Not logged inRybka Chess Community Forum
Up Topic Correspondence Chess / Correspondence Chess / CC game centaur vs. Rybka 3 1day/move (Draw)
1 2 3 Previous Next  
- - By billyraybar (***) [us] Date 2008-09-10 22:17
1. e4  c5

rnbqkbnr/pp1ppppp/8/2p5/4P3/8/PPPP1PPP/RNBQKBNR w KQkq c6 0 2
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-11 04:39
Hi :-),

I do not like the header. It is not me against Rybka. At least I am only playing as centaur and do not even think that I will make any move not suggested by at least one engine. From my corr practice I do not believe in 24 hr, analyses, but prefer to use different engines and trying to go deeper into the lines, especially when they are more or less forced.

I will do smoe comments from time to time to make it more entertaining and you could post a diagram after each move or every third move.

Second move was already not so easy. In OTB I like to play the Morra against unprepared opponents, there are just an awful lot of traps to avoid. But in a long time game against the strongest engine there cannot be expected more than a draw in that one.

The same for 2.c3 Tiviakov (living now in the Netherlands) has some success with, but admits that against the best lines after 2..d5 and 2..Nf6 there is hardly more than a tiny advantage for white, although without much risk.

Grand Prix Attack 2.f4. Here I do not like the response 2..d5, which seems good enough for black to equalize.

Closed Sicilian 2.Nc3 I only played with black pieces and good results overall.

So my choice:

1.e4 c5 2.Nf3

Best Regards
Heinz
Parent - - By Bouddha (****) [ch] Date 2008-09-11 08:29
Thanks for your comments.

Well, the Mora seems relatively easy even OTB to neutralize and have a good games for black.

There is not so many variations to know.
Parent - By Heinz van Kempen (***) [de] Date 2008-09-11 15:11
Hi Bouddha :-),

in fact I got some 20 or 25 move wins with white playing on servers or in the club competitions against other clubs who do not know my repertoire. This is against players around 2000 ELO who are not familiar with the Smith-Morra. In any case white should get enough for a draw even against good defence. But some also do not accept the Morra by playing 3..d3 after 3.c3 or 2..e6 etc..

Some comments about the traps by in IM Lenderman are here:

http://webcast.chessclub.com/Manest/SmithMorra/part1/manest_Smith_Morra_part1.html

Part 2 and 3 only for registered users.

Best Regards
Heinz
Parent - - By Kappatoo (*****) [de] Date 2008-09-11 13:24
You forgot to mention why you didn't play the Grand-Prix after 2.Nc3 and 3.f4. :)
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-11 15:17
Hi Kappatoo :-),

yes, this is also in interesting option.

What do you think is the better of two promising strategies in the opening for this match?

-Going for a sideline to tear Rybka out of her book as quickly as possible, hoping that she will play weaker without the book and maybe still having some GM games to consult and to compare.

-Or going for one of the main lines in maybe Najdorf or Sveshnikov where we have novelties almost evey week in TWIC for example and maybe hit her with one of those? Okay, I do not know if Billy is allowed to update his book in the course of this game also with this novelties.

Best Wishes
Heinz
Parent - - By Kappatoo (*****) [de] Date 2008-09-11 15:38
Hi Heinz,
I wish I could make a substantial concrete proposal, instead of some general wishy-washy remarks.
To all appearances, the main lines in the Najdorf have been analyzed (and booked) to an insane extent in the engine's room on Playchess. If you don't stumble upon a killer novelty there, I wouldn't recommend following them, especially if you want to win. I think especially the lines in the English attack and the Bg5 line are not recommendable thus. One option would be to avoid the open Sicilian and play something like 3.Bb5+, aiming for a small but stable advantage. The downside is that the engines are no longer that bad at playing those positions, and the positions are rather drawish. Another option would be to go for a not so extremely heavily analyzed variation in the Najdorf. E.g. the 6. Bc4 e6 7.Bg5 line - it's comparatively young, and there is probably still a lot to be explored. Or Nisipeanu's 6.Be3 e5 7.Ne2!? Or 6.f4. Or even something slightly crazy like 6.h3, which was recently played by Carlsen and Karjakin or alternatively the related idea 6.Rg1, which used to be quite fashionable for a short time some years ago.
In each case, you should first have a look what latest theory has to say about these lines, and then judge whether you think there is still some potential in one of them.

Good luck for the game! Regards,
Jens
Parent - By Heinz van Kempen (***) [de] Date 2008-09-11 17:18
Hi Jens :-),

thanks for the proposals. In fact I thought about the Rossolimo with 3.Bb5 and had an eye on 6.Rg1 in the Najdorf where GM M. Wahls gave some comments on. 6.Bc4 e6 7.Bg5 I will take into account if we get a Najdorf. Before I decide I will surely investigate some of this lines with different engines.

Surely I will try to get a win and not play for a draw, even taking the risk to lose of course. I think that I have some advantages, for example I can check what Rybka 3 would probably play after 24 hours, just letting her calculate for one hour (usually engines do not change the move so often after one hour). The main thing is to get enough time every day for analyzing a lot, what I doubt will be always possible.

Best Wishes
Heinz
Parent - - By Uly (Gold) [mx] Date 2008-09-11 20:17

> I do not like the header. It is not me against Rybka.


What header would you like to have?
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-11 22:40
Hi Vytron :-),

simply CC game centaur vs. Rybka 3 1day/move or something like this.

Best Regards
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-12 04:25
Heinz,

I would change the header if I could.  I think my Marketing training got the best of me.  Anyhow, I am very excited to watch this game unfold.  I hope you win. My book's response is 1. e4 c5 2 Nf3 Nc6

r1bqkbnr/pp1ppppp/2n5/2p5/4P3/5N2/PPPP1PPP/RNBQKB1R w KQkq - 0 3
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-12 11:40
Hi :-),

so not a Najdorf. I checked some games with 3.Bb5 now. Not so bad, but I decide for 3.d4

1. e4 c5 2 Nf3 Nc6 3,d4

if 3..cd4: then 4.Nd4:

Best Wishes
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-12 13:03
Hmm... thought for sure you would play 3. Bb5.  Well...

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6

r1bqkb1r/pp1ppppp/2n2n2/8/3NP3/8/PPP2PPP/RNBQKB1R w KQkq - 0 5
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-12 14:52
Hi :-),

I did not like 3.Bb5 g6 too much. Seems to be sufficient for Black according to the latest games I found.

Now hardly any choice:

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nx3

Probably a Sveshnikov or Richter Rauser now where it will not be so easy to avoid long theoretical lines, but I hope we willl be out of book before move 25.

Readers feel free to suggest any secret weapons :-). I am also checking Jeroen´s book of course.

Best Wishes
Heinz
Parent - By billyraybar (***) [us] Date 2008-09-12 15:21
Hi :),

Yeah, I hope we are out of book soon as as well.  We'll see.

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5

If 6. Nxc6 then bxc6 and if 6. Ndb5 then d6

r1bqkb1r/pp1p1ppp/2n2n2/4p3/3NP3/2N5/PPP2PPP/R1BQKB1R w KQkq e6 0 6
Parent - - By Uly (Gold) [mx] Date 2008-09-12 21:42

> simply CC game centaur vs. Rybka 3 1day/move or something like this.


Changed.

Also, why is this at the flip side? Should I move this to Chess (because it's a chess game), Computer chess (because it's against a computer), or Rybka main? (because it's against Rybka.)
Parent - - By billyraybar (***) [us] Date 2008-09-12 21:50
Thanks for changing the header.  I wasn't sure where this belonged so I chose the 'Flip Side' because it encompasses everything.  Moving this to 'Rybka main' would get the most exposure so I would be for that, but it is totally up to you. Wherever you think this should be is fine with me.
Parent - - By Uly (Gold) [mx] Date 2008-09-12 22:02
I moved it to main (the flip side is for discussions not related to chess/computer chess, or personal discussions), if it gets badly buried I'll move it to computer chess.
Parent - - By billyraybar (***) [us] Date 2008-09-12 22:07
Thanks
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-13 02:36 Edited 2008-09-13 02:38
Hi :-),

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5

if 7..Nxd5 then 8.exd5:

I am going for a sideline here. Yes, I know, it is very rarely played by Top GM`s, but the main lines after 9.Bxf6 or 9.Nd5 are analyzed to death. I checked some games by Shirov from the latest years and even after hours Rybka did not find a novelty that would be crushing or at least very promising in all subvariations. Some engines like Zappa, DS11 and Hiarcs overestimate the white position in a few main lines in the Sveshnikov, probably because of the weakness of the e6 square. Rybka or Naum evaluate black´s counterchances more realistically.

