Rybka 3: The dark truth behind the hype

[Cubeman]

Alburt, It is very important to know approximately (atleast) how the evaluations were going, and which moves of Fritz were mistakes.
After the exchange of queens was it already a mating attack?


I won't try to tell you how much better this is over Rybka 232a. Suffice it to say this is a performance of a full +308 Elo over Fritz 11. Seventy-five wins to four... It kind of boggles the mind. If this commentary sounds like hyperbole, let me ask you how you would describe it.

Here is a cute game it played, where a seemingly innocuous queenless middlegame (at least to poor innocent me... ) quickly turned into a mate attack:

1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 e6 5. e3 Nbd7 6. Bd3 dxc4 7. Bxc4 b5 8. Bd3 Bb7 9. O-O {end of book}

[D]r2qkb1r/pb