It is currently Tue Feb 19, 2019 9:31 am

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 69 posts ]  Go to page Previous  1, 2, 3, 4
Author Message
 Post subject: Re: LZ help for position
Post #61 Posted: Fri Feb 01, 2019 1:03 am 
Pippen wrote:
How big of an disadvantage is -5%?
Ask a merchant banker. Or a climatologist. Or a deal-maker. Or a trader. Or a ball-breaker poker player. Or His Golden Royal Excellent Empire Duckness, who is all them rolled up into one.
Lady Bracknell wrote:
to lose 5% on one move may be considered unfortunate, but to lose 5% on each of 20 moves looks like carelessness.

Top
  
 
Offline
 Post subject: Re: LZ help for position
Post #62 Posted: Fri Feb 01, 2019 1:04 am 
Judan

Posts: 5573
Location: Cambridge, UK
Liked others: 313
Was liked: 2940
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
Pippen wrote:
Tryss wrote:
With LZ #197, at around 10k visits, it's roughly :

-4.5%
-5%
-7.5%
-7%



How big of an disadvantage is that?


As a human dan player you will sometimes be capable of correctly distinguishing one move as better than another when LZ sees the difference as small as 1%. But other times you won't realise you made 10% mistakes. Your blunders could be 20-30% or even more. Very roughly I'd guess you might make 30 mistakes in a game >= than 5%. Also bear in mind these % mistakes are quite a different scale to number of points loss of the human feeling of how big a mistake is: for example a 1 point mistake in an endgame you were winning by half could be -60%, an unconditional loss of 5 points in a middlegame you are winning by loads just -4%.

See also forum/viewtopic.php?p=240586#p240586 for a 2% mistake ddks shouldn't make.

Top
 Profile  
 
Offline
 Post subject: Re: LZ help for position
Post #63 Posted: Fri Feb 01, 2019 1:21 am 
Judan

Posts: 5573
Location: Cambridge, UK
Liked others: 313
Was liked: 2940
Rank: UK 4 dan
KGS: Uberdude 4d
OGS: Uberdude 7d
Seeing as I think saying a 5% mistake is your 30th worst (on % scale) move this game is a fairly intuitive way of understanding how bad it is, I quickly ran one of Pippens's games (https://lifein19x19.com/viewtopic.php?p=240856#p240856) though mylizzie which has a nice table to show mistakes over a selectable threshold, Using LZ 157 (so winrate changes generally smaller than more recent 40 block LZ) and mistakes >5% I got the following. That's 32 white mistakes >5%, my guess of 30 was not bad eh ;-) .

Attachment:
pippen mistake table.PNG
pippen mistake table.PNG [ 102.31 KiB | Viewed 248 times ]


And for comparison, here's the same thing for the best 2 humans on the planet, Ke Jie vs Park Junghwan's recent game. Not so different (though it was a longer game and it's debatable the meaning of such mistakes in late endgame where doesn't change result, also was only about 1k analysis so LZ 157 could be wrong)! Not so many >20% mistakes though.
Attachment:
ke park cctv new year mistakes.PNG
ke park cctv new year mistakes.PNG [ 76.77 KiB | Viewed 244 times ]

Top
 Profile  
 
Offline
 Post subject: Re: LZ help for position
Post #64 Posted: Fri Feb 01, 2019 10:29 am 
Lives in gote

Posts: 631
Liked others: 6
Was liked: 28
KGS: 3D
Tygem: 5D
So basically it comes down to: fuseki doesn't matter below pro-level play. :)

And if I am not mistaken then LZ giving White 46.51% and Black 53.49 after the split move K17 means (in theory) that from 1.000 games Black would win ~535 and White would win ~465, so enough chances for White still, right?

Top
 Profile  
 
 Post subject: Re: LZ help for position
Post #65 Posted: Fri Feb 01, 2019 11:08 am 
Pippen wrote:
from 1.000 games Black would win ~535 and White would win ~465, so enough chances for White still, right?
None shall Pass. It ain't over 'till the fat lady sings.

Top
  
 
Offline
 Post subject: Re: LZ help for position
Post #66 Posted: Fri Feb 01, 2019 12:24 pm 
Honinbo

Posts: 8044
Liked others: 2267
Was liked: 2826
Pippen wrote:
So basically it comes down to: fuseki doesn't matter below pro-level play. :)


Right. For amateurs only the endgame matters. :lol:

Edit: Seriously, if we take the fuseki to go up to move 50, the pros made 2 mistakes in the fuseki while the amateurs made 5. I'm not sure if I counted the total number of mistakes correctly, but I think that the number for the amateurs was 64 and the number for the pros was 55. In that case the fuseki accounted for about 4% of the pros' errors but for about 8% of the amateurs' errors.