Ok, I also do not claim that the line I chose will give anything to white. It is just a try.

Best Regards
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-13 03:58
Hi,

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8

If memory serves, my book's lines could very well end soon here depending on your response.

rnbqkb1r/pp3ppp/3p4/1N1Pp3/8/8/PPP2PPP/R1BQKB1R w KQkq - 0 9
Parent - - By Uri Blass (*****) [il] Date 2008-09-13 06:59
Why do you trust book?

What about giving rybka to calculate for 24 hours without book?

I have big doubt if theory at move 5 is the best move
Here is some analysis that I made of 1.e4 e5 2.Nf3 Nf6 3.Nxe5 d6 4.Nxf7 Kxf7

Rybka insist on 5.Nc3 when the theory move in noomen's book is 5.d4
Are you sure that 5.d4 is better than 5.Nc3?

so k - Rybka 3 1-cpu 32-bit, Friend mode
rnbq1b1r/ppp2kpp/3p1n2/8/4P3/8/PPPP1PPP/RNBQKB1R w KQ - 0 1


Analysis by Rybka 3 1-cpu 32-bit :

5.Bf1-c4+
  µ  (-1.24)   Depth: 2   00:00:00
5.Nb1-c3
  µ  (-1.11)   Depth: 2   00:00:00
5.Nb1-c3
  µ  (-1.17)   Depth: 3   00:00:00
5.Nb1-c3
  µ  (-1.11)   Depth: 4   00:00:00
5.Nb1-c3
  µ  (-0.91)   Depth: 5   00:00:00
5.Nb1-c3 Bc8-e6
  µ  (-0.90)   Depth: 5   00:00:00
5.Nb1-c3 Bc8-e6 6.d2-d4
  µ  (-0.88)   Depth: 6   00:00:00  5kN
5.Nb1-c3 Bc8-e6 6.d2-d4 c7-c6 7.Bf1-e2
  µ  (-0.84)   Depth: 7   00:00:00  12kN
5.Nb1-c3 c7-c6 6.Bf1-e2 Nb8-a6 7.0-0
  µ  (-0.80)   Depth: 8   00:00:01  33kN
5.Nb1-c3 Qd8-e7 6.d2-d4 Qe7-e6
  µ  (-0.76)   Depth: 9   00:00:02  60kN
5.Nb1-c3 c7-c6 6.Bf1-e2 b7-b5 7.a2-a3 Nb8-d7 8.0-0 Nd7-c5
  ³  (-0.70)   Depth: 10   00:00:05  110kN
5.Nb1-c3 c7-c6 6.Bf1-e2 Bf8-e7 7.0-0 Rh8-f8 8.d2-d4 Kf7-g8 9.Qd1-d3 Bc8-e6
  ³  (-0.61)   Depth: 11   00:00:17  370kN
5.Nb1-c3 c7-c6 6.Bf1-e2 Bf8-e7 7.0-0 Rh8-f8 8.d2-d4 Kf7-g8 9.Qd1-d3 Bc8-e6
  ³  (-0.61)   Depth: 12   00:00:27  575kN
5.Nb1-c3 c7-c6 6.Bf1-e2 Bf8-e7 7.0-0 Rh8-f8 8.d2-d4 Kf7-g8 9.Qd1-d3 Bc8-e6
  ³  (-0.61)   Depth: 13   00:00:36  786kN
5.Nb1-c3 c7-c6 6.Bf1-e2 Qd8-a5 7.0-0 b7-b5 8.a2-a3 b5-b4 9.Be2-c4+ d6-d5 10.e4xd5 Bc8-a6 11.Bc4-a2
  ³  (-0.57)   Depth: 14   00:01:18  1639kN
