Life In 19x19 http://www.lifein19x19.com/ |
|
#267 rlaalswo v. fevobod152 http://www.lifein19x19.com/viewtopic.php?f=37&t=17149 |
Page 6 of 6 |
Author: | fevobod152 [ Tue Dec 24, 2019 6:16 am ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | Allen [ Tue Dec 24, 2019 7:36 am ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
for observers |
Author: | dfan [ Tue Dec 24, 2019 9:11 am ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Allen wrote: for observers |
Author: | rlaalswo [ Tue Dec 24, 2019 9:19 am ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | fevobod152 [ Tue Dec 24, 2019 11:08 am ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | rlaalswo [ Tue Dec 24, 2019 1:28 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | fevobod152 [ Tue Dec 24, 2019 3:58 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | rlaalswo [ Tue Dec 24, 2019 6:53 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | fevobod152 [ Tue Dec 24, 2019 9:01 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | rlaalswo [ Tue Dec 24, 2019 9:05 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | fevobod152 [ Tue Dec 24, 2019 9:07 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Author: | rlaalswo [ Tue Dec 24, 2019 10:34 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
I resign. Thank you for the game. |
Author: | fevobod152 [ Tue Dec 24, 2019 10:46 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
Thank you for the game, and have a Happy New Year! |
Author: | TelegraphGo [ Sun Dec 29, 2019 4:42 pm ] |
Post subject: | Re: #267 rlaalswo v. fevobod152 |
I hope it's not too rude to post my interpretation of an AI review on this game, because I'm doing it! First, let me congratulate both players on a VERY well played game. On my first pass through of this game with my low-playouts slightly old LZ version, there was about as much +% change as -% change from move to move. That means that you guys played at a level very close to AI's 'instinct' level Here's a picture of the evaluation graph after more playouts: https://imgur.com/a/tfs9zuc _________________________________________ Black took a slight lead in the opening, apparently. It seems to me that AI doesn't particularly like the probe that fevebod152 played in the lower right. After that, it seems like it wanted white to take more influential variations in the top left, to harmonize with white's very centrally-oriented bottom side. One blue move that surprised me: Both the players and I thought that W couldn't possibly afford to take gote in this totally unimportant area, and were impressed by fevebod's grasp at sente. LZ thought the taking this good shape was worth the full move - rating it very slightly better than fevebod's move. The idea appears to be taking the left side in security, and claiming the middle will end up as an investment in dame. Here's a sample variation I picked: _________________________________________ The drop in black's winrate near the middle was well explained by fevebod in his post, suspecting a poor move order. He was completely correct. Also, rlaalswo made a second mistake later in the variation, on one of seemingly forced moves. I show an easy-to-evaluate variation suggested by LZ: Black can now capture - white will answer, since black's ignoring any ko threat would be sente. Compared to the game, we have the same shape but for the ko aji of disconnection and the attachment to white's stone. Black will continue with a wedge at H14, which is clearly more powerful than the move black played in game at K16. _________________________________________ I was horribly wrong about the move that fevebod played and I complained about - LZ says it's the only move and every other move loses >5%!! I took about an hour to convince myself of this, and to be honest don't have the energy to test my communication skills by relaying my new understanding of the shape in text. I do recommend studying this move in particular by yourself, though - I learned something tangible. If you wrongly agreed with me before, you'd do yourself a favor to put the position on a real board and do some deep reading until you're in fevebod's camp. My criticism of rlaalswo's deep invasion was not so clearly wrong, but just slightly silly - both moves are ~71% winrate for white. When later I asked him whether he considered the cut, I was more curious about his decision making process than suggesting it as a good move. His verification that the cut was inaccurate was correct. _________________________________________ Black hits his final peak winrate when he cuts. My AI was misreading this continuously - on 1-5k playouts it made a mistake every three moves or so. Analyzing became very tedious, and my conclusion may be inaccurate. That said, I suspect that with perfect play every complicated variation goes up in flames for black without hope of compensation. There doesn't seem to be a way to avoid a complicated position here, so this 62.5% position is just 100% winning for white - there's not even a way to find a 2 point loss as a compromise. So the downward trend that black fell through after that? That same thing would have happened if this was an AI v. AI game, though perhaps it would have been a slightly shallower slope. There's still something I want to show off, a tesuji that I missed, my AI initially missed, and that the players might not have given due consideration. _________________________________________ So overall, this was a fantastic game, with the only clear mistakes being the pair of minor move order mistakes by rlaalswo around move 50. I would like to forcefully upgrade fevobod152 to high-dan, if not already secretly a professional. That was simply a flawless game by white, and near flawless by black. |
Page 6 of 6 | All times are UTC - 8 hours [ DST ] |
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |