gowan wrote:
Forgot to post the SGF?
Thanks.
Working on diagrams. I thought I had reserved a post for them. I'll do that now.
| Life In 19x19 http://www.lifein19x19.com/ |
|
| A nearly perfect opening http://www.lifein19x19.com/viewtopic.php?f=15&t=17055 |
Page 1 of 1 |
| Author: | Bill Spight [ Thu Nov 14, 2019 9:53 am ] |
| Post subject: | A nearly perfect opening |
Theses days, pros play nearly perfect openings, because they have gone to school on today's top bots. However, you don't expect ancient pro games to have nearly perfect openings, according to today's bots, for a couple of reasons. First, humans have improved over time, even withou the help of AI. Second, many plays that were considered natural or good, the bots now tell us are mistakes. Third, not having komi is a significant difference that affects strategy. In no komi games White often made inferior plays in order to complicate the game for Black. But recently, while browsing the Elf GoGoD commentaries, I have run across a couple of pro games with openings containing only a few minor errors, according to Elf. Here is one between Shusai Meijin Honinbo and Segoe Kensaku, who was 5 dan at the time. |
|
| Author: | gowan [ Thu Nov 14, 2019 10:10 am ] |
| Post subject: | Re: A nearly perfect opening |
Forgot to post the SGF? |
|
| Author: | Bill Spight [ Thu Nov 14, 2019 10:19 am ] |
| Post subject: | Re: A nearly perfect opening |
gowan wrote: Forgot to post the SGF? Thanks. Working on diagrams. I thought I had reserved a post for them. I'll do that now. |
|
| Author: | Bill Spight [ Thu Nov 14, 2019 10:20 am ] |
| Post subject: | Re: A nearly perfect opening |
The game took place on December 3, 1917. It was the second game of a ten game match. Based on the rank difference, Shusai was the favorite to win the game. Edit: Qualification changed in response to John Fairbairn's comment below.  was the only play that matched Elf's first choice. The other plays up through  were within what I think is Elf's margin of error, less than 4½% worse than Elf's top choice.  was a good direction of play move.  is a minor error, losing 7% to a, according to Elf. Indeed, one wonders why Shusai did not make such a dual purpose move. jumps into the corner.  and  are Elf's first choices. Elf regards  as a minor error, losing 7% to the AI joseki play at a. Variation for is in next diagram.  ,  and  are Elf's top choices. Elf regards  as a minor error, losing 5½% to 20, which aims to take the corner in exchange for outside strength.Elf's variation for After ,  approaches the top right corner. Then Black reduces White's moyo in sente and switches to the top side.All moves except  and  are Elf's top choices. Elf thinks that loses 7% to the tesuji at 26. Then play would continue.  - ,  -  , Ba - Wb. ,  ,  ,  and  are all Elf's top choices. All moves except  are Elf's top choices. looks natural, but Elf thinks that it loses 5% to a.Elf's main variation for By  the point of is clear, to attack the floating Black group. and  attack strongly, then White switches to bolster his weak stones in the bottom left corner, then White comes back to cut with  .BTW, Segoe won by resignation. Edit: Added a diagram for the next four moves, which were all Elf's top choices. Seeing 11 moves straight that were all Elf's top choices in a classical game was quite something. |
|
| Author: | John Fairbairn [ Thu Nov 14, 2019 11:40 am ] |
| Post subject: | Re: A nearly perfect opening |
Bill, with your approval I will, in step with you, add comments from two human commentaries I can instantly put my hands on. One is by Segoe himself but is very brief. The other is by your perennial favourite Takagawa. Moves 1-10 Segoe did not comment on any move before 35, so I think we can infer that he was satisfied with the opening. Takagawa commented: "For White 10 I would want to combine an extension with a pincer at 34 on the lower side [your 'a']." Moves 11-20 Takagawa: "If 12 blocks at 13, play will proceed as per joseki and the location of 10 will be found wanting." Background Segoe had been 5-dan for just over a year, and this match was on B-2 handicap. This game was Game 2 of a scheduled 10-game match sponsored by the father of one of Segoe's pupils, Yoneyama Toku, a 4-dan. Segoe won the first game (no longer extant), and the match was called off after this second game because of what he described as "certain circumstances after this game." Not sure it's fair to say Shusai was favourite BTW. Shusai had to wait until 1925 before he ever beat Segoe, and clocked up a good dozen failures to crack Segoe's defences. The early games were handicap games to be sure, but Segoe had defeated Shusai on just Black just a few months before this match, after a mere 89 moves. |
|
| Author: | Bill Spight [ Thu Nov 14, 2019 12:11 pm ] |
| Post subject: | Re: A nearly perfect opening |
John Fairbairn wrote: Takagawa commented: "For White 10 I would want to combine an extension with a pincer at 34 on the lower side [your 'a']." Moves 11-20 Takagawa: "If 12 blocks at 13, play will proceed as per joseki and the location of 10 will be found wanting." Many thanks, John. Takagawa and Elf agree about . As for  , Elf prefers the block at 13, but only by 2%, certainly within the margin of error. The bots don't care much, if at all, about our human sense of direction of play.  At least this early in the game.
