Human versus Machine

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Harvey Williamson, bob

Uri Blass
Posts: 8009
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: Human versus Machine

Post by Uri Blass » Fri Oct 27, 2017 7:55 pm

Lyudmil Tsvetkov wrote:
Nay Lin Tun wrote:Saw the 10 preview pages and you wrote about winning against stockfish 4 with stonewall .Well, Stonewall is notoriously drawish opening (since dinosaur age )against computers, of course at least 99% good players already knew and might have tried against computer already. (similar to a player nicked name Grandfather who frequently posted his drawish games with stone wall here). I can tell the reason behind the loss of computer, it was because black did not develop queen bishop on c8. However I highly doubt that the same opening will work against current stockfish or Komodo of 2017 version. If you play stonewall in either 1. e3, d4, bd3 , f4 , Nf3,Ne5 in whatever order, black will respond with d5, nf6 bf5 or bg4 in either sequence. Black will not badly play like e6 and stop developing c8 bishop in that opening. Of course you may still get a draw 1 in 100 against modern computers and it should be according to your rating difference vs computers. ( your 2100+ FIDE rating vs 3100+ computers)
I don't know when people on this forum and elsewhere will finally learn that my objective current chess strength, especially under quiet conditions, is at least 2800?

that 2100+ estimate(actually 2200+) dates back from 2004-2006, and during that period I had a couple of 2400 performances.
I simply played too few rated games during that period, mostly at the weekends or in the evening, after a strenuous working day. Lack of concentration and fatigue play an enormous role in deciding a chess game.

in the course of 12 years, I have added too much chess strength, I am even afraid to think of how large that quantity is nowadays. Matter of fact is that in the last 5 years I have played and analysed at least 3 times more games than during my entire chess training span before that.

I don't know what are you talking me drawing 1 out of 100 against Stockfish. Out of 100, I am usually drawing at least 67.

I acknowledge you are right, though, concerning Bf5 in the commented Stonewall game. It is the much better choice, Komodo does play that move, but not Stockfish.
Just check, current Stockfish will again choose e6 instead.
I believe that if you play in tournament against humans and improve your rating and get the GM title more people are going to buy your books.

If you think that you improved significantly from the last time that you played against humans then I see no reason for you not to prove it by playing against humans(and if you are really strong then you can expect also to earn money from playing.

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Part 2

Post by Lyudmil Tsvetkov » Sat Oct 28, 2017 2:08 am

Here just reposting a brief reply of mine to a disbelieving member from a chess forum
(so that people are aware in what way a human may get an advantage against the top engines in the 4 mentioned structures):

Forum member wrote:
It does not make much sense to play king's Indian defense as the computer is especially strong in sharp positions
Playing the stonewall as white is known to be bad as it almost guarantees black a way to trade off his light squared bishops leaving white permanantly crippled and as black it is supposed to be good only under certain circumstances, which you will not get very often.
It's incredibly difficult, if not impossible to get the c6, d5, e4 pawn structure as black in the opening if white doesn't play badly and even if you do get it, you get positions similar to the Caro-Kann and French, which are both well respected and it's hard to see any fantastic advantage you obtain in getting these structures.
The 'central bind structure' is playable as white, but doesn't give much advantage and with white playing properly, it's very difficult to get as black.
However, the computer does have difficulty in playing against the king's Indian attack-like setup you mentioned before (with e4, e6, d3, d5, nc3, d4) and I think that you can get an advantage against it. But an advantage is all and I fail to see how anyone besides another engine can convert it into a win against such powerful defenders.


The king's Indian Defence involving d6-e5-f4 pawns is a closed one and far from sharp, so that is precisely
the position a human would like to get.
KID=KIA with black, so if the KIA is good, the KID is good too. One tempo is not of such a critical significance
at the current level of top engines.
Concerning the Stonewall Attack, indeed, white has fully equal, draw, at most, if black plays Bf5 early on to trade
light square bishops, but, fortunately, even current Stockfish development still prefers e6 and Bb7/a6(not Komodo though).
c6-d5-e4 is not that hard to get, both Stockfish and Komodo like a line like 1. d4 d5 2.c4 c6 3. Nc3 Nf6 4. Nf3 g6!(this is the trick,
definitely strongest continuation) 5. e3 Bg4!(again, best) 6. h3(that is how top engines play) Bg4 7. Qf3(bishop pair lacking,
but the queen is very displaced here) e6, then Bd6/g7, and at some point e6-e5 break is pushed.
de5 is rarely good, so there are excellent chances black will push e5-e4 later on, getting the abovementioned structure.
Of course, as the game is closed, engines see nothing, black will get decisive advantage only 20 moves later after a lot
of regrouping.
On the contrary, the central bind is best possilbe disposition for white at all, as 1. c4 is definitely white's best possible move.
For example, 1. c4! e5 2. Nc3 Nf6 3. e4!(g3 first, followed by Bg2 is also possible), and white gets big advantage, not sure if
winning though.
With black, you can get that for example from the Sicilian, Nimzovich-Rossolimo variation(see the game I just posted), as top engines
still prefer 1. e4 c5 2. Nf3 Nc6 3. Bb5 g6 4. Nc3? Later black plays Bg7 and e5, and the bind is there. Very simple.

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Part 2

Post by Lyudmil Tsvetkov » Sat Oct 28, 2017 2:16 am

Actually, it seems, the structures are not 4, but at least 8, as those could be mirrored on the queen side.
For example, a standard d6-e5-f4 black KID chain could be replicated by an e6-d5-c4 chain on the queen side, with white king castling long, etc.
That seems to happen very rarely in practice, though.

But those are only the structures I myself have discovered. I am pretty much certain there are many other ways of trying to handle the engines
easily, they just wait to be discovered by someone.

As said, if Stockfish does not improve by at least 400 elos in the next year or 2, too many humans will start beating it too easily. :D

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Human versus Machine

Post by Lyudmil Tsvetkov » Sat Oct 28, 2017 2:34 am

Uri Blass wrote:
Lyudmil Tsvetkov wrote:
Nay Lin Tun wrote:Saw the 10 preview pages and you wrote about winning against stockfish 4 with stonewall .Well, Stonewall is notoriously drawish opening (since dinosaur age )against computers, of course at least 99% good players already knew and might have tried against computer already. (similar to a player nicked name Grandfather who frequently posted his drawish games with stone wall here). I can tell the reason behind the loss of computer, it was because black did not develop queen bishop on c8. However I highly doubt that the same opening will work against current stockfish or Komodo of 2017 version. If you play stonewall in either 1. e3, d4, bd3 , f4 , Nf3,Ne5 in whatever order, black will respond with d5, nf6 bf5 or bg4 in either sequence. Black will not badly play like e6 and stop developing c8 bishop in that opening. Of course you may still get a draw 1 in 100 against modern computers and it should be according to your rating difference vs computers. ( your 2100+ FIDE rating vs 3100+ computers)
I don't know when people on this forum and elsewhere will finally learn that my objective current chess strength, especially under quiet conditions, is at least 2800?

that 2100+ estimate(actually 2200+) dates back from 2004-2006, and during that period I had a couple of 2400 performances.
I simply played too few rated games during that period, mostly at the weekends or in the evening, after a strenuous working day. Lack of concentration and fatigue play an enormous role in deciding a chess game.

in the course of 12 years, I have added too much chess strength, I am even afraid to think of how large that quantity is nowadays. Matter of fact is that in the last 5 years I have played and analysed at least 3 times more games than during my entire chess training span before that.

I don't know what are you talking me drawing 1 out of 100 against Stockfish. Out of 100, I am usually drawing at least 67.

I acknowledge you are right, though, concerning Bf5 in the commented Stonewall game. It is the much better choice, Komodo does play that move, but not Stockfish.
Just check, current Stockfish will again choose e6 instead.
I believe that if you play in tournament against humans and improve your rating and get the GM title more people are going to buy your books.

If you think that you improved significantly from the last time that you played against humans then I see no reason for you not to prove it by playing against humans(and if you are really strong then you can expect also to earn money from playing.
:D :)

Hey, Uri, who has so much time to waste on competitions?

You are right of course, but you need at least couple of years to get IM/GM title officially.

All I know is that, when I am commenting the games now for the book, I immediately see all available tactics, even in more complicated positions, while, when playing the very same commented games couple of years ago, frequently, I would sometimes wonder why the hell did Stockfish resign?
i.e., I have not been seeing then the relevant tactical lines, even in simpler situations.

I would assess that at at least 200-300 elo, from wondering 'What the hell is going on' to immediately seeing all tactics.

I guess also a writer's quality/strength could be assessed even by the very books written, don't you think so?

I guess you will agree there might be even players who have never played a single competitive game in their life, who don't have rating, but are actually stronger than Carlsen. Possible?

My predicament is that: to have an abundance of titles for the things I am not doing and no special titles for the very thing I am most busy with, what to do?

Still, I believe in the quality of my books.

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

A blast from the past

Post by Lyudmil Tsvetkov » Mon Oct 30, 2017 7:35 am

Most people here should be aware, but for those new, I just want to
bring back to life an intriguing thread on talkchess, involving a live chess game between me and Stockfish, played in late 2014:
http://www.talkchess.com/forum/viewtopi ... be27a1a99c
This is just to show how much analytical effort has gone into developing the right strategies to overpower the top engines.
With each move consistently analysed for half an hour, and Stockfish using 16 threads, the amount of knowledge one gets from similar sessions is certainly tremendous.

And that is only one of maybe more than a thousand similar analytical threads on talkchess during the last 5 years.

Some might try to raise cheating allegations against me, but I am worth
what I am worth.

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: A blast from the past

Post by Lyudmil Tsvetkov » Mon Oct 30, 2017 7:38 am

And this is the place to thank Louis once again, as otherwise this game would not have been possible.

tpoppins
Posts: 558
Joined: Tue Nov 24, 2015 8:11 pm
Location: upstate

Re: Part 2

Post by tpoppins » Tue Oct 31, 2017 3:45 am

Lyudmil Tsvetkov wrote:both Stockfish and Komodo like a line like 1. d4 d5 2.c4 c6 3. Nc3 Nf6 4. Nf3 g6!(this is the trick,
definitely strongest continuation) 5. e3 Bg4!(again, best) 6. h3(that is how top engines play) Bg4 7. Qf3(bishop pair lacking,
but the queen is very displaced here) e6
That's two illegal moves in a seven-move sequence.

tpoppins
Posts: 558
Joined: Tue Nov 24, 2015 8:11 pm
Location: upstate

Re: Part 2

Post by tpoppins » Tue Oct 31, 2017 5:30 am

Image

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Part 2

Post by Lyudmil Tsvetkov » Tue Oct 31, 2017 5:39 am

tpoppins wrote:
Lyudmil Tsvetkov wrote:both Stockfish and Komodo like a line like 1. d4 d5 2.c4 c6 3. Nc3 Nf6 4. Nf3 g6!(this is the trick,
definitely strongest continuation) 5. e3 Bg4!(again, best) 6. h3(that is how top engines play) Bg4 7. Qf3(bishop pair lacking,
but the queen is very displaced here) e6
That's two illegal moves in a seven-move sequence.
Is not it obvious that it is 6...Bf3?

Lyudmil Tsvetkov
Posts: 6033
Joined: Tue Jun 12, 2012 10:41 am

Re: Part 2

Post by Lyudmil Tsvetkov » Wed Nov 01, 2017 6:33 am

tpoppins wrote:Image
:D :D haha, Tim, very funny.

There will be Part III, but I don't know when and that will certainly not be
the topic.

Below a real review, by a real reader, who has read the book and understands chess, not just playing around with computers and fully trusts them.
That is what MarshallArts posts on Rybka forum:

The exhortation not to buy his books is very mean-spirited and basically unfounded. I can understand if someone does not like the playing conditions which these games were conducted under, but the criticism is going too far.

His books are quite good actually. I only skimmed through the Secret of Chess, but the newer human vs machine game books seem packed with good and crisp explanations that can help elevate a reader's play even against other human players. The games themselves are H vs M masterpieces, regardless of whatever handicaps were used by the author. A very high level understanding of chess transpires when looking at these well commented games. I was positively surprised by the quality of
these games and their annotations.



http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=32312

So that, basically, Tim, I would say, read the book, learn better chess, and then post such hilarious edited screenshots.

Post Reply