It is currently Thu Aug 24, 2017 1:43 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 51 posts ]  Go to page Previous  1, 2, 3
Author Message
Offline
 Post subject: Re: DH records
Post #41 Posted: Sun May 14, 2017 6:17 am 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
thanks for this Bill, it took me a bit of time, but I see what you mean now. And probably diagram 4 is never better.

The language I use for endgame does seem to confuse myself. It comes from pro commentary estimates, and my own understanding with a rough overview of the theory, whereas your system is clearer, comparing everything to a given zero point.


dhu163 wrote:
I can agree that in general your 3rd diagram is marginally better for white than the 4th, but by 1/6. I think you mean that after white a, then c is worth 2 2/3, so therefore a is worth 3 points with a 2 2/3 gote follow up, and therefore is worth 4 1/3 (instead of the 5 I first wrote down). The 3rd diagram is equivalent to if W gets a and B gets c, so this is better by (3- 2 2/3)/2 = 1/6.


There is a lot of rubbish in what I wrote here. It should be:

after white a, then c is worth 4 2/3, so therefore a is worth 3 points with a 4 2/3 gote follow up, and therefore is worth 3+ (4 2/3) /2 = 5 1/3 (instead of the 5 I first wrote down). The 3rd diagram is equivalent to if W gets a and B gets c, so it is better by ((4 2/3 )/2- 3)/2 = 1/3.

Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #42 Posted: Sun May 14, 2017 6:50 am 
Judan

Posts: 6261
Liked others: 1476
Was liked: 2376
dhu163 wrote:
The language I use for endgame does seem to confuse myself. It comes from pro commentary estimates, and my own understanding with a rough overview of the theory, whereas your system is clearer, comparing everything to a given zero point.


Yes, I do think that the usual method of evaluating plays causes confusion. But this is not my system. It uses what I learned as miai counting, what O Meien today calls absolute counting ( 絶対計算 ). :)

_________________
"Drooling Banjos"

Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #43 Posted: Wed May 17, 2017 9:15 am 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
final pandanet game, and I've gone all season undefeated, playing almost all matches

this game, my play quality was low, or at least my alertness, judgement, reading were all down due to exam prep exhaustion, just relying on feeling. Also my opponent played well, making a nice plan to attack a weak group of mine (I should probably have sacrificed a large group to save that one) and perhaps it should have died. In the end, I thought I was slightly behind, but score estimate says I was ahead, I threatened a slow ko to kill my opponent's corner. Instead he threw the ko in himself, so I took the ko first for free, he played a fake ko threat in time trouble, and I won.

I made a quick commentary.



And a practical way to avoid the complacency vs Andrew Kay is probably to imagine a significant reverse komi, to keep playing good moves, and still challenge my opponent to think, to still leave a high quality game, than some conservative rubbish.




Attachments:
20170516 paxdeorum mathmo.sgf [8.49 KiB]
Downloaded 185 times
Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #44 Posted: Thu May 18, 2017 3:33 pm 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
I just had a good pro game study session. Perhaps with uni pressure and computer go, my creative spirit in go has gotten lost, and I've stopped improving.

With computer go, I've fallen for the trap of thinking learning their ideas and techniques will help my go, when I forget how much of a gap I have on the understanding of every move of high level games and every move they notice but don't play. Perhaps it is also that the high level of judgement and intuition of computer go really brings me back to my ideas of the logic of fuzzy concepts and intuition that I found so interesting before in go. But it turns out to have limited value in improving go skill. I need to put more emphasis improving my personal understanding rather than mimicking high level moves when my understanding of lower level moves is insufficient. I've reduced my leela usage to a minimum, and I should try to think more independently of alphago.

At today's go club, I got the chance to think more deeply about pro moves on a real board, guessing the next move, and seeing how confusing they are, and how big the gap is. Normally I just skim games without thinking too much.

I've looked at the first 30 moves quite in depth and no more. The file size is bigger than my average self-commentary.

Also https://www.youtube.com/watch?v=-yhaEj5ylL0
Lee Changho plays a very cool new move against Lee Sedol in some Korean cup. I can't find the sgf on go4go yet.



Attachments:
__go4go_20170514_Shin-Jinseo_An-Dongxu.sgf [9.88 KiB]
Downloaded 167 times
Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #45 Posted: Fri May 19, 2017 5:46 am 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
not too bad a game. Looked into move 28 in depth.



Attachments:
20170519 bronson111 mathmo.sgf [7.37 KiB]
Downloaded 152 times
Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #46 Posted: Thu Jun 08, 2017 2:41 am 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
I won 6/7 at challengers, to come joint top with uberdude, who I will be playing in the final of the British Championship.

I felt I played much much better than at candidates, and by games 4,5,6,7, my opponents seemed to have no chance.
So along with nice study sessions with Andrew Simons, Tim Hunt and Andrew Kay each evening over alphago self-play, I was in a good mood the whole tournament despite losing on day 1 to Uberdude, and getting a ridiculous £80 fine on morning 2 for forgetting my discount railcard at home while using it for my oyster card on the London Underground.
Quite relaxed whether I got through or not. And daring to experiment with the crazy ideas from alphago self-play

varying amounts of comments, I'll probably add more later







Attachments:
chall2017 3.sgf [5.09 KiB]
Downloaded 104 times
chall2017 2.sgf [4.97 KiB]
Downloaded 104 times
chall2017 1.sgf [8.83 KiB]
Downloaded 109 times


Last edited by dhu163 on Thu Jun 08, 2017 4:06 am, edited 2 times in total.
Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #47 Posted: Thu Jun 08, 2017 2:47 am 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d






Attachments:
chall2017 6.sgf [8.21 KiB]
Downloaded 115 times
chall2017 5.sgf [4.48 KiB]
Downloaded 114 times
chall2017 4.sgf [16.01 KiB]
Downloaded 114 times
Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #48 Posted: Thu Jun 08, 2017 2:48 am 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d


Attachments:
chall2017 7.sgf [13.16 KiB]
Downloaded 116 times
Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #49 Posted: Sun Aug 20, 2017 1:41 pm 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
I'm very pleased to have won game 1 of the British Championship, and played well. I had gotten far too nervous beforehand, so this has calmed me down.

Slow games prove very good for me. And my self-review file is full of variations, at a record size.



Attachments:
2017 Brit champ final 1.sgf [43.99 KiB]
Downloaded 52 times


Last edited by dhu163 on Mon Aug 21, 2017 4:01 am, edited 1 time in total.

This post by dhu163 was liked by 3 people: daal, dfan, ez4u
Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #50 Posted: Sun Aug 20, 2017 3:49 pm 
Oza
User avatar

Posts: 2360
Liked others: 1165
Was liked: 1051
Good job holding onto that grain of sand! Thanks for sharing such a thorough and interesting review.

_________________
These moves are not part of a regular dan repertoire... - Knotwilg

Top
 Profile  
 
Offline
 Post subject: Re: DH records
Post #51 Posted: Wed Aug 23, 2017 3:18 am 
Lives with ko

Posts: 149
Liked others: 12
Was liked: 170
Rank: UK 2d Dec15
KGS: mathmo 4d
IGS: mathmo 4d
I don't much trust leelabot these days.

But using it points out some of the issues in my judgement, especially when emotional.

1) At the start, it preferred black in my joseki choice: perhaps move 13 is better at C17; at move 16, it doesn't like my D15 idea as it thinks the D14 cut works instead of needing to atari D18; move 20, 21, 22 could all be D15; move 35 mistake drops B's win chance from 52.6% to 38.3%

2) my whole sequence from move 52 to 62, taking gote for insufficient profit, is questionable, e.g. it preferred move 54 at D3. I had already suggested in my review move 52 at L17


btw, upon calculation, move 68 better endgame is actually to make good shape at C6

3) At move 69, B's chance jumps from 36% to 39% (it didn't notice how good this move was) at it suggests J17 for W and cut. It really doesn't like my strategy with move 70, I suppose it says I lose too much on the upper right without sufficient profit from any attack. By move 79, it likes B with K15, at 62%! But perhaps it is a bit confused, as even though move 88-94 follows its line it rapidly drops from 60% to B to 60% to W. Leelabot is bad at semeai. At move 124, it is up to 86% to W, and doesn't really change afterwards.

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 51 posts ]  Go to page Previous  1, 2, 3

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: Google [Bot] and 2 guests


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