Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / what does the cluster think the best move is here?
1 2 Previous Next  
- - By chessorpoker (**) Date 2009-09-27 17:30
r3k2r/2q1bpp1/p2p3p/P1nPp3/1p4PP/1N3P2/1PPQ4/1K1R1B1R w kq -


my rybka3 goes on a long sequence of 1.Qxb4 0-0 Nxc5 Rfc8
Parent - - By nine castles (**) Date 2009-09-28 20:53
More importantly, why should anyone help you cheat in your correspondence game?
Parent - - By chessorpoker (**) Date 2009-09-28 20:59
actually, the game is already finished. i just wanted to see what the cluster or strong hardware would play. anyhow if you knew me at all you would know im strong enough, that i dont need to cheat.
Parent - - By Yoav Dothan (**) Date 2009-09-29 13:13
I checked in my updated games database and could not find your position. Could you send the link to the game? if it is in the ICCF server it would be ideal. Otherwise I am sorry to join the view that you ask for help in a current game.
Parent - - By chessorpoker (**) Date 2009-09-29 21:58 Edited 2009-09-29 22:02
i didn't really want to state what the right move is here as its taken me 8 months to finish this game. i was just wondering what cluster rybka or strong hardware thinks about this position as all i have is a dual core pc. i can promise you that Qxb4 is not the best move and i seriously doubt that any program is going to find the best reply. i'm just curious if the cluster can figure it out. i don't cheat and don't need to cheat as im a chess expert with victories and draws against titled players.

so, does rybka cluster or strong hardware recognize that 1.f4!! is the best move? or is it a fail low etc.

my thinking here is that :

1. blacks king is in the center
2. the last move played on the wing-therefore playing in the center is ideal
3. blacks bishop is a bad bishop
4. QRBNK vs QNB has give white the advantage; yes whites bishop is developed at f1
5. white has an advantage in position and time which is more valuable than material

honestly, what you all need to understand is that programs play weak chess and any strong postal player loves when someone uses one.
Parent - - By yanquis1972 (****) Date 2009-09-30 04:53
ehmm, sure.

Analysis by Rybka 3:

1. +/=  (0.43): 1.Qxb4 0-0 2.Nxc5 Rfc8 3.c3
2. +/=  (0.31): 1.f4 Bf6 2.Nxc5 Qxc5 3.Bd3 exf4 4.Rhe1+

thats after abiyt 2 minutes of minor interactive analysis (ie forwarding rybka's variations) in multi pv mode. it might take more work to find that f4 is better than Qxb4 (that's assuming it is) but any postal player worth his salt could do it easily. rybka could probably do it by herself while you're sleeping.  i mean i can see f4 is a legitimate, possibly winning move in seconds, simply by being a rybka kiddie. but keep thinking she plays weak chess -- whatever CC players there are here will certainly appreciate knowing that you do.
Parent - - By chessorpoker (**) Date 2009-09-30 07:26
i've beaten rybka3 37 times, drawn 11 games now in postal chess. i think, tactically it is very nice in the middlegame, however it does not understand why moves are played in the opening for positional gain and its endgame could use a lot of improvements.

also what hardware are you using?
Parent - By yanquis1972 (****) Date 2009-09-30 07:35
well i agree with you on those counts. however i am 100% certain anyone playing serious correspondence chess would not rely wholly on rybka for opening analysis.  i am also certain, btw, you did not beat rybka 3 without any computer assistance all those times, or most, & possibly not any. obviously rybka 3 + a strong, or even just fairly smart, human is going to be stronger than rybka 3 alone.

i was using a quad but the results would be similar on 1 cpu i think. f4 shows up as the #2 or #3 move more or less instantly in the pv.
Parent - By Uri Blass (*****) Date 2009-10-01 18:28
If you beat rybka in most of your games then I guess that the reason is that your hardware is faster and your opponents do not give rybka a long time to analyze.

Uri
Parent - - By jaboo2 (**) Date 2009-10-01 20:24 Edited 2009-10-01 20:33

> i've beaten rybka3 37 times, drawn 11 games now in postal chess. i think, tactically it is very nice in the middlegame, however it > does not understand why moves are played in the opening for positional gain and its endgame could use a lot of improvements.


> also what hardware are you using?


You've beaten Rybka 3 37 times? Then you must be a GM. I've beaten Rybka 3 only 1 time but on weak hardware and it was the 32 bit version.

I have the feeling though that these engines are not perfect and many times have trouble in the opening. Chess programs don't understand openings as well as humans do and if you let the program play the opening without its opening book then it will play the opening very poorly.

I believe human is still better than computer. Kramnik and Anand are still stronger than Rybka alone. The day will come though, probably in 10 years from now when Rybka will be so strong as to clearly surpass the strongest human. I think that will be Rybka 5 or 6 running on 16 cores or more.
Parent - By InspectorGadget (*****) Date 2009-10-02 07:16

> Kramnik and Anand are still stronger than Rybka alone. The day will come though, probably in 10 years from now when Rybka will be so strong as to clearly surpass the strongest human.


Wow!
Parent - By Uri Blass (*****) Date 2009-10-02 19:26
1)I think that you miss the fact that it is a correspondence game when usually people use rybka in correspondence games.
The claimthat I understand  is that the team of the poster and rybka beat rybka 37 times in correspondence games.

2)I think that it is obvious that kramnik or anand are weaker than rybka in OTB chess(average time of some minutes per move )
I believe that rybka is also stronger than them in correspondence chess when the players get 24 hours per move or 48 hours per move.

3)I do not believe the poster that he beat rybka 37 times in correspondence games unless he used better hardware or more computer time.
I believe that the gap between rybka+human and rybka with no help in correspondence chess is relatively small and you need more than 100 games to get 37 wins against rybka and I doubt if the poster had time to play more than 100 correspondence games(I think that serious correspondence players even had no time to play even 37 games from the time that rybka came out).

I can say that I started only 26 correspondence games from the time that rybka3 is available and I feel that it is clearly too many games(I finsihed 11 of them at this time).

Uri
Parent - - By Mark (****) Date 2009-10-01 10:40
Here's about a day of IDeA analysis.  f4 is starting to pull ahead of Qxb4.
Parent - - By chessorpoker (**) Date 2009-10-01 15:52
thanks mark, apparently idea analysis IS worth something after all...
Parent - By Felix Kling (Gold) Date 2009-10-01 17:04
Well, Larry, Dzindzi, Iweta and Dennis Breder are some masters using IDeA. Seems to be useful ;)
Parent - - By Mark (****) Date 2009-10-01 18:05
Yes, I really like IDeA.  It's not always faster than interactive infinite analysis, but it works well if you have plenty of time to run it.  I'm usually running some position from the forum constantly.  Also, I like how it creates a tree with all of the analysis.
Parent - - By chessorpoker (**) Date 2009-10-01 19:49 Edited 2009-10-01 19:52
i wish i could get it to work. as soon as i try aquarium crashes every time. i have all the current updates, i really dont get it. maybe the next version of aquarium will be more helpful.

thanks for not posting all the analysis after 1. f4
Parent - - By Mark (****) Date 2009-10-01 20:38
Too bad you're having trouble with Aquarium.  I guess you've already tried uninstalling and reinstalling.  Could it be trying to load a corrupted file?  Did you completely delete all analysis files, trees, etc?

If you're interested in the IDeA tree, I could send it to you.
Parent - By yanquis1972 (****) Date 2009-10-01 20:40
i'm thinking this guy's just an ass. how about not sending him jack squat?
Parent - - By George Tsavdaris (****) Date 2009-10-01 22:03 Edited 2009-10-01 22:06
IDeA seems indeed very strong(in fact that was the main reason i bought Aquarium) but can it find for example the following mate in 127 after some(long) time?

Mate in 127 for white:
8/7p/7p/p4n1p/b3Q2p/K2p3p/p1r5/rk5n w - - 0 1


Or this a little tougher one:
Mate in 292 for white:
q5nn/1p2p3/p1k1P1p1/6Pp/PKp1p1pP/8/2P1P1PP/3Q4 w - - 0 1


I hope with this much easier mate it will have some luck:
Mate in 77 for white:
8/1p4Pp/1p6/1p6/1p5p/5r2/4Kpkp/6br w - - 0 1
Parent - By Mark (****) Date 2009-10-02 10:32
Just for fun, I let IDeA run for 12 hours on the mate in 77.  No luck in finding the solution.  IDeA only sees a draw...
Parent - - By Uri Blass (*****) Date 2009-10-01 18:30
I doubt it.
I have no computer time to give it to analyze with infinite analysis but
it is possible that rybka can find the same move also with infinite analysis.

Uri
Parent - - By Felix Kling (Gold) Date 2009-10-01 18:50
In theory you are right, Rybka should be stronger in IA. But with longer time search becomes more uneffective and the time to the next depth is so long that it's unpracticable.
Parent - - By Highendman (****) Date 2009-10-01 19:49
Hi,

Why would IA be better than IDEA crawling through variations, running on the same h/w for the same amount of time?

If someone would run the same position for 10 hours on IA, or invest the same 10 hours on IDEA crawling through a position with reasonable settings of depth/time per position, I would assume IDEA would get deeper and wider?

After some depth in IA the time it takes to extend to the next depth becomes soooo long. And assuming the IDEA algorithm for deciding which lines to extend is rational, IDEA should end up reaching better coverage.

Obviously some people here have tested this?
Parent - By Felix Kling (Gold) Date 2009-10-01 20:04
First the move selection algorithm of IDeA is rather primitive, so Rybka's search is more efficient than IDeA. It would be deeper than IDeA. However, as you don't want a single line and wait hours for the next depth, IDeA is better for long analysis.

With the interactive part (marking moves red/green) it becomes even more efficient than Rybka's search in some cases.
Parent - - By Banned for Life (Gold) Date 2009-10-01 21:11
Vas has complete flexibility in how he implements IA. If the IDeA algorithm were more effective at coming up with strong moves (as measured by Vas' test methodology), Vas could use it as a replacement for the current IA algorithm. I suspect that in reality, IDeA is not as strong, in the general case, as IA, so this won't happen.

Of course IDeA still has advantages, even if it isn't as strong as IA. The biggest of these is that the IDeA tree lets you see scores for a very large number of PVs rather than just the one you get with IA.
Parent - - By Uly (Gold) Date 2009-10-01 23:26
I think there should be a distinction made when talking about interaction, IDEA with the I means that the user is interacting actively with it, while DEA without the I would mean it's fully automated, that's the one people have been referring to, I think.

The I part can be critical, and then its power could be much stronger than IA, as I'm being shown in my game against Rowlando, he plays really strong moves by using IDEA even though he's using a significantly slower hardware than me, and has neutralized my use of IA enhanced with Persistent Hash. So it's difficult to discuss IDEA when the most important part is the person using it, knowing how to interact is the key.
Parent - - By Banned for Life (Gold) Date 2009-10-02 05:15
I'm not sure if this conflicts with what I've written above. If an engine used IDeA for its search, and had an aquarium like interface, it would theoretically be possible to interact by suggesting new lines and cutting off analysis of undesired lines. This would be interesting to see, and would result in a tree structure which is generally a lot more useful than a single PV, but I'm guessing that playing strength would be compromised regardless of whether there was interaction or not (assuming interaction with IDeA would be measured against interaction with IA).

Regarding your game against Rowlando, isn't he a much better chess player than you are (assuming no engine assistance)? We tend to ignore this small detail...
Parent - By Uly (Gold) Date 2009-10-02 17:41

> Regarding your game against Rowlando, isn't he a much better chess player than you are (assuming no engine assistance)? We tend to ignore this small detail...


Yes, OTB I still miss when opponents hang their pieces against me (I consider it worse than hanging them myself.) So with his strength he'd be able to know when the engine suggest a stupid move and prune it, and force smart moves that the engine is not considering ASAP, while I have to rely on engine evaluation, and when the best moves are hated by engines at high depth (or when long variation needs to be forced for the engines to realize they're best) I'd just miss them.

Perhaps this has more importance than hardware and that's the reason I was never able to succeed by using IDEA.
Parent - By buffos (Silver) Date 2009-10-03 16:34
Comparing wrong things.

IDEA needs a compass to walk the labyrinth of variations. This compass is engine evaluation. The engine search, uses move generation once, and reuses it for that position again and again. It has memory and extends its result structure.
IDEA does not have that benefit. To find a new move in a position, it needs to ask the compass for directions again. Its not efficient at all BUT the benefits are clear. The analysis are preserved, it will probably produce more knowledge in the same time as IA and ofcourse there is human interaction

IA is not designed for human analysis. Its a byproduct of engine play. When and if chess engines will focus analysis (rather than play OTB) then we will have more things to talk about.
The montecarlo approach is a more "rough" approach to the same problem. IDEA is the smarter one.

My vision (hopes). In the following year, IDEA will provide some API, to read write trees, and the engines will support IDEA and adding moves, extending moves... etc will be done by the engine MUCH faster.
Parent - By nine castles (**) Date 2009-10-01 20:26
LOL you are a world champion then ? :rolleyes:
Parent - - By Lukas Cimiotti (Bronze) Date 2009-10-01 21:14
r3k2r/2q1bpp1/p2p3p/P1nPp3/1p4PP/1N3P2/1PPQ4/1K1R1B1R w kq - 0 1


Analysis by Rybka 3v61 cluster NL:

1.Qxb4 Nxb3 2.Qa4+ Qd7 3.cxb3 h5 4.g5 Bd8 5.Qb4 0-0 6.Bd3 Bc7
  +/=  (0.60)   Depth: 8   00:00:00  248kN
1.Qxb4 Nxb3 2.Qa4+ Qd7 3.cxb3 h5 4.g5 Bd8 5.Qb4 0-0 6.Bh3 Qc7 7.Rc1 Rb8
  +/-  (0.72 !)   Depth: 9   00:00:00  396kN
1.Qxb4 Nxb3 2.Qa4+ Qd7 3.cxb3 Qxa4 4.bxa4 Kd7 5.Rc1 Rhc8 6.Rxc8 Kxc8 7.Bd3 Kb7 8.b4 g6 9.h5 Rf8
  +/-  (0.71)   Depth: 9   00:00:00  441kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.f4 exf4 4.Re1 dxc5 5.Qe4 Bd6 6.g5 hxg5 7.hxg5 g6 8.Qa4 Qb7 9.Bg2 Rab8
  +/=  (0.60)   Depth: 10   00:00:00  1752kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qe4 g6 5.h5 Rab8 6.Qa4 Rb4 7.Qa3 e4 8.fxe4 Bf6 9.b3
  +/=  (0.48)   Depth: 11   00:00:01  4964kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qe4 g6 5.Qa4 Rab8 6.Bxa6 Rc7 7.c4 Kg7 8.h5 Qe3 9.hxg6
  +/=  (0.55)   Depth: 12   00:00:02  5878kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qe4 g6 5.Qa4 Rab8 6.Bxa6 Rc7 7.Bd3 Rcb7 8.b3 Rb4 9.Qa2 Qxd5 10.Be4 Qb5 11.h5 Ra4 12.Qb2 Qc5
  +/=  (0.42)   Depth: 13   00:00:05  13256kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qe4 g6 5.Qa4 Rab8 6.Bxa6 Rc7 7.Bd3 Rcb7 8.b3 Rb4 9.Qa2 Qxd5 10.Be4 Qb5 11.h5 Ra4 12.Qb2 Qc5
  +/=  (0.42)   Depth: 14   00:00:05  14778kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qe4 g6 5.Qa4 Rab8 6.Bxa6 Rc7 7.Bd3 Rcb7 8.b3 Rb4 9.Qa2 Qxd5 10.Be4 Qb5 11.h5 Ra4 12.Qb2 Qc5
  +/=  (0.42)   Depth: 15   00:00:14  39906kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qxc5 Rxc5 5.c3 Bd8 6.h5 Bg5 7.Kc2 Rxa5 8.b4 Ra2+ 9.Kb3 Rf2 10.Be4 Re2 11.Ra1
  +/=  (0.54 !)   Depth: 16   00:00:19  53049kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qxc5 Rxc5 5.c3 Bd8 6.h5 Bxa5 7.Kc2 Bb6 8.f4 exf4 9.Rhf1 a5 10.Rxf4 Rf8 11.Kb3 Bd8 12.Rf2 Re8 13.Re2 Rxe2 14.Bxe2 g6 15.Ka4 Kg7