|
|
| Author: | Knotwilg [ Thu Nov 14, 2019 2:04 pm ] |
| Post subject: | Re: A nearly perfect opening |
A great find! (It didn't disturb the reading but you may want to correctly assign (not  ) and (not ) in the above)
|
|
| Author: | John Fairbairn [ Fri Nov 15, 2019 2:27 am ] |
| Post subject: | Re: A nearly perfect opening |
More comments from the humans: Moves 21-30 Takagawa: A quieter way to play instead of White 24 would be the stretch to 27, Black 26, and then something like White 83 [r13]. The choice of how to play in the game is unreasonable. Moves 31-40 Segoe: With Black 35 and 37 and so on, both sides get immersed in a fighting game of power go. Takagawa: Instead of the attachment at 35, invasion at 37 is also good. If White 38 cuts at 46, then he will lose the capturing race after Black 48, White 56 [f5], Black 47, White 49, Black 10-15 [k5], White 9-15 [j5], Black 9-14 [j6], White 8-15 [h5], Black 50. Moves 41-50 and Moves 51-54 None. Bill: There was just one more comment from Segoe, for move 202 ("Even without the oversight of White 202, there was no way to budge Black's victory." There are several more comments from Takagawa, but nothing more to do with the opening. |
|
| Author: | Bill Spight [ Fri Nov 15, 2019 8:33 am ] |
| Post subject: | Re: A nearly perfect opening |
John Fairbairn wrote: More comments from the humans: Moves 21-30 Takagawa: A quieter way to play instead of White 24 would be the stretch to 27, Black 26, and then something like White 83 [r13]. The choice of how to play in the game is unreasonable. Yes, at 27 does fit Takagawa's style, doesn't it? Interesting that he recommended approaching the top right corner at R-13 instead of Elf's play, the third line attachment at P-17. That attachment is one of the first things I noticed about pro games in the 19th century when I was a DDK. Interesting that the bots seem to like it.
|
|
| Author: | kvasir [ Sat Nov 16, 2019 12:22 am ] |
| Post subject: | Re: A nearly perfect opening |
Why nearly perfect? I'm not sure this game is any different from other no komi games from that period. Guess white just wants to survive the opening without letting black become thick and take enough points to coast to the endgame. was awful and Leela#249 wants to punish at move 33 by leaning the other way than what happened in this game. 1-7 in the diagram below seem brilliant.Basically white goes from 57% to 41% in this opening if is Leelas move. Might one call this a slack opening for white if it was a komi game?These images might be little bit over-sized but it seems easier to post them than create diagrams  Attachment: Capture.PNG [ 1.44 MiB | Viewed 5872 times ] Attachment: Capture3.png [ 863.34 KiB | Viewed 5872 times ] |
|
| Author: | Bill Spight [ Sat Nov 16, 2019 4:11 am ] |
| Post subject: | Re: A nearly perfect opening |
kvasir wrote: was awful and Leela#249 wants to punish at move 33 by leaning the other way than what happened in this game. 1-7 in the diagram below seem brilliant.Basically white goes from 57% to 41% in this opening if is Leelas move. Might one call this a slack opening for white if it was a komi game?Interesting. We seem to have a disagreement between Elf of the commentaries and LZ#249. Elf gives a winrate estimate for White of 36.7% with 225.9k playouts. LZ#249 gives White a winrate estimate after the shoulder blow of 40.8% with 92k playouts. However, LZ#249 gives a winrate estimate for White of 47% with 1.6k playouts. is LZ#249's second most visited play. I think that the 6% difference means that it thinks that is a mistake. The shoulder blow does not show up in Elf's commentary, so Elf gave it fewer than 1.5k playouts. Elf shows four alternative plays that got more than 1.5k playouts, each with a winrate estimate within 2% of that of .Given the playout disparity between the shoulder blow and , 92k vs. 1.6k, LZ#249's winrate estimate for is much less reliable than its estimate for the shoulder hit. I suppose that you have a readout after has been played. What is White's winrate estimate then? You inherit the winrate estimate from LZ#249's top choice for  . That will give us a better comparison. Thanks. Edit: The White winrate estimate for Elf's top choice for is 36.8% with 248.6k playouts, within 0.1% of the winrate estimate for . Basically no change.
|
|
| Page 1 of 1 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|