It is currently Tue Mar 19, 2024 1:47 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 137 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7
Author Message
Offline
 Post subject: Re: Uberdude's journal
Post #121 Posted: Mon Feb 04, 2019 11:11 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
The final game with Lizzie analysis, I played from around move 140 with under a minute and a 10 second Fischer increment so it was pretty crazy! Not a smooth LZ cruise to victory, LZ said the lead swapped hands 20 times in the middlegame madness! Inviting the g13 atari mistake is based on a ladder which amusingly only just works for me.



Compare winrate graph for this game:
Attachment:
vs Peikai.PNG
vs Peikai.PNG [ 306.21 KiB | Viewed 16687 times ]


with previous one!
Attachment:
vs Alastair.PNG
vs Alastair.PNG [ 209.95 KiB | Viewed 16687 times ]


This post by Uberdude was liked by: Elom
Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #122 Posted: Tue Feb 05, 2019 4:45 pm 
Lives with ko
User avatar

Posts: 205
Liked others: 49
Was liked: 36
Rank: EGF 2k
KGS: MKyle
Quote:
First game was against this forum's MikeKyle, maybe he will upload our Lizzie review.


Sure.

I didn't save the sgf in lizzie at the time but I think i got the order right later on.

For anyone missing the context, this is a huge miss-match (even game 4d v 2k (egf)). Always great to get a game against a stronger player, and grateful for the review too - Bots are fun, but humans can explain using concepts, not just sequences. If anyone has other comments, I'd appreciate that.



Attachment:
frodshamr12019-02-02elf.sgf [13.72 KiB]
Downloaded 1811 times


This post by MikeKyle was liked by: Uberdude
Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #123 Posted: Wed May 29, 2019 4:38 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
I'm through to another British championship match vs Andrew Kay, as expected given we were the only 4ds in a field of 1k-2ds. I lost to Andrew in the first round with a big mistake in a fight and no way back. Won the rest of the games but quite a few (vs Hunt, Wall) I gave my opponent a big chance but they kindly let me recover so didn't play particularly well. Here's the LZ reviews.

White vs Kay. I didn't live with 150 but stupidly died at top, played a load more and resigned when the theoretical ko at bottom became seki.


Black vs Rix (was doing well but mishandled my weak group on left, until he bizarrely let me kill his top left corner when I played a liberty increasing probe and then couldn't kill my middle)


White vs Hunt, doing well until started ko too early. Ended up killing his middle for a resign as I allowed the right side to be a ko.


Black vs Wall or how to overattack and turn your 10 point territory into opponents 35 point territory! He resigned at 241


White vs Mitrovic (most interesting game)


Black vs Fearnley (played terrible on left side!) Played more endgame, ended up (unnecessarily) killing top right with the j15 h14 tesuji combo I'd been aiming at.


White vs Scarff (closer than I hoped, though LZ gives me 90+ from early)


Attachments:
simons scarff.sgf [17.85 KiB]
Downloaded 1586 times
simons fearnley.sgf [9.29 KiB]
Downloaded 1583 times
me vs boris.sgf [17.77 KiB]
Downloaded 1575 times
simons wall.sgf [32 KiB]
Downloaded 1596 times
simons hunt.sgf [14.35 KiB]
Downloaded 1659 times
simons rix.sgf [17.21 KiB]
Downloaded 1652 times
simons kay.sgf [22.58 KiB]
Downloaded 1629 times

This post by Uberdude was liked by 2 people: Bill Spight, dfan
Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #124 Posted: Wed May 29, 2019 6:07 am 
Gosei

Posts: 1589
Liked others: 886
Was liked: 527
Rank: AGA 3k Fox 3d
GD Posts: 61
KGS: dfan
Uberdude wrote:
I'm through to another British championship match vs Andrew Kay, as expected given we were the only 4ds in a field of 1k-2ds.

Congrats! But what happened to all the very strong players? Are there no 7d-level players around or do they consider this tournament not interesting? (I have seen that happen locally but it is a bit more surprising at a national level.) I imagine that there must be plenty of strong players around who grew up in the CJK countries; am I mistaken (I might be extrapolating too much from US demographics), or are they not interested, or is there some other reason they don't participate? I saw no Asian names in the crosstable. The top of the US rating list is dominated by people with Asian names: see https://www.usgo.org/search-ratings.

I remember you complaining about this in the past, but I don't remember if there were more details.

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #125 Posted: Wed May 29, 2019 6:43 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
dfan wrote:
But what happened to all the very strong players? Are there no 7d-level players around or do they consider this tournament not interesting? I imagine that there must be plenty of strong players around who grew up in the CJK countries; am I mistaken (I might be extrapolating too much from US demographics), or are they not interested, or is there some other reason they don't participate? I saw no Asian names in the crosstable. The top of the US rating list is dominated by people with Asian names.

First of all this is the British championship (not Open), so there is a requirement to be a British citizen or have lived here 5 years. So that excludes the various strong (usually ~5d) Chinese students who tend to study at university for just a few years. A few years ago we did have one, Junnan Jiang 5d, who had lived here long enough for his PhD to compete and win (I patted the racoon belly against him in the Challenger's but lost later!). Whether many of them would play if they were allowed I'm not sure, the timing is bad with university exam season, beating up us native weakies might not be so interesting, and it's a big time commitment. There is talk of changing the format given the declining entry.

There are no eligible 7d-level players (I mean EGF 7d, AGA 7d is about EGF 5d). Matthew Macfadyen 6d has dominated British go since the 70s but he is not so strong now and has retired from competitive play. Shutai Zhang 7d was a Chinese player who lived in England in the 90s and interuppted Matthew's win streak. Bei Ge 5d beat Matthew in the 2000s. There have been the odd 7d Asians here (e.g. 7d Koreans Oh Chi Min, Kim Youngsam studied here briefly, some Chinese 7d students (e.g. Qinmeng Zhang, Quichu Lu) occasionally turn up at London Open or other tournaments. But afaik there are no 7d level long-term residents of Asian origin. There is for example Li Shen 6d, who was active as a 6d child (he won the first Cambridge tournament I played in), briefly resurfaced while a student at Cambridge but has since disappeared from the Go scene again. 2 years ago Daniel Hu 5d (who was active on L19) did win.

European Go demographics are quite different to US. If you look at the top 6d and 7d Europeans the vast majority are natives rather than Asian immigrants as in the US. So UK is quite similar to other European countries, except our top active players are 4d with some inactive 5ds instead of 6-7d. Something I have noticed is a lot of the top Europeans started as children, whereas the current generation of top Brits like me started at university so it's unsurprising we aren't as strong. Our youth Go support/activity has probably been pretty poor compared to other countries though it has improved in the last few years, but still no home-grown dan players.


This post by Uberdude was liked by 4 people: dfan, ez4u, sorin, Struttnoddy
Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #126 Posted: Sat Jun 01, 2019 12:26 pm 
Dies in gote

Posts: 50
Liked others: 12
Was liked: 9
Rank: 1kyu
KGS: struttnoddy
Congrats and all the best for the final!

Quick question, I must be missing something but I can't see the seki at the bottom of the board in your game against Andrew? It still looks like a two step ko to me

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #127 Posted: Sun Jun 02, 2019 12:20 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
Struttnoddy wrote:
Quick question, I must be missing something but I can't see the seki at the bottom of the board in your game against Andrew? It still looks like a two step ko to me

I left it as the hopeless approach ko for most of the game, started the ko in endgame, he ignored my threat for just a dozen or two points so then it was seki at best or dead but either way was still far behind so resigned.

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #128 Posted: Tue Oct 29, 2019 4:24 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
Uberdude in Oct 2016 wrote:
So my goals are:
- at least play in if not win British Championship title match next year
- get 2400 rating
- get solid 5d KGS from non-blitz


Well, I finally got to 2401, though maybe I will lose it once my loss to Andrew Kay in the first title match game is retrospectively entered into the database after the match concludes. I had a nice result at the Wessex tournament: narrow loss to a Chinese 5d after a decent lead, beat a 2d former 4d, got my first real-life tournament win against a rusty Matthew Macfadyen 6d who I've only beaten online and in a friendly game before (reasonably close game for a while in which I sacrificed some bits in a fight, but then when I was blitzing through the 10 second Fischer increment he speed up and made several blunders), and then beat another Chinese 5d after he played perhaps a bit too lazily (he never attacked my AI-style psuedo-thickness and just started the endgame early) but he let me capture some reduction stones in the middlegame and then I won by a point on the board as white.

Image


This post by Uberdude was liked by 2 people: ez4u, yakcyll
Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #129 Posted: Tue Oct 29, 2019 4:37 am 
Lives in gote

Posts: 575
Location: Adelaide, South Australia
Liked others: 207
Was liked: 262
Rank: Australian 2 dan
GD Posts: 200
Uberdude wrote:
Well, I finally got to 2401, though maybe I will lose it once my loss to Andrew Kay in the first title match game is retrospectively entered into the database

And then maybe you'll get it back again after the second and third games :-) By the way, is there a date for game 2 yet?

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #130 Posted: Tue Oct 29, 2019 5:01 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
xela wrote:
Uberdude wrote:
Well, I finally got to 2401, though maybe I will lose it once my loss to Andrew Kay in the first title match game is retrospectively entered into the database

And then maybe you'll get it back again after the second and third games :-) By the way, is there a date for game 2 yet?


Yeah, not feeling so hopeless after those practice games :) . Next game is Saturday 9th November.

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #131 Posted: Tue Oct 29, 2019 8:38 am 
Honinbo

Posts: 9545
Liked others: 1600
Was liked: 1711
KGS: Kirby
Tygem: 커비라고해
Good luck on the upcoming game!

_________________
be immersed

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #132 Posted: Sun Nov 10, 2019 12:50 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
I won the second game to set up a decider. :) Happier with how I played this game: no stupid blunders, just one big misjudgement born out of too much fighting spirit trying to get what I thought was a good exchange before connecting (actually there was another later but he didn't punish) leading to the ko (but I managed to trick him with g18 hane to get back to even) and then an oversight in endgame from a bad exchange that let him save a bunch of dead stones, but luckily it wss gote and I generally played a better endgame so won by 8.5.



This post by Uberdude was liked by 2 people: Bill Spight, xela
Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #133 Posted: Sun Nov 10, 2019 3:00 pm 
Lives in gote

Posts: 575
Location: Adelaide, South Australia
Liked others: 207
Was liked: 262
Rank: Australian 2 dan
GD Posts: 200
Sorry I wasn't able to watch this one live and cheer for you, but thanks for posting it. Good luck with the final of the final!

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #134 Posted: Mon Nov 11, 2019 9:02 am 
Honinbo

Posts: 9545
Liked others: 1600
Was liked: 1711
KGS: Kirby
Tygem: 커비라고해
Seems they should rename the British Championship to the Andrew Title. You guys enter the finals pretty often!

_________________
be immersed

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #135 Posted: Mon Nov 11, 2019 9:44 am 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
In the last 8 years (since Matthew Macfadyen retired) of the 16 finalists the names have been distributed as follows:

Andrew: 10 (5 each for Kay and Simons, 3 times vs each other)
Sam: 1
Daniel: 1
Charles: 1
Junnan: 1
Alex: 1
Nick: 1

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #136 Posted: Mon Nov 11, 2019 9:48 am 
Honinbo

Posts: 9545
Liked others: 1600
Was liked: 1711
KGS: Kirby
Tygem: 커비라고해
Uberdude wrote:
In the last 8 years (since Matthew Macfadyen retired) of the 16 finalists the names have been distributed as follows:

Andrew: 10 (5 each for Kay and Simons, 3 times vs each other)
Sam: 1
Daniel: 1
Charles: 1
Junnan: 1
Alex: 1
Nick: 1


That's fascinating. Do you have insight into what Matthew Macfadyen did that set him apart? Are the Andrews following a similar path?

_________________
be immersed

Top
 Profile  
 
Offline
 Post subject: Re: Uberdude's journal
Post #137 Posted: Mon Nov 11, 2019 1:56 pm 
Judan

Posts: 6725
Location: Cambridge, UK
Liked others: 436
Was liked: 3719
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
Commentary/review of the game, a combination of my thoughts at the time plus LZ/KataGo/Elf review:

KataGo graph of winrate (green) and point (purple), white (me) towards bottom. It pretty much matches my feelings during the game:
- nice little lead in the opening,
- oh crap why am I such a dummy to make this ko",
- "phew he fell for my trick, game ok again"
- attack for profit
- "I'm ahead so long as I cut this moyo down to size in sente"
- "la la leading in the endgame happy times"
- "oh detritus that peep meant he can save his dead stones, did I just blunder this game?"
- gosh did I get away with that thanks to better endgame.

Though I didn't realise how risky my timesujis with the ko in endgame were if he resisted.

Attachment:
2019 g2 KG.PNG
2019 g2 KG.PNG [ 169.55 KiB | Viewed 14964 times ]


- Parallel 4-4s are to be expected between us when I'm white: AK has only played 4-4s for years, even before bots I've like double 4-4s as white (play a speedy opening and make komi count), I used to do one or two 3-4s as black but post-AI often do double 4-4 as black which is something I never used to do.
- Move 6 counter approach is standard LZ idea, it's a bit mirror-go-esque in that white plays symmetrically which tries to make komi valuable. Matthew commented that the classical theory was that this is good for black, presumably because he has sente to play around the side as a move which combines pincer and extension from his approach, but the bot's distaste of side moves over corner moves (it's easier for them to become inefficient) probably means that double-purpose move is not as good as we thought.
- Move 7 q9 2-space high is AK's favourite and I had planned for it with a little bot research (LZ #250 says -2%), so we played to move 24 quickly.
- Move 8 3-3 is LZ's favourite and simple and what I had prepared, interestingly KataGo prefers double approach quite a bit. It thinks black's best move is to ignore the fact his corner is about to be surrounded and kiai double approach my top right corner (NB I use left/right per the game record, in reality this black right side was the bottom side from my perspective), but I would expect AK to q6 attach and although KG thinks it's fine for white to play the joseki with the q7 p6 exchange (because white gets sente, very important) I don't like making black thick like that and giving him the m3 aim.
- Although I take the corner in gote, black's r5 connection is bad shape: with either the 1-space low or 3-space high pincer the modern AI style is to s6 crawl not r5 connect, and in the latter then spend another move at r7, but such a shape is inefficient with the 2-space high. So black's side is far from territory and can even be attacked when I activate r6, which bots like to do early. Of course AK knows about r6 aji and that the side isn't territory, but when white contests that it leads to fighting which he likes. Many of our games follow this pattern.
- Move 15 r5, LZ actually thinks this is a slight minus and better to double approach directly, KataGo says loses 0.2 of a point, which is peanuts given the massive mistakes later. Still it's interesting to broaden one's mind that such a move could be aji keshi, because the exchange loses black potential on the lower side, and perhaps with the double approach fighting at the top right black might end up wishing to not make territory on the right side.
- Move 18, which attachment to play is not so clear but the LZ version I had studied with liked this one, as does Elf v2 and KataGo. But o16 is also possible, and even more recent LZ 250 actually prefers it after a while. But using my preparation my plan was to play this quickly and save time for more important things later.
- Move 24 o16 attach was as far as my preparation went. My understanding of the criteria for this move being better than q17 connect is when q17 doesn't make a good miai of attacking whichever black group he doesn't defend, and here q9 makes it harder to fight on that side. It also works well when white's potential on the top side is more than black's on the right. Here my plan is to allow black the s17 profit, because I will use r6 later to live on the side and trash those points. It seemed to surprise AK a bit and he didn't play the usual bot joseki of q15 next.
- Move 25 n16 hane the first move that took a while, and the biggest mistake of the game so far (LZ 250 -3%) so I am happy with how my plan is going: a slight lead and ahead on the clock. It might look painful for white to get the atari and empty triangle, but white gets a strong group and bots value settled groups.
- 30 capture is honte, old me might have worried it was passive as black gets some nice outside kikashi and tried something overplayish like L16 pincer but from bot study I know this result is good for white. I saw Matthew criticised my result here as overconcentrated, but bots and I (and VIXX KGS 8d in kibitz) disagree: not having a running group is very valuable, as is moves like black p13 being totally gote as it means I can trash the right side from r6 with freedom. AK was probably happy as black too, although LZ 250 gives white already up to 68% and KG white a 4 point lead (but starts with 1 point lead with the 7.5 komi, though it doesn't understand white pass stone last BGA rules) we aren't strong enough to preserve such slim margins.