  +/=  (0.46)   Depth: 16   00:00:21  58277kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qxc5 Rxc5 5.c3 Bd8 6.h5 Bxa5 7.Kc2 Bb6 8.f4 exf4 9.Rhf1 a5 10.Rxf4 Rf8 11.Kb3 Rb8 12.Ka4 Bd8 13.Rf2 Bf6 14.Re2 Be5 15.Bf5
  +/=  (0.46)   Depth: 17   00:00:26  75518kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5 4.Qxc5 Rxc5 5.c3 Bd8 6.h5 Bxa5 7.Kc2 Bb6 8.f4 exf4 9.Rhf1 a5 10.Rxf4 Rf8 11.Kb3 Rb8 12.Ka4 Bd8 13.Rf2 Bf6 14.Re2 Be5 15.Bf5
  +/=  (0.46)   Depth: 18   00:00:30  85133kN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5[] 4.Qxc5 Rxc5 5.c3 Rxa5 6.b4 Rb8 7.Kb2 Rc5 8.Kb3 Bf6 9.c4 Rc7 10.Ka4 g6 11.Rb1 Kg7 12.Ka5 Bd8 13.Kxa6 Ra8+ 14.Kb5 Ra3 15.Rbd1 Rb7+ 16.Kc6
  +/=  (0.51)   Depth: 19   00:00:59  173mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5[] 4.Qxc5 Rxc5 5.c3 Rxa5 6.b4 Rb8 7.Kb2 Rc5 8.Kb3 Bf6 9.c4 Rc7 10.Ka4 g6 11.Rb1 Kg7 12.Ka5 Bd8 13.Kxa6 Ra8+ 14.Kb5 Ra3 15.Rbd1
  +/=  (0.51)   Depth: 20   00:01:31  274mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Bd3 Qxc5[] 4.Qxc5 Rxc5 5.c3 Rxa5 6.b4 Rb8 7.Kc2 Ra2+ 8.Kb3 Rf2 9.Rdf1 Rxf1 10.Bxf1 Ra8 11.Bd3 Kf8 12.Ka4 Ke8 13.g5 Kd7 14.gxh6 gxh6 15.h5 Bg5 16.Rg1
  +/=  (0.54)   Depth: 21   00:03:37  681mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.c3 Qxc5 4.Qxc5 Rxc5 5.Rh2 Bd8 6.Rc1 Rb8 7.b4 Rxa5 8.Bc4 g6 9.Kc2 Ra3 10.Kb2 Ra5 11.Kb3 Kg7 12.Rhh1 Rb6 13.Rcd1 Rb8 14.Kc2 Ra3 15.Ra1 Rxa1 16.Rxa1
  +/=  (0.54)   Depth: 22   00:05:46  1053mN, tb=1
1.Qxb4 0-0 2.Nxc5 Rfc8 3.c3 Qxc5 4.Qxc5 Rxc5 5.Bd3 Rxa5 6.b4 Rb8 7.Kc2 Ra2+ 8.Kb3 Rf2 9.Rdf1 Rxf1 10.Bxf1 Ra8 11.Bd3 Kf8 12.Ka4 Bd8 13.c4 Bb6 14.Rf1 Rc8 15.Ra1 Ke7 16.Ra2
  +/=  (0.53)   Depth: 23   00:09:39  1791mN, tb=5
1.Qxb4 0-0 2.Nxc5 Rfc8 3.c3 Qxc5 4.Qxc5 Rxc5 5.Rh2 Rxa5 6.b4 Rb8 7.Kb2 Bf6 8.Kb3 Rc5 9.Bxa6 e4 10.Rc2 exf3 11.h5 f2 12.Rf1 Rxd5 13.Rcxf2 Rg5 14.Rg1 Re5 15.Rf3 d5 16.Bd3
  +/=  (0.52)   Depth: 24   00:52:36  9912mN, tb=911
1.f4 Bf6 2.Nxc5 Qxc5 3.Re1 Kf8 4.Re4 exf4 5.Rxb4 Be5 6.Rc4 Qe3 7.Qxe3 fxe3 8.Re4 e2 9.Bxe2 Rb8 10.b3
  +/=  (0.65 !)   Depth: 24   03:46:44  37732mN, tb=2775

(Cluster,  01.10.2009)
Parent - By Roland Rösler (****) Date 2009-10-01 22:12
Wow! We have 26 pieces on the board and CR shows the first tbs hit after 5'+46". This is really amazing!
Parent - - By chessorpoker (**) Date 2009-10-02 00:44 Edited 2009-10-02 00:54
lukas all i have to say is wow. it took me weeks of thinking to get f4 Bf6 Nxc5 Qxc5 Re1, as the pin dictates the rest of the game. my initial idea here was to go f4 Bf6 g5?! hg5 fg5 Bd8 g6 and destroy the kings shelter but the line wasnt going anywhere. thx for the analysis time on CR, and it must be + 3-400 elo to understand this move f4!

whats even more amazing is that all of a sudden at depth 24 it changes over from Qxb4 to f4, very amazing machine and program you guys made on the rybka team!
Parent - By Banned for Life (Gold) Date 2009-10-02 05:18
Seems like some people may owe you an apology. It took the cluster 03:46:44 to see the advantage of f4, which is one of the deeper moves we've seen here.
Parent - - By Ernst (***) Date 2009-10-02 05:43
Deep Shredder XP, q6600@3200MHz.

14/30  0:01   +0.93    1.Dxb4 Pxb3 2.Da4+ Dd7 3.cxb3 Dxa4
                       4.bxa4 Kd7 5.Ka2 Ta7 6.Tc1 Tb8 7.Tc6 Tab7
                       8.Th2 Tb4 9.Txa6 Txa4+ 10.Kb1 Tb7
                       11.Ld3 (2.976.099) 2326

15/30  0:01   +0.95    1.Dxb4 Pxb3 2.Da4+ Dd7 3.cxb3 Dxa4
                       4.bxa4 O-O 5.Tc1 f5 6.g5 hxg5 7.hxg5 Lxg5
                       8.Tc6 Tf6 9.Kc2 e4 10.Lxa6 Kf7
                       11.Th5 Kg6 (4.310.393) 2361

16/38  0:04   +0.66    1.Dxb4 O-O 2.Pxc5 Tfc8 3.c4 Tcb8
                       4.Da3 dxc5 5.g5 h5 6.Ld3 Tb4 7.g6 fxg6
                       8.Lxg6 Tab8 9.Lxh5 Txb2+ 10.Dxb2 Txb2+
                       11.Kxb2 Dxa5 (9.535.951) 2378

17/42  0:14   +0.33    1.Dxb4 O-O 2.Pxc5 Tfc8 3.Ld3 Dxc5
                       4.Dxc5 Txc5 5.c3 Ta7 6.Kc2 Txa5 7.h5 Kf8
                       8.Ta1 Txd5 9.Txa6 Txa6 10.Lxa6 Lg5
                       11.Ld3 Ke7 12.b4 Le3 13.Te1 (35.366.224) 2361

18/41  0:22   +0.41    1.Dxb4 O-O 2.Pxc5 Tfc8 3.Ld3 Dxc5
                       4.c3 Ld8 5.Dxc5 Txc5 6.b4 Txc3 7.Kb2 Tc7
                       8.Kb3 Kf8 9.Tc1 Tb7 10.Ka4 Tab8
                       11.Tb1 Ta8 12.h5 Lg5 13.b5 axb5+ (53.410.350) 2351

19/41  0:28   +0.41    1.Dxb4 O-O 2.Pxc5 Tfc8 3.Ld3 Dxc5
                       4.c3 Ld8 5.Dxc5 Txc5 6.b4 Txc3 7.Kb2 Tc7
                       8.Kb3 Kf8 9.Tc1 Tb7 10.Ka4 Tab8
                       11.Tb1 Ta8 12.h5 Lg5 13.b5 axb5+ (68.337.151) 2365

20/48  0:44   +0.47    1.Dxb4 O-O 2.Pxc5 Tfc8 3.Ld3 Dxc5
                       4.c3 Ld8 5.Dxc5 Txc5 6.b4 Txc3 7.Kb2 Tc7
                       8.Kb3 Kf8 9.Tc1 Tb7 10.Ka4 Tab8
                       11.Tb1 Ta7 12.Thc1 f5 13.Tc6 fxg4
                       14.fxg4 (105.532.081) 2375

21/51  1:12   +0.47    1.Dxb4 O-O 2.Pxc5 Tfc8 3.Ld3 Dxc5
                       4.c3 De3 5.De4 Dxe4 6.fxe4 Tc5 7.b4 Txc3
                       8.Kb2 Tc7 9.Kb3 Ld8 10.Tc1 Tb7
                       11.Ka4 Tab8 12.Tb1 Ta8 13.Tb2 Kf8
                       14.h5 (172.976.341) 2369

22/56  3:44   +0.36    1.Dxb4 O-O 2.Pxc5 Tfc8 3.Ld3 Dxc5
                       4.c3 Df2 5.De4 Lxh4 6.Tc1 Lg5 7.Tc2 De3
                       8.Dxe3 Lxe3 9.b4 Kf8 10.Ka2 Ke7
                       11.Kb3 Ta7 (501.835.826) 2233

22/60  21:16  +0.50    1.f4 exf4 2.Dxf4 Lf6 3.Th3 Le5
                       4.Dxb4 Tb8 5.Dc4 De7 6.Pxc5 Txb2+
                       7.Kc1 dxc5 8.Te3 Tb4 9.d6 Lb2+
                       10.Kd2 Dxe3+ 11.Kxe3 Txc4 12.Lxc4 Kd7
                       13.Lxa6 Ld4+ (2.962.537.080) 2320

beste zet: f3-f4  tijd: 22:55.273 min  n/s: 2.323.227  CPU 396.1%   n/s(1CPU): 586.525   nodes: 3.194.999.435
Parent - - By Asylum (**) Date 2009-10-02 07:01
Is this with a "Deep Shredder 11" ?
Parent - By Ernst (***) Date 2009-10-02 22:17
A close-to-12 Shredder.
Parent - - By CSullivan (**) Date 2009-10-04 00:37
r3k2r/2q1bpp1/p2p3p/P1nPp3/1p4PP/1N3P2/1PPQ4/1K1R1B1R w kq - 0 1


Analysis by Zappa Mexico II (Q6600 @ 3.1 Ghz, 1 GB hash table):

1.Qxb4 0-0 2.Nxc5 Rfc8 3.Rh3 Qxc5 4.Qxc5 Rxc5 5.f4 exf4 6.b4 Rcc8 7.Rd4 Re8 8.Bd3 g5
  +/=  (0.70)   Depth: 18/55   00:01:08  188mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Rh3 Qxc5 4.Qxc5 Rxc5 5.f4 exf4 6.b4 Rcc8 7.Rd4 Re8 8.Bd3 g5
  +/=  (0.70)   Depth: 18/55   00:01:16  212mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Rh3 Qxc5 4.Qxc5 Rxc5 5.f4 exf4 6.b4 Rcc8 7.Rd4 Re8 8.Bd3 g5
  +/=  (0.70)   Depth: 19/57   00:01:49  300mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Rh3 Qxc5 4.Qxc5 Rxc5 5.f4 exf4 6.b4 Rcc8 7.Rd4 Re8 8.Bd3 g5
  +/=  (0.70)   Depth: 19/57   00:02:11  357mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Qa4 Qxc5 4.Bd3 Rcb8 5.b3 Rb4 6.Qc6 Rc8 7.Qxc5 Rxc5 8.Bxa6 Bf6 9.Be2 Rxa5
  +/=  (0.65)   Depth: 20/59   00:04:09  674mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Qa4 Qxc5 4.Bd3 Rcb8 5.b3 Rb4 6.Qc6 Rc8 7.Qxc5 Rxc5 8.Bxa6 Bf6 9.Be2 Rxa5
  +/=  (0.65)   Depth: 20/59   00:05:36  899mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.g5 Qxc5 4.Qxc5 Rxc5 5.gxh6 gxh6 6.f4 Bf6 7.fxe5 Bxe5 8.Rg1+ Kf8 9.c3 Rxa5 10.Kc2 Ra4
  +/=  (0.57)   Depth: 21/61   00:10:13  1628mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.g5 Qxc5 4.Qxc5 Rxc5 5.gxh6 gxh6 6.f4 Bf6 7.fxe5 Bxe5 8.Rg1+ Kf8 9.c3 Rxa5 10.Kc2 Ra4
  +/=  (0.57)   Depth: 21/61   00:15:36  2442mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Qa4 Qxc5 4.Bd3 Rcb8 5.b3 Rb4 6.Qc6 Rc8 7.Qxc5 Rxc5 8.Be4 Rxa5 9.Kb2 Rc5 10.c3 Rb8
  +/=  (0.48)   Depth: 22/63   00:28:52  4500mN
1.f4
  +/=  (0.49)   Depth: 22/63   00:58:59  9039mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Qa4 Qxc5 4.Bd3 Rcb8 5.b3 Rb4 6.Qc6 Rc8 7.Qxc5 Rxc5 8.Be4 Rxa5 9.Kb2 Rc5 10.c3 Rb8
  +/=  (0.48)   Depth: 22/63   01:16:28  11728mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Rh3 Rab8 4.Qa3 Qxc5 5.Qxc5 Rxc5 6.Bxa6 Rxa5 7.Be2 Rba8 8.c3 Bd8 9.Rhh1 Kf8 10.Kc2 Bb6 11.Bc4 R5a7
  +/=  (0.47)   Depth: 23/65   01:57:40  18202mN
1.f4
  +/=  (0.48)   Depth: 23/65   02:31:45  23431mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Rh3 Rab8 4.Qa3 Qxc5 5.Qxc5 Rxc5 6.Bxa6 Rxa5 7.Be2 Rba8 8.c3 Bd8 9.Rhh1 Kf8 10.Kc2 Bb6 11.Bc4
  +/=  (0.47)   Depth: 23/65   03:19:29  30849mN
1.Qxb4 0-0 2.Nxc5 Rfc8 3.Rh3 Rab8 4.Qa3 Qxc5 5.Qxc5 Rxc5 6.Bxa6 Rxa5 7.Be2 Rba8 8.c3 Bd8 9.Rhh1 Kf8 10.Kc2 Bb6 11.Bc4 Ke7 12.b4
  +/=  (0.47)   Depth: 24/67   03:46:39  35199mN
1.f4
  +/=  (0.48)   Depth: 24/67   05:31:37  49182mN
1.f4 Bf6 2.Rh3 exf4 3.Qxf4 Nxb3 4.Rxb3 0-0 5.g5 Be5 6.Qxb4 Rfb8 7.Qd2 h5 8.Re1 Rxb3 9.cxb3 Qb7 10.Bc4 Rb8 11.Ka1
  +/=  (0.54)   Depth: 24/67   07:15:51  65135mN
1.f4 Bf6 2.Rh3 exf4 3.Qxf4 Nxb3 4.Rxb3 0-0 5.g5 Be5 6.Qxb4 Rfb8 7.Qd2 h5 8.Re1 Rxb3 9.cxb3 Qb7 10.Bc4 Rb8 11.Ka1
  +/=  (0.54)   Depth: 24/67   07:37:01  68537mN
Parent - By CSullivan (**) Date 2009-10-04 11:54
Zappa II Mexico, after 25 iterations and 13:41:22, gives the same line (1.f4 Bf6 2.Rh3, etc.) as +0.50.
Parent - - By Mark (****) Date 2009-10-02 18:41
After 7 hours, 4 billion nodes, and 25 ply, R3 still has Qxb4 as the best move.
Parent - - By chessorpoker (**) Date 2009-10-02 21:01
i really didn't think that the cluster would recognize 1. f4!, but it sure did and i'm pleasantly surprised. apparently shredder sees it without too much difficulty which may be a factor of its "king attacks", i'm definitely going to analyze with this engine more often. i wish we knew more about what R3 and CR was thinking about and why.

also lukas if you would like this opening with all the analysis for the rybka teams opening book i will give it to you guys.
Parent - - By Mark (****) Date 2009-10-02 21:15
I'll let R3 continue to chew on this overnight to see if it gets to 26 ply.
Parent - - By Mark (****) Date 2009-10-03 13:00
No luck after 27 ply, 22+ hours, and 12.7 billion nodes...

1. Qxb4 O-O 2. Nxc5 Rfc8 3. c3 Qxc5 4. Qxc5 Rxc5 5. h5 Bg5 6. Rh2 g6 7. hxg6
fxg6 8. Be2 Bf4 9. Rg2 Rxa5 10. b4
Parent - - By Highendman (****) Date 2009-10-03 13:20
So two very interesting conclusions:

IDEA found the best move, while IA couldn't, on the same H/W. Right?
Cluster Rybka with Lukas private build is amazingly strong. Finding the best move in about 3 minutes.
Parent - - By Lukas Cimiotti (Bronze) Date 2009-10-03 14:08
it was more than 3 hours ;)
Parent - By Highendman (****) Date 2009-10-03 14:10
Well, I won't let the facts get in the way of a good story :)
Parent - By Roland Rösler (****) Date 2009-10-03 20:47
It´s a pity you stopped immediately after the line was shown. I´m sure you would have got another line at depths 24 with better eval. You can see it at the exclamation mark after the eval (here: (0.65 !)). But I can understand your hurry! It was late and 3.75 hours are very much yet.

1.f4 Bf6 2.Nxc5 Qxc5 3.Re1 Kf8 4.Re4 exf4 5.Rxb4 Be5 6.Rc4 Qe3 7.Qxe3 fxe3 8.Re4 e2 9.Bxe2 Rb8 10.b3
  +/=  (0.65 !)   Depth: 24   03:46:44  37732mN, tb=2775


PS: The most astonishing move in this line is 10. b3!?. R3 loves b4.
Parent - - By Mark (****) Date 2009-10-03 16:28
Right, IDeA did find the best move, however, I did have some relatively minor interactivity.  I'll try it with IDeA again with no interactivity for a day and see what happens.  (I'll delete the infinite_analysis tree to make sure IDeA doesn't get any info from that.)
Parent - - By Highendman (****) Date 2009-10-03 16:32
Thanks. This position seems like a very good test position to demonstrate the merits of IDEA vs. IA.
Now that we have the verdict of Lukas' Cluster on what the best move is. I'm very curious to see what you find out.
Up Topic Rybka Support & Discussion / Rybka Discussion / what does the cluster think the best move is here?
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill