installed a clean 20070703 copied the mt exes to the gnubg install directory and renamed them. N. No virus found in this outgoing message. Checked by AVG Free Edition. Version: 7.5.476 / Virus Database: 269.9.14/885 - Release Date: 3/7/2007 10:02 ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Installation problem - 0.16 instead the wanted 0.15
On 04.07.2007 09:21 Massimiliano Maini wrote: Hi, I think that may be my fault. I've recompiled the old 0.15 and updated it on gnubg.org, you may try ro download it again and check. In any case, it was just a minor issue (naming of th version was 0.16 instead of 0.15): no 0.16 elements were in. MaX. Hi MaX, indeed, now after installing the 0.15 it also says it's 0.15. Thanks. Will also investigate how stable 0.16 runs on my system and let you all know. Thanks a lot!!! Hardy ;-) -- Hardy's Backgammon Pages -- www.hardys-backgammon-pages.com ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
Michael Petch wrote: I recently installed gnubg (latest windows binaries) on a virgin windows system. Default was Zadeh. Hmmm I can see two reasons that might cause this: 1. There is a gnubgautorc file shipped with the distribution that says Zadeh. or: 2. Reading the g11 table fails, and zadeh is the fallback. MaX? is there a gnubgautorc in your distributions? Michael? Is it possible for you to read any other MET? -Øystein signature.asc Description: OpenPGP digital signature ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Default met
* Joseph Heled [EMAIL PROTECTED] [070704 09:59]: On 7/4/07, Achim Mueller [EMAIL PROTECTED] wrote: Didn't you also make a _very_ long simulation comparing mets? Yes, but I played each set of dice rolls twice, identical rolls but swapped. This handles the variance better, and I could not find a large difference if I recall correctly. I made a printout of doubling windows for the different equity tables. Just a few examples: Snowie Zadeh n3 - n2 (1) 57,7 - 73,0 55,3 - 71,7 n3 - n4 (1) 37,8 - 62,5 41,5 - 65,9 !! n4 - n2 (1) 65,8 - 81,2 63,3 - 79,4 !! n4 - n4 (1) 50,0 - 72,0 50,0 - 70,9 n4 - n4 (2) 50,0 - 66,4 50,0 - 67,7 The results of gnubg matches against itself with different mets must somehow reflect these differences. I'll check whether I can do a longer simulation of gnubg(g11) - gnubg(zadeh) here at work. Do you still have the framework of your rollouts? Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Re: MT/SMP bug disappeared??!
Christian Anthon wrote: Hi Jon, I did a bit of investigation. I think the problem is in the multi thread loop. The number of games isn't always correct (off by 1) and that hits both the average and the std err in the beginning of a rollout, but only the std. err. is affected in the long run because it is calculated recursively. It's possible - working out how many games have been played is difficult (when there are multiple threads working). What seems strange is that it only effects cube decision rollouts, or is that not the case? Jon signature.asc Description: OpenPGP digital signature ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Re: MT/SMP bug disappeared??!
On 7/4/07, Jonathan Kinsey [EMAIL PROTECTED] wrote: Christian Anthon wrote: Hi Jon, I did a bit of investigation. I think the problem is in the multi thread loop. The number of games isn't always correct (off by 1) and that hits both the average and the std err in the beginning of a rollout, but only the std. err. is affected in the long run because it is calculated recursively. It's possible - working out how many games have been played is difficult (when there are multiple threads working). What seems strange is that it only effects cube decision rollouts, or is that not the case? Jon I think it does only affect cube rollouts, and there is a difference in the increment criteria for cube rollouts. Christian. ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
[Bug-gnubg] Results from the bot shootout
From GammonU: BGBJellyfish Snowie 4.5 GnuBG 0.14 BGB -534504 481 Jellyfish466- 428 407 Snowie 4.5 496 572 - 489 GnuBG 0.14 519 593511 - According to Frank Berger the matches were setup bei Torsten Schoop. I don't know any details about the settings. Ciao Achim ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Results from the bot shootout
Achim Mueller wrote: From GammonU: BGBJellyfish Snowie 4.5 GnuBG 0.14 BGB -534504 481 Jellyfish466- 428 407 Snowie 4.5 496 572 - 489 GnuBG 0.14 519 593511 - It's 25 point matches, isn't it? It looks like GNU Backgammon beats them all. Is that the right interpretation? This was about the result I would expect in a match series. The point is that the positions where GNU Backgammon outperforms Snowie is more frequent than the positions where Snowie outperforms GNU Backgammon. Would anyone estimate the standard deviation of these numbers? -Øystein signature.asc Description: OpenPGP digital signature ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
Re: [Bug-gnubg] Results from the bot shootout
Or loses to them all. Who did those runs? On 7/5/07, Øystein Johansen [EMAIL PROTECTED] wrote: Achim Mueller wrote: From GammonU: BGBJellyfish Snowie 4.5 GnuBG 0.14 BGB -534504 481 Jellyfish466- 428 407 Snowie 4.5 496 572 - 489 GnuBG 0.14 519 593511 - It's 25 point matches, isn't it? It looks like GNU Backgammon beats them all. Is that the right interpretation? This was about the result I would expect in a match series. The point is that the positions where GNU Backgammon outperforms Snowie is more frequent than the positions where Snowie outperforms GNU Backgammon. Would anyone estimate the standard deviation of these numbers? -Øystein ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg
RE: [Bug-gnubg] Results from the bot shootout
On Behalf Of Joseph Heled Or loses to them all. Who did those runs? BGBJellyfish Snowie 4.5 GnuBG 0.14 BGB -534504 481 Jellyfish466- 428 407 Snowie 4.5 496 572 - 489 GnuBG 0.14 519 593511 - The bot in the left column (or row) against the bot in column 2,3,4,5 as an example: BGB won 534 matches against Jellyfish (quote from Frank Berger on bgonline.org, http://www.bgonline.org/forums/webbbs_config.pl?read=7784) N. On 7/5/07, Øystein Johansen [EMAIL PROTECTED] wrote: Achim Mueller wrote: From GammonU: BGBJellyfish Snowie 4.5 GnuBG 0.14 BGB -534504 481 Jellyfish466- 428 407 Snowie 4.5 496 572 - 489 GnuBG 0.14 519 593511 - It's 25 point matches, isn't it? It looks like GNU Backgammon beats them all. Is that the right interpretation? This was about the result I would expect in a match series. The point is that the positions where GNU Backgammon outperforms Snowie is more frequent than the positions where Snowie outperforms GNU Backgammon. Would anyone estimate the standard deviation of these numbers? -Øystein ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg No virus found in this incoming message. Checked by AVG Free Edition. Version: 7.5.476 / Virus Database: 269.9.14/885 - Release Date: 3/7/2007 10:02 No virus found in this outgoing message. Checked by AVG Free Edition. Version: 7.5.476 / Virus Database: 269.9.14/885 - Release Date: 3/7/2007 10:02 ___ Bug-gnubg mailing list Bug-gnubg@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnubg