9x9 BayesElo ratings
xela: I was curious about the /9x9GrandSlam idea, but instead of a synthetic tournament, I'd like to try a different analysis method. I converted the Minigo game list into a text file, so that I could feed the 684 games that have results into Rémi Coulom's
BayesElo software. Note that this is not good science! It's just for fun. The BayesElo ratings assume that players' strengths are constant throughout the 16 year period that the games were played, and that all games were played under the same conditions. Of course both these assumptions are not quite right.
The result is that the top 5 players are people who only played a relatively small number of games, so their Elo ratings are only accurate to within +/- 200. At number 6, with 56 games and an 80% win/loss rate, is Yamada Kimio. So perhaps he is the strongest at 9x9, or perhaps Cho Chikun is stronger, there aren't enough games to decide either way for sure.
Full results:
Rank Name Elo + - games score oppo. draws 1 Cho Chikun 2373 338 211 11 91% 2081 0% 2 Wang Qun 2355 572 306 3 100% 2125 0% 3 Suzuki Keiji 2340 477 262 5 100% 2067 0% 4 Takemiya Masaki 2284 209 180 12 67% 2184 0% 5 Ma Xiaochun 2283 265 193 11 82% 2082 0% 6 Yamada Kimio 2238 107 94 56 80% 2028 0% 7 Kurahashi Masayuki 2211 98 91 55 71% 2061 0% 8 Ishida Yoshio 2204 199 188 12 58% 2149 0% 9 Iguchi Hideichiro 2197 385 385 2 50% 2197 0% 10 Yuki Satoshi 2197 100 91 42 81% 2031 0% 11 Wang Jiankun 2196 376 314 3 67% 2126 0% 12 Yata Naoki 2184 106 96 50 72% 2033 0% 13 Okada Yumiko 2175 356 278 4 75% 2062 0% 14 Yamazaki Yoshihiro 2155 571 331 2 100% 1978 0% 15 Cho Hunhyun 2154 188 196 11 45% 2182 0% 16 Ruan Yunsheng 2143 389 389 2 50% 2144 0% 17 Den Isei 2133 252 238 7 57% 2092 0% 18 Fang Tianfeng 2125 372 311 3 67% 2056 0% 19 Yamada Takuji 2120 316 244 5 80% 1974 0% 20 Shimojima Yohei 2104 355 302 3 67% 2039 0% 21 Song Li 2104 292 249 6 67% 2014 0% 22 Ishii Shinzo 2101 254 249 4 50% 2128 0% 23 Yamada Wakio 2099 135 124 25 68% 2005 0% 24 Inori Yoko 2091 186 176 12 58% 2040 0% 25 Tanimura Yoshiyuki 2090 122 115 30 63% 2017 0% 26 So Yokoku 2088 289 262 5 60% 2041 0% 27 Yamada Noriyoshi 2088 134 127 28 61% 2017 0% 28 Tsukuda Akiko 2087 197 185 12 58% 2036 0% 29 Fujii Shuya 2084 225 198 9 67% 1996 0% 30 Son Hideyo 2078 404 620 1 0% 2197 0% 31 Ha Youngil 2078 404 620 1 0% 2197 0% 32 Chin Kaei 2078 404 620 1 0% 2197 0% 33 Iyama Yuta 2078 404 620 1 0% 2197 0% 34 Hoshikawa Takumi 2078 404 620 1 0% 2197 0% 35 Mannami Kana 2078 404 620 1 0% 2197 0% 36 Suzuki Ayumi 2078 404 620 1 0% 2197 0% 37 Aragaki Shun 2077 108 105 38 55% 2051 0% 38 Sasaka Shiro 2076 104 100 43 58% 2024 0% 39 Hane Naoki 2060 298 234 5 80% 1941 0% 40 Kudo Norio 2054 198 274 10 20% 2238 0% 41 Hasegawa Hiro 2053 181 179 10 50% 2071 0% 42 Ueki Yoshio 2049 131 125 27 59% 2000 0% 43 Kato Yuki 2046 361 304 3 67% 1980 0% 44 Cho Riyu 2043 277 252 5 60% 1998 0% 45 Saito Tadashi 2025 117 116 33 52% 2018 0% 46 Kim Pyonmin 2023 182 178 13 54% 2005 0% 47 Goto Shungo 2020 171 171 14 50% 2031 0% 48 Ishida Atsushi 2020 102 101 44 50% 2024 0% 49 Takagi Shoichi 2015 397 395 2 50% 2026 0% 50 Oya Koichi 2015 397 395 2 50% 2026 0% 51 Yamada Shiho 2012 124 123 30 50% 2020 0% 52 Takei Takashi 2011 164 153 15 60% 1968 0% 53 Oda Hiromitsu 2003 152 156 17 47% 2023 0% 54 Sano Takatsugu 2002 125 126 26 50% 2010 0% 55 Enda Hideki 2002 121 122 30 47% 2033 0% 56 Kubo Hideo 1999 378 531 1 0% 2101 0% 57 Iwamaru Taira 1998 246 229 5 60% 1972 0% 58 Yanaka Katsunori 1997 144 139 20 55% 1966 0% 59 Furuta Naoyoshi 1990 209 216 9 44% 2022 0% 60 Yokota Shigeaki 1988 264 264 4 50% 1989 0% 61 Araki Issei 1983 142 142 20 50% 1991 0% 62 Cho U 1982 402 612 1 0% 2099 0% 63 Ri Yo 1975 190 204 10 40% 2034 0% 64 Takabayashi