9x9 Skills [#8679]
2a0b:f4c0:016c:0004: 9x9 Skills
(2020-02-03 02:29) [#11487]
Monteo: He is the strongest 9x9 player based on a synthetic tournament (real games organized as a tournament).
Anonymous: What does "real games organized as a tournament" mean?
Monteo: It means a group of actual games of the actual players recorded during the Yomiuri TV show program. I put them into a tournament format, called 9x9 Grand Slam, to see who is the winner, and he is Yamada Kimio. See TheWayto9x9/9x9GrandSlam? for more information.
Anonymous: So you invented a tournament by using games from a non-tournament format (Yomiuri TV's 1987-2002 Mini-Go 1-Game Matches (ミニ碁一番勝負)) whose rules varied over the years and chose the winner based on undisclosed weighting by yourself?
Monteo: Thatís right. You can also invent another tournament and see whether you will reach the same conclusion or not, with the limited game records we have had so far. Or, to be more reliable, you can conduct a big data project to identify bad or divine moves in such pro games, to see how strong they are.
: The claim "Yamada is strongest at 9x9" is actually plausible
(2020-02-03 12:58) [#11488]
"You can also invent another tournament and see whether you will reach the same conclusion or not". Personally I don't think a knockout tournament is a good way to decide who's strongest. So I tried something different: see TheWayto9x9/9x9BayesEloRatings. It's indeed possible that Yamada is strongest (amongst pros active up to 2002). Or maybe Cho Chikun or Takemiya or Ma Xiachun is stronger. There's not enough data to reach a firm conclusion.
18.104.22.168: New Technique
(2020-02-14 18:54) [#11494]
Monteo A firm conclusion can be drawn using an advanced method called Superhuman Performance Analysis Technique. I will show the full results in the book ďWho is the Strongest Go Player in History?Ē. This method is somewhat valid because it shows that AlphaGo Zero is the strongest, followed by AlphaGo Master. Ke Jie is stronger than both Honinbo Shusaku and Honinbo Dosaku.
: ((no subject))
(2020-02-15 13:25) [#11495]
Firm conclusion, really?
I dont want to sound agressive, but did you studied experience of similar project (trying to asses strenght of chess players through history using super strong chess AI (rybka 3.x) ?). It looks like no. From the data and analysis, it was drawed conclusion that strongest players were Fisher(strongest 1 year period) and Capablanca (strongest periods from 2 year to 15 year). But actually best played game was by Tal (yeah, i am looking and pointing at your conclusion that Cho Hunhyeon was little stronger than Shusaku). Also Alekhine looked worse than Capablanca, yet somehow he managed to win 34 games match against him convincingly (what is probability that better player lose match played to 6 wins? pretty low, even if difference in strenght is small).
You dont even say what AI you used (if AGZ is strongest i assume you used AGZ, but somehow its strange to believe that you have access to it).
Also Shusaku and KeJie played completely different games. Hard to compare them, and using AI to jugde Shusaku game is even harder. Judging Dosaku is even more harder from obvious reasons.
(2020-02-15 17:02) [#11496]
Monteo I donít want to talk much. You can send me two game records from a downloadable Go database with hidden player names and ranks, I can figure out who is the stronger player between Black in game 1 and Black in game 2, without prior knowledge of their strength except their moves. If I tell you correctly, you must give me $30 at least (although actually I should charge $500) for wasting my time.
: Re: Reply.
(2020-02-15 17:19) [#11497]
To make that bet i would need to know who was stronger in advance. I dont have such a knowledge.
Anyway, it is Sensei Library, not "Free Place To Advertise For Anybody Even If He Is Very Strong Player". If you would explain your method in details (also a bot you claim to say is stronger than AGZ), it could be accepted here. Now it looks like claims of some at least "extravagant freak". You make very strong claims, but you dont back it with any strong proofs or arguments. Sorry, but its not a place for it. Your method and bot might be very good (though when you said your claims is backed by just 1 game make suspicions that your knowledge of scientific method is not sufficient*), but you dont make it in way easy for other ppl to believe you.
- I could pick very poor game of Cho Hunhyun and claim he was at most weak pro (he played more than 2000 games, at least in few he must have blundered heavily, right?). Cherrypicking never nice. You didnt even said which game of Shusaku you analyzed. I dont know your method much, but i have strong suspicious that if you would choose Shusaku first Castle Game against Sakaguchi Sentoku ( http://ps.waltheri.net/database/game/17242/), you would get Shusaku stronger than from game you chose.