Re: [Computer-go] CGOS rating has drifted a lot
Hi Remi, This has latest 19x19.pgn and some tools to make bayes.html. http://www.yss-aya.com/20210112cgos_bin.tar.bz2 I got these tools from Joshua Shriver. Thanks, Hiroshi Yamashita Hi Hiroshi, I tried to compile my old code, but it turns out to be a bit too complicated because it was using tensorflow. I am not using tensorflow any more, and don't want to try to reinstall and recompile it. But I would like to try to take a look at the data. Can I download it anywhere? Can you send the PGN of the 3 lists to me? Thanks, Rémi ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] CGOS rating has drifted a lot
Hi Hiroshi, I tried to compile my old code, but it turns out to be a bit too complicated because it was using tensorflow. I am not using tensorflow any more, and don't want to try to reinstall and recompile it. But I would like to try to take a look at the data. Can I download it anywhere? Can you send the PGN of the 3 lists to me? Thanks, Rémi ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] CGOS rating has drifted a lot
Hi Hiroshi, Thanks for your work. In both lists CrazyStone-18.04 is ahead of CrazyStone-81-TitV, which is really completely wrong. I will reconnect CrazyStone-18.04 soon. It may help to adjust things a little bit. Rémi ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] CGOS rating has drifted a lot
Hi, Deleting losing on time games, and deleting losing on time games and same program matches are follows. 19x19 BayesElo (deleting losing on time games) http://www.yss-aya.com/cgos/19x19/bayes_20210109_del_timeout.html 19x19 BayesElo (deleting losing on time games and same program matches) http://www.yss-aya.com/cgos/19x19/bayes_20210109_del_timeout_same_program_match.html Top rating and ELFv2 (delete on time) black_hole_v75245 LZ_05db_ELFv2_p800 4005 Top rating and ELFv2 (delete on time and same program matches) black_hole_v64958 LZ_05db_ELFv2_p800 3855 Top rating and ELFv2 (original) Katago_40b_s37_4V4762 LZ_05db_ELFv2_p800 3686 http://www.yss-aya.com/cgos/19x19/bayes.html I had thought top rating of "delete on time and same program matches" was less than original, but it is not. The matches between these names are deleted. For example, if black name has "LZ" and white name has "ELF", this game is deleted. LZ | lz | Leela | ELF kata | Kata | KG | k40b | k160 | Ka Zen CrazyStone | CS Rnz | rn | Rn Aya myCtest AQ Dirichlet CGI DCNN- Odin 731369 ... number of games 721122 ... number of games after deleting timeout 592249 ... number of games after deleting timeout and same program matches Thanks, Hiroshi Yamashita ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] CGOS rating has drifted a lot
Hi Remi, I agree with you losing on time against weak program is bad effect for ratings. And I think same program matches with different playouts, like 400 playouts vs 800 playouts are worse. There are too many Leela Zero vs Leela Zero, and recently KataGo vs KataGo. Maybe for human, current top program rating is not so high. I will try to calculate BayesElo without losing on time and same program matches before very long. Thanks, Hiroshi Yamashita ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
Re: [Computer-go] CGOS rating has drifted a lot
您好!我是江苏君华特种工程塑料制品有限公司的总经理李军,我的手机号码为13382868677,感谢您发来的邮件!我会尽快处理答复您! Hello,I am Jolly Li, the general manager of Jiangsu Junhua High Performance Engineering Plastic Products Co.,Ltd. Thanks for your mail ,it is well received,I will get back to you as soon as possible. ___ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go