Masahiro 1974 113 119 33 39% 2049 0% 65 Imai Kazuhiro 1973 164 181 14 36% 2058 0% 66 Kashiwabara Yasuto 1973 403 615 1 0% 2090 0% 67 Chinen Kaori 1971 358 358 2 50% 1969 0% 68 Shigeno Yuki 1971 344 608 2 0% 2164 0% 69 Hayashi Kozo 1969 123 127 27 44% 2008 0% 70 Konishi Kazuko 1965 136 147 22 36% 2054 0% 71 Yanagawa Hiromasa 1963 113 116 35 43% 2019 0% 72 Ko Reibun 1960 339 601 2 0% 2148 0% 73 Okada Shinichiro 1959 404 618 1 0% 2077 0% 74 Yamamoto Kentaro 1959 174 181 12 42% 1998 0% 75 Nakamura Shinya 1958 168 175 15 47% 1989 0% 76 Yukawa Mitsuhisa 1956 212 220 9 44% 2003 0% 77 Yoshida Shoji 1954 192 213 11 36% 2044 0% 78 Kageyama Toshiyuki 1950 128 129 23 48% 1974 0% 79 Umezawa Yukari 1940 395 585 1 0% 2053 0% 80 Kitano Ryo 1940 146 158 19 37% 2023 0% 81 Takanashi Seiken 1939 274 348 4 25% 2057 0% 82 Izumo Tetsuya 1933 171 187 13 38% 2002 0% 83 Takiguchi Masaki 1930 383 383 2 50% 1930 0% 84 Yamanouchi Masaki 1926 248 249 6 50% 1929 0% 85 Minematsu Masaki 1925 128 139 26 35% 2032 0% 86 Sasaki Tsuyoshi 1922 255 342 5 20% 2078 0% 87 Yashiro Kumiko 1914 305 359 3 33% 1979 0% 88 Hashiguchi Mika 1906 241 285 6 33% 1995 0% 89 Tsuda Kae 1902 357 602 2 0% 2110 0% 90 Catalin Taranu 1896 399 599 1 0% 2011 0% 91 Ashida Isoko 1893 383 548 1 0% 1998 0% 92 Mizuno Hiromi 1887 275 362 4 25% 2003 0% 93 Enda Yoichi 1887 134 151 24 29% 2029 0% 94 Nakano Yasuhiro 1886 169 198 14 29% 2023 0% 95 Fujiwara Katsuya 1878 334 580 2 0% 2059 0% 96 Sakakibara Fumiko 1873 144 156 20 35% 1981 0% 97 Tanaka Hideharu 1871 168 181 13 38% 1947 0% 98 Kiyonari Tetsuya 1867 276 360 4 25% 1987 0% 99 Tanimura Kuniko 1866 276 358 4 25% 1982 0% 100 Tafu Kae 1865 324 533 2 0% 2031 0% 101 Shiraishi Kyoko 1861 201 258 9 22% 2021 0% 102 Furuya Yutaka 1855 202 268 7 14% 2018 0% 103 Mitsunaga Junzo 1833 402 609 1 0% 1950 0% 104 Xu Rongxin 1824 306 587 3 0% 2056 0% 105 Honda Goro 1824 219 332 9 11% 2083 0% 106 Sumi Shinsuke 1822 200 269 10 20% 2013 0% 107 Yahata Koichi 1820 177 247 14 14% 2057 0% 108 Tanaka Chieko 1809 213 324 10 10% 2087 0% 109 Kirimoto Kazuo 1803 221 317 8 13% 2033 0% 110 Sekiyama Toshimichi 1784 188 257 12 17% 2004 0% 111 Yoshida Mika 1769 342 580 2 0% 1960 0% 112 Kori Toshio 1750 276 342 4 25% 1875 0% 113 Maeda Ryo 1744 277 541 5 0% 2051 0% 114 Sakakibara Masateru 1742 260 545 7 0% 2093 0% 115 Konno Tameto 1723 289 559 4 0% 1994 0% 116 Katsuma Shiro 1669 285 483 4 0% 1938 0%
To use BayesElo, you need to convert the result list to PGN format. The first five lines of the PGN file look like this:
[Black "Aragaki Shun"][White "Kurahashi Masayuki"][Result "1-0"] 1-0 [Black "Den Isei"][White "Kurahashi Masayuki"][Result "0-1"] 0-1 [Black "Oda Hiromitsu"][White "Den Isei"][Result "1-0"] 1-0 [Black "Ueki Yoshio"][White "Den Isei"][Result "0-1"] 0-1 [Black "Hayashi Kozo"][White "Ueki Yoshio"][Result "1-0"] 1-0
Then you use the following commands in BayesElo (settings recommended in a forum post by Coulom, sorry I don't have the link to hand):
readpgn gamelist.pgn elo offset 2000 advantage 0 drawelo 0.01 mm exactdist ratings
Monteo Interesting analysis. Today we can see who among them is the strongest 9x9 player, by using AI with the strongest neural net. I compared the performance of Cho Chikun, Yamada Kimio, and Go Seigen, by using AlphaGo Zero performance benchmarking method with an AI that plays 9x9 Go at a superhuman level for any komi and any board sizes. Among the three players when having Black, Cho Chikun is closest to the vicinity of Go divine.
Initial results:
Rank Name %Best Move %Performance benchmarking 1 Cho Chikun 33.33 -56 2 Go Seigen 24.52 -70 3 Yamada Kimio 33.33 -210
WGA: I think Yuki Satoshi 9p is probably the strongest player in the above list, based on the game results at 9 Dan Big Matches