5.Nb1-c3 c7-c6 6.Bf1-e2 Bf8-e7 7.0-0 Rh8-e8 8.d2-d4 Be7-f8 9.f2-f4 Nf6xe4 10.Nc3xe4 Re8xe4 11.Be2-d3 d6-d5 12.Bd3xe4 d5xe4 13.f4-f5 Nb8-d7 14.Bc1-e3
  ³  (-0.50)   Depth: 15   00:05:41  7390kN
5.Nb1-c3 Bc8-g4 6.f2-f3 Bg4-e6 7.f3-f4 Kf7-g8 8.d2-d4 Nb8-d7 9.Bf1-e2 c7-c5 10.d4-d5 Be6-f7 11.0-0 Qd8-b6 12.Be2-c4 a7-a6 13.Qd1-d3 Qb6-c7
  ³  (-0.46)   Depth: 16   00:16:06  21451kN
5.Nb1-c3 Bc8-g4 6.f2-f3 Bg4-e6 7.f3-f4 Kf7-g8 8.d2-d4 Nb8-d7 9.Bf1-e2 c7-c5 10.d4-d5 Be6-f7 11.0-0 Qd8-b6 12.Be2-c4 a7-a6 13.Qd1-d3 Qb6-c7
  ³  (-0.46)   Depth: 17   00:25:19  34495kN
5.Nb1-c3 Bc8-g4 6.f2-f3 Bg4-e6 7.d2-d4 d6-d5 8.e4-e5 Nf6-h5 9.Bc1-e3 c7-c5 10.Nc3-e2 Nb8-c6 11.g2-g4 c5xd4 12.Ne2xd4 Bf8-c5 13.c2-c3 Nc6xe5 14.g4xh5 Qd8-h4+ 15.Be3-f2 Qh4xh5 16.Bf1-e2 Rh8-e8 17.0-0 Ne5-c4 18.f3-f4 Qh5-g6+ 19.Bf2-g3
  ³  (-0.44)   Depth: 18   00:40:55  55845kN
5.Nb1-c3 Bc8-e6 6.d2-d4 c7-c5 7.d4-d5 Be6-c8 8.Bf1-e2 Qd8-e7 9.f2-f3 Nb8-d7 10.0-0 a7-a6 11.Qd1-e1 Nd7-e5 12.Bc1-e3 Bc8-d7 13.a2-a3
  ³  (-0.38)   Depth: 19   01:42:59  145177kN
5.Nb1-c3 Bc8-e6 6.d2-d4 Kf7-g8 7.Bf1-d3 c7-c5 8.d4-d5 Be6-f7 9.0-0 Nb8-d7 10.Bc1-e3 Qd8-e7 11.b2-b3 h7-h6 12.Qd1-d2 Nd7-e5 13.h2-h3 Kg8-h7
  ³  (-0.39)   Depth: 20   02:37:49  226040kN
5.Nb1-c3 Bc8-e6 6.d2-d4 Kf7-g8 7.Bf1-d3 c7-c5 8.d4-d5 Be6-f7 9.0-0 Nb8-d7 10.Bc1-e3 Qd8-e7 11.b2-b3 h7-h6 12.Qd1-d2 Nd7-e5 13.h2-h3 Kg8-h7
  ³  (-0.39)   Depth: 21   04:24:25  387552kN
5.Nb1-c3 Bc8-e6 6.d2-d4 Kf7-g8 7.Bf1-d3 c7-c5 8.d4xc5 Nb8-c6 9.c5xd6 Qd8xd6 10.h2-h3 Ra8-d8 11.0-0 Nc6-b4 12.Bc1-e3 Nb4xd3 13.c2xd3 Qd6xd3 14.Qd1-f3 Qd3-c2 15.Rf1-c1 Qc2xb2 16.Be3-g5 Bf8-e7 17.Ra1-b1
  ³  (-0.35)   Depth: 22   11:43:28  1051065kN
Parent - - By billyraybar (***) [us] Date 2008-09-13 13:27
Uri,

I do not know what you mean.  Rybka with no book vs a very strong CC player?  If Rybka does well with this game then next game maybe can be under those conditions.
Parent - - By Uri Blass (*****) [il] Date 2008-09-13 13:32
My point is that a book may be a disadvantage at correspondence game because the book may include blunders.
I think that it is better to use smaller book at longer time control because slower time control increase the probabilbility that the program is going to find better moves than the book moves.

Uri
Parent - - By benben (***) [au] Date 2008-09-13 13:36
I believe theory moves say at move 5 are stronger than rybka at depth 25 or so.
Parent - - By Uri Blass (*****) [il] Date 2008-09-13 14:02
1)I believe that in most cases rybka is going to choose theory move and the choice of the right theory move may be important when there is more than one theory move.

2)I see no basis to believe that theory moves at move 5 are stronger than rybka at depth 25.
The only case when I believe in it is when there are some long forced line that were analyzed by humans but I expect rybka to be better when there is no long forced line.

Uri  
Parent - - By billyraybar (***) [us] Date 2008-09-13 14:22
I agree about using smaller book.  The book I am moving is rather small.
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-13 15:25 Edited 2008-09-13 15:29
Hi :-),

better to use no book. Without it a weak engine like Rybka would probably have fallen to the known opening trap:

7.Nd5 Nxd5 8.exd5 Ne7 9.c4 a6? 10.Qa4 Bd7?? 11.Nd6 mate, LOL.

Ok, got your move now and will answer tomorrow. Checking quickly my two main databases Megabase 2008 and TWIC there are no players rated above 2600 who played this with white. The best I found are:

Solak (2564) 9.Qf3
Groszpeter (2539) 9.g4 (probably dubious)
Kholmov (2505) 9.Be2

Jeroen´s book features anyway seven alternatives: 9.c4, 9.Bd3, 9.Be2, 9.Be3, 9.a4, 9.Qd3 and 9.Qf3, the last two ones with the idea 9..Be7 10.Qc3 or 9..a6 10.Qa3. Here I will go deeper into the lines with Rybka 3 Human/Dynamic and Zappa as long as line are more or less forced. All top engines after one hour now show the moves given in Rybka3.ctg, Zappa additionally 9.Bc4, 9.Nc3 (two moves I do not like) and even for some time 9.f4 which it dropped now.

Best Regards
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-13 15:37
Heinz,

I am confused.  Do you mean that no players rated over 2600 have played the moves you have played so far? This is absolutely wrong.  Polgar, Svidler, Anand, Leko, Short are GMS over 2700 that have played your moves.
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-13 15:59
Hi :-),

you are right of course. I found a lot more after installing a decent opening key to the databases I have in ChessBase 9. So still possible that we get a lot of "theory moves". But already at this stage I will analyze with engines and not trust exclusively in the top GM´s.

Best Regards
Heinz
Parent - By billyraybar (***) [us] Date 2008-09-13 16:11
Understood
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-13 15:08
Hi Uri :-),

the Cochrane Gambit 4.Nxf7 is probably underestimated.

I have added this to the generic set with only the first four moves and there will be 12 games totally 40/120 repeated for Rybka 3 against the six opponents so far. Of course the games will not go to the rating list, if white loses all or almost all, what I doubt. Maybe a time control of 40/120 is not sufficient to prove the strength of the move.

The games I will post next week.

Kind Regards
Heinz
Parent - By Uri Blass (*****) [il] Date 2008-09-13 17:59
Thanks
I also do not believe that white is going to lose almost all games and for me the interesting question is if white can get at least 50%

Uri
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-13 22:11 Edited 2008-09-13 22:37
Hi :-),

my choice is 9.a4

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8 9.a4

Against the main line with 9.c4 GM did fairly well and I found no improvements with the engines. 9.c4 is the main choice of Rybka 3 Human and Zappa after a few hours. The move 9.Qf3 proves not to be a good one on further analyses, although white got some quick wins (but here opponents played badly).

Some of the analyses I did and games I looked at are here:

http://www.husvankempen.de/nunn/Replay/cr3.htm

You can view it better in Chess Base or whatever you have after download (press "hier" button).

Will be updated from time to time when the game proceeds.