But if you consider the number of amateur errors that are larger than any of the pro errors, in terms of win rate, the amateurs made 15 or 16 of those, if I counted correctly, and all of them were in the middle game.

_________________
There is one human race.
----------------------------------------------------

The Adkins Principle:

At some point, doesn't thinking have to go on?

— Winona Adkins

Top
 Profile  
 
 Post subject: Re: LZ help for position
Post #67 Posted: Fri Feb 01, 2019 1:22 pm 
Bill Spight wrote:
if you consider the number of amateur errors that are larger than any of the pro errors, in terms of win rate, the amateurs made 15 or 16 of those, if I counted correctly, and all of them were in the middle game.
which suggests two things:
1. Black Knight Pippen can get away with fundamentally illogical fuseki; because unlike King Arthur, her opp will blunder in the middle game and let her off;
2. If there's anywhere that separates the wheat from the chaff, it's the middle game

The second point suggests that it's the middle game that is the phase that deserves most assistance from any AI that can lend a helping hand.

Rbot...
https://www.lifein19x19.com/viewtopic.p ... 28#p240928

... is on the case viewtopic.php?p=240856#p240856

Top
  
 
Offline
 Post subject: Re: LZ help for position
Post #68 Posted: Fri Feb 01, 2019 3:27 pm 
Lives in gote

Posts: 631
Liked others: 6
Was liked: 28
KGS: 3D
Tygem: 5D
jaca wrote:
Black Knight Pippen can get away with fundamentally illogical fuseki


Ilogical? Maybe it's time for a challenge. If I got it correctly this is the worst of my split fusekis, according to LZ:

Click Here To Show Diagram Code
[go]$$B
$$ +---------------------------------------+
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . 3 . . . . . 4 . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . 1 . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . , . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . , . . . . . , . . . . . 2 . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ | . . . . . . . . . . . . . . . . . . . |
$$ +---------------------------------------+[/go]


Let's play 10 games with this setup, usual komi (Chinese 7.5, Japan 6.5), with LZ vs. LZ (with some decent playouts) and see how many times White wins. If it is more or equal to 4, you can't say this fuseki is illogical, if it's 2 or less you can (3 is kind of meh). I'd propose Uberdude as the guy who runs the competition, posting the games here. The price will be to be right. ;) Maybe that's the start of my glory as the inventor of the 21th century "Pippen-Fuseki". *dream*

Top
 Profile  
 
 Post subject: Re: LZ help for position
Post #69 Posted: Fri Feb 01, 2019 3:49 pm 
Pippen wrote:
the start of my glory as the inventor of the 21th century "Pippen-Fuseki
you are already world-famous, and we will all do everything we can to further glorify your holy name in tablets of stone.

However,

Please accept my apologies, but i'm never going to use the Pippen-Fuseki in my own games.

Isn't it obvious why it isn't a good idea to ignore common sense, which says that 2 open friendly sides are better than one, and that to make yourself the meat in a sandwich is just asking to be eaten?!

But i will munch upon my own Humble Pie if your experiment proves me wrong. As it happens, i am usually wrong, so you are in with a good chance! :)

Pippen wrote:
If it is more or equal to 4, you can't say this fuseki is illogical,
i don't care that you are trying to cheat here, claiming that a win rate of 40% is a logical proof! :o

And Now for the weather forecast... It will be hot today, as the steam from Uberdude's PC adds to all the hot air Pippen produces to set a world record for climate change acceleration and we will all go to Hell in a handbasket sooner than you thought.

This will finish off Mankind once and for all, and "good bloody riddance!" will say all the other species.

And, later, a new intelligence will evolve, that will tell its children about the Great Overheating of the past that melted all the icecaps and evaporated the entire Pacific Ocean, all caused by the Pippen-Fuseki.

Stop Press

omg, it's started already!
Attachment:
omg.png
omg.png [ 46.68 KiB | Viewed 143 times ]

Top
  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 69 posts ]  Go to page Previous  1, 2, 3, 4

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: bernds, Google [Bot], mithra and 5 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