- 34 s6 is a lesson from bots, they like to activate this stone early (e.g. https://lifein19x19.com/viewtopic.php?p=238695#p238695) even if they sometime then tenuki. I think the point is later black may resist more strongly if he has support, so it's an early probe. LZ agrees it, p6 and f17 are all excellent moves so that's nice. KG prefers to answer at the top right but LZ is stronger so I'll believe it. When black blocked on top (LZ would t6 hane under, 2% better) I treated it as a probe of black saying he wanted to play thickly for the centre and thus my choice of...
- 36 d10 pincer as a counter measure to q6's declared intention of centre development. If I simply backed off at f17 then I was sure AK would play d10 and even though bots are happy to take more points at c3 I was not happy to give AK the thickness for moyo he likes. Elf does sometimes like this pincer, but here it doesn't (-4%), still as I am not as good as a bot at reducing, and am playing AK who likes moyos I remain content with my decision. The only regret is I do now appreciate more how f17 simply makes the m17 kikashi unlikely to become useful. Bots also like to continue locally at q7, but I felt no rush as I didn't fear black spending a move there anytime soon.
- 37 f17 double approach as expected because just 3-3 is too kind to me. We both played correct choice of bot joseki to c16.
- 45 d12 could be c12, AK style to be more fighty and separating me.
- 46 was a key moment. I had several ideas and suspected that the d14 attach I went for was not the best, but more manageable. First though was f11 knight move and this is LZ's intuition too but this leads to complex fighting when black tries to break the line and I use g17 weakness. I also considered c6 approach: if black kindly backs off at f3 then that's almost surely a good exchange, but then I wasn't so sure because if I continue with f11 then black might live with c10 attach, which before would have been bad to make white stronger with the d9 connection, but now makes c6 inefficient and the wrong side. Also black f3 would surely be better as tenuki like f12 jump and that would be AK's style. But LZ says it was better than d14 anyway. The bots recommend an interesting move that is sometimes on my radar but wasn't here which is a good lesson: c4 attach, if you want to c6 approach direction but worry it can become overconcentrated then go closer to the corner! If they hane inside you get what you want, and hane outside is locally a nice enough result for you and then you can deal with the outside potential you give them later. The downside of the attach is it makes black stronger when he answers as he did (I did worry a bit about tenuki but decided it was find for me), and that I can't easily get the b14 attach next because he can resist at e14 before answering.
- 48 Matthew criticised me thinking a long time here rather than 2 moves earlier as that was the strategic direction decision point, and fair enough. I had been planning to high pincer at j16 to make sense of d14 defending my cut (at the cost of strengthening black) but wasn't so happy with all the possible variations (e.g. h16 bump if I extend up then black can settle on the side without too much trouble and the m17 kikashi annoy, LZ confirms that j17 extend down was indeed better for me) so then went for the low j17 for profit whilst attacking and also takes care of the m17 stones. But it was my worse move of the game so far (LZ250 -8%, KG -2 points), but much worse soon to come! The sequence went as expected to when I end at l17, which nicely makes n18 a Belgian ko so settles any worries for that group too e,g black r18 is now totally gote so I am freer to play on right side.
- 49 should g15. All 3 bots start off liking the j16 attach of the game, so if super-human bots' intuition needs thousands of playouts to realise push up is better (it is the vital point later) it's not so bad for a 4d even though AK did say he regretted it after and in hindsight wanted to h17 kick, but that's worse. KG suggests pole connection at h15 for black 51.
- Move 55 was a critical turning point. I had expected black g15 honte was proper, to which I planned to b14, because now if black resists with e14 I will b13 because the f15 stones are light after black spend a move at g15, so black has to submit at b13 and I happily get the c15 kikashi which makes black b15 gote and then run out and we fight. LZ confirms my analysis and expected sequences: black's good moves are g15, or k15 m16 exchange first then g15, or b15 (and then I get g15 sente shape point then run out). So when black played the e14 atari first my reasoning was if I connect at d15 then I have lost the option to tenuki it later when I b14 as black's already got e14 in sente: he's made me commit to saving the f15 stones. So after I connect he can g15 key shape point in sente not gote now as I need to poke out at f14 (yes a broken shape for black, but g15 in sente is worth it) and then he can treat e14 lightly, and my b14 c15 combo has become bad. LZ confirms this sequence analysis was correct; however, this result (LZ 30%) is quite a bit worse for black than g15 directly and my b14 c15 combo (38%) so this was a misjudgement on my part that I felt I should not allow black to get the e14 forcing move AND g15. So then I thought about can I get g15 in sente first, if black connects then I d15 and it's a nice exchange for me to stop his honte there next. Of course I considered his resistance like what happened in the game and I read he couldn't net my g14 empty triangle and the corner could become a ko, but I forgot that he would have plenty of local ko threats vs g15 and the ko was direct so terrible for me. Also I had imagined him playing d19 not d18 connection and not realised what a difference it makes to his liberties.
- 56: So my g15 was a massive mistake, LZ -47%, KG -32% and -13 points, way bigger than all the mistakes so far. This is a weakness of mine in that I see if I follow my opponent's plan he gets some nice exchange so I try to resist and make one of my own, but it can lead to a fight that's way worse. There's another example later. AK of course went for the fighting trade. And even more annoyingly after the game he said he wasn't going to play the good move (as LZ agreed with me) of g15 honte like I wanted to avoid but was going to do some weird loose move at h13 in which case I could have got g15 sente anyway later if I wanted, or even better said it was aji keshi and keep the threat of pushing through with ideas like g13 or f13 (LZ/KG choice).
- 58 Once I'm on the train I don't get off, but the bots say I could just admit my mistake and q7 or b15.
- 72 My idea back at g15 had been that ko is ok because if I win it I kill all his stones too, but now I realised he has too many local threats. Bots say b17 now is better to die with some liberties, but at least direct ko gave AK chances to screw up too, which he did. I was feeling pretty silly to throw away my carefully accrued small lead with this craziness now. KataGo says black leads 18 points now, compared to 4 points for me before; I knew it was bad, but not that many multiples reversed!
- 77 k15 push. But thankfully AK seems to believe my bluff that winning the ko is somehow small and greedily goes for more on the outside: if he just plays h13 as a clear local threat, or maybe j13 even better (LZ agrees both good 92%), then whatever I do: resolve ko and sacrifice cutting stones, or vice versa is clearly very good for him.
- 78 If I naturally hane at L15 then the h13 net works. There's some ladders which don't work for me towards q9 that LZ struggles to read. KG thinks L15, let black net the cutting stones and win the ko is best of a bad job, and maybe it does lead to the least worst points outcome, but LZ with it's win% objective function thinks the same, but for a human I disagree: I have to save the stones and keep fighting and encourage my opponent to make mistakes. And miraculously it worked!
- 81 m18: I thought AK might make this mistake out of greed: he should just take the ko, I L15 cut, he captures corner to resolve ko, I L14 extend to complete capture of 4 cutting stones, and black gets sente to c6 or r7 and is 20 points ahead. He was probably tempted by the thought of m18 launching an attack on mu upper right group as well as the top side fighting, but that group is resilient: it can live submissively at r18, live chunkily in ko at s18, and still has centre access. If after the m16 connection I played L14 (what bots want at low playouts because they fail to see black's tesuji as quickly as I did!) and black took the ko then if m18 is dead then compared to before black just played 2 dead stones at m18 and m16 compared to taking and winning the ko directly, a huge mochikomi and loss of sente. KG says m18 lost 10 points so I'm half way back to even.
- 83 m16 is another chance for AK to take the ko, but given he played m18 that's unlikely because I'm pretty sure he's looking at the h18 shortage of libs tesuji.
- 84 is another crucial moment, and gratifyingly I played better than KataGo would with 25000 playouts. L14 looks natural and is what KG wants and believe leads to a black 8 point lead, expecting black to take the ko, white k14 and black resolve ko. But black has a better move: h18 hane and white can't block at j18 or j19 hane is a classic tesuji that squeezes out the white libs (28 points behind) so white has to compromise with k14 and allow black to connect under IN SENTE and black is 16 points ahead, a very bad misread for a bot as strong as KataGo. LZ 250 on the other hand sees the problem with L14 and prefers g18 in under 1000 playouts. Against g18 I thought black's best moves were either 1) to take the ko, I L14, he resolves ko, and then I could n14 honte or p13 turn to try to gobble up the m18 stones as mochikomi. LZ and KG agree with this and put black 10 points ahead. Or 2) L14 capture to stop m18 being stupid, I win the ko and try to kill d18 but actually that's not a clean kill but an approach ko, and black can use m18 to attack my upper right, KataGo says that's considerably better at a 17 point black lead. And 3) I dreamt that black would allow my trick and block at f18...
- 85 which he kindly did. I don't really understand why seeing as I thought he'd seen the h18 j19 tesuji and allowing the g18-f18 exchange meant I fixed that in sente. LZ says a -40% mistake almost as bad as my g15 and I'm now back to 50%. Phew!
- 87-90 are also bad exchanges given that he's going to take and resolve the ko, because not only do they lose ko threats they destroy the option of m18 still connecting up on the first line with the L18 turn. Now bots say white is leading and I was feeling much better in the game too after the debacle of g15 and the ko.
- 92 Had been my planned ko threat ever since the ko started (but then stopped). LZ agrees with it, KG prefers the thick p13 turn which helps take care of black running out with the m14 attachment, a move I hadn't read much particularly but knew I had bad aji there. And I correctly thought that AK would answer the fighting in this area so I would have sente to patch up the top later if I wanted. There is a huge difference between the position after white p6 following black's m18 mistake, and if black had simply taken the ko back then:

Click Here To Show Diagram Code
[go]$$Bc What black should have got (black to play)
$$ ---------------------------------------
$$ | . O . . . . . . . . . . . . . . . . . |
$$ | X O X X . . . . . . . . . O . . . . . |
$$ | . X . X X X X O O O O X O . O . O X . |
$$ | X . . , X O X O X O X . X O O O X . . |
$$ | . X . X O O O X X X O . . X . . X . . |
$$ | O O X X X . O . . . O . . . O O X . . |
$$ | . X . X . . . O . . . . . . . X . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . X O O . |
$$ | . . . . . . . . . . . . . . . . X X . |
$$ | . . . X . . . . . , . . . . X X X O . |
$$ | . . . . . . . . . . . . . O . O O O . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


The marked stones show black got 2 (almost) dead stones, white got 2 useful stones as ko threat and follow up. It's nice to be on the giving not receiving end of a rip-off occasionally!

Click Here To Show Diagram Code
[go]$$Bc What black got (black to play)
$$ ---------------------------------------
$$ | . O . . . . X . O . . . . . . . . . . |
$$ | X O X X . X . X O . . B . O . . . . . |
$$ | . X . X X X X O O O O X O . O . O X . |
$$ | X . . , X O X O X O X B X O O O X . . |
$$ | . X . X O O O X X X O . . X . . X . . |
$$ | O O X X X . O . . . O . . . O O X . . |
$$ | . X . X . . . O . . . . . . . X . . . |
$$ | . . . X . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . O . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . X . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . W . . . |
$$ | . . . . . . . . . . . . . . W X O O . |
$$ | . . . . . . . . . . . . . . . . X X . |
$$ | . . . X . . . . . , . . . . X X X O . |
$$ | . . . . . . . . . . . . . O . O O O . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ ---------------------------------------[/go]


WIP


This post by Uberdude was liked by 3 people: Bill Spight, dfan, xela
Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 137 posts ]  Go to page Previous  1 ... 3, 4, 5, 6, 7

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group