Kind Regards
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-15 05:45
Hi ;),

Here we go. Out of book.  Just now entered the move (I was out of town) and Rybka is now thinking.  I am out of town every other day, usually returning 24-36hrs later.  So, there is a small chance that, if you post your move just as I leave out of town that I will have to either let Rybka think less than 24hrs on given move or post the move a little late.  I'll do my best to reply within 24hrs.  I expect to get called to work in 12-18hrs so I'll just post Rybka's response at that time.  Anyhow so far we have

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8 (last book move for Rybka) 9.a4

rnbqkb1r/pp3ppp/3p4/1N1Pp3/P7/8/1PP2PPP/R1BQKB1R b KQkq a3 0 9
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-15 11:29
Hi :-),

more interesting that Rybka is already out of book on move 9. There are still a lot of GM games with 9.a4, but maybe Rybka will find something better than those, because according to Megabase 2008 stats white scores good here (64%). My guess is that we will have a balanced position after a few moves more.

About this 24-36 hrs....We have all the time in the world. Usually such a game lasts a few months. I will also not be able to check the forum every few hours and so maybe will only get the new move 12 hours after it was posted and so will also have to use 36 hrs, leaving the engines and me with 24 hrs. to analyze. Please do not let Rybka calculate less than 24 hrs. for one move, except when it is forced. When it calculates a few hours more from time to time this will not make a lot of difference usually. At this depths it lasts very long to gain one ply more.

Kind Regards
Heinz
Parent - By Mark (****) [us] Date 2008-09-15 12:28
For these long time controls, it will be interesting to see how often the move at 12 hours differs from the move at 24 hours.
Parent - - By billyraybar (***) [us] Date 2008-09-15 17:13
Hi,

Yes, out of book - I 'optimized' my .ctg book, thus making  moves to have occurred in a minimum of 3 games to be played.  My book only contains 2 games for black's response here. 

Rybka is currently at depth 28, 8826mN, after 12hrs of thinking.  I will try to let it think for the full 24hrs before posting the response.
Parent - - By billyraybar (***) [us] Date 2008-09-15 22:33
Hi,

Well I accidently hit this space bar on my keyboard, causing the current move to be played, so I used 18hrs on this position.  I don't see any other choice than to make the move now.  Rybka is thinking and awaiting your reply.

rnbqk2r/pp2bppp/3p4/1N1Pp3/P7/8/1PP2PPP/R1BQKB1R w KQkq - 0 10


1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8 (last book move for Rybka) 9.a4 Be7
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-16 22:08 Edited 2008-09-16 22:11
Hi :-),

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8 (last book move for Rybka) 9.a4 Be7 10.Be3

Most GM play 10.Be2 here (this would be also Zappa´s choice). But there is one critical variation where Black scores good and engines do not like it too much on further analyes for white: 10.Be2 0-0 11.0-0 Sd7 12.Kh1 a6 13.Sa3 f5

So I opt for 10.Be3 proposed by Rybka 3 Human and also played in some GM games, aiming for a setup Qd2-Be3 against an oncoming f5 and with the knight going to c4 after an expected a6 Na3 at some point and also trying to have control over the b6 square if possible.

Kind Regards
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-18 03:39
Hi again,

Rybka plays Nd7.  She reached depth 27 with evaluation of .06 in 6 1/2 hours and remained at that ply for 16 more hours. 

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8 (last book move for Rybka) 9.a4 Be7 10.Be3 Nd7

r1bqk2r/pp1nbppp/3p4/1N1Pp3/P7/4B3/1PP2PPP/R2QKB1R w KQkq - 0 11
Parent - - By Uly (Gold) [mx] Date 2008-09-18 04:55

> and remained at that ply for 16 more hours.


That's when forcing the move and waiting for Rybka to reach depth 27 in the next move (in this case, depth+1) is faster. For that reason alone Centaur > Rybka 3, while Rybka 3 will not use the 24 hours fully (e.g. in this case the move had the quality of a 6 1/2 hours move), an interactive centaur will use the 24 hours much better.

Specially when Rybka gets a fail low and would need 24 hours and 1 minute to switch to a better move, while a centaur would notice it and find a better move immediately.

What is interesting is to know if this, and other advantages for the centaur are enough to win.
Parent - - By Dragon Mist (****) [hr] Date 2008-09-18 10:36
Vytron, such a Centaur would be Good-like. To see what R3 didn't see up to depth 26? And to make a decision what move is better at that depth? One would need at least Sampled Search, multi PV or something alike. Or all together.
Parent - By Uly (Gold) [mx] Date 2008-09-18 18:55

> such a Centaur would be Good-like.


Not that much.

> To see what R3 didn't see up to depth 26?


The centaur is able to extend the lines as much as he wants and to reduce them, getting effective pruning beyond what the engine can do by itself. I'm sure that most of the time the analysis is going to be on par with the engine, and then the centaur only needs to see some key move that the engine is missing to win the game, and each move is a chance for the centaur to see it, move it and win. I'm talking about people experienced in doing this, not just some random person that is given an engine and might only weaken it.

> One would need at least Sampled Search, multi PV or something alike. Or all together.


That's the key, the centaur has 24 hours to do whatever he wants to the position to analyze it, making use of Monte Carlo, IDeA, DPA, Persistent Hash, effective move excluding, using other engines that may have a better grasp of the position, etc. as needed while if Rybka's branching factor goes over 4 she may play a simple 6 hours infinite analysis move.
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-19 06:48
Hi :-),

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8 (last book move for Rybka) 9.a4 Be7 10.Be3 Nd7 11.Be2

An alternative would be 11.Qd2, but on further analyses I did not like variations too much, for example 11..a6 12. Na3 f5 13.g3 0-0 14.Nc4 b6 or 13.f3 0-0 14,Nc4 b6 17.a5 b5 18.Nb6 like it was played in some games. With about equal chances like in the game now.

Kind Regards
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-19 15:59
Hi Heinz,

This move might take a long while.  I have to travel to Cleveland and may not be back for 36hrs or so.  If I see Rybka coming up with a different move after 24hrs then I will not play the new found move.  Currently Rybka is at depth 28.

P.S. Rybka expected Qd2 here and is now evaluating the position at 0.
Parent - - By Heinz van Kempen (***) [de] Date 2008-09-19 16:08 Edited 2008-09-19 16:12
Hi Billy (or what´s your name) :-),

does not matter so much if 36hrs or 24hrs. You should always give the best move Rybka comes up with even when she calculated a bit more. For me it is not so important to win or lose but to get tough opposition and seeing how my chances would be with the strategy I am using. This will be mainly the following...letting R3 Human and Zappa calculate for 6 to 12 hours and when they differ giving the moves to the other one and maybe also try other engines depending on the position. When there is something more or less forced going deeper into the lines maybe also using all features of Rybka for analyzing.

Best Regards
Heinz
Parent - - By billyraybar (***) [us] Date 2008-09-20 19:49
Hi :)),

0-0, is Rybka's move, depth 28, reached at 8 1/2 hrs and this branching factor is insane -- no new depth for the next 23 hrs.

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 Nxd5 8. exd5 Nb8 (last book move for Rybka) 9.a4 Be7 10.Be3 Nd7 11.Be2 0-0

r1bq1rk1/pp1nbppp/3p4/1N1Pp3/P7/4B3/1PP1BPPP/R2QK2R w KQ - 0 12
Parent - - By Dragon Mist (****) [hr] Date 2008-09-21 00:13
Rybka 3 branching factor is quite ok. I am "the branching factor expert", trust me. Or ask Vas. :-)

While 2.3.2a used to take bf of 50 or even more sometimes without any reason, R3 is decent, and is on the average of 2, or even less. Sometimes she goes for bf 3-4, but this can mean one of 2 things:
- your move is failing low :-(
- there is a major change in the PV relatively close to the root position.

In your example, you are close to bf 4, nothing to be excited about yet.
Parent - - By Uly (Gold) [mx] Date 2008-09-21 06:59

> or even less.


Well, I don't like this, I've seen her getting a branching factor of like 0.1, often, so she takes 30 seconds for the whole analysis and then suddenly the next iteration is finished in 5 seconds. I don't like it because it means either I have to wait 5 seconds every time just to check if the next iteration would end fast (and these 5 seconds are cumulative meaning I lose more time every time she doesn't) or I don't and so I'm losing an extra ply every time the branching factor would be this low.

For people like me that analyze a lot of positions interactively it's important to know when to jump to the next move, and a parameter of "time predicted for next iteration" would be nice, because assuming a branching factor of 2 isn't efficient. It's possible to predict it and show estimated time, old DOS Rebel did it.
Up Topic Correspondence Chess / Correspondence Chess / CC game centaur vs. Rybka 3 1day/move (Draw)
1 2 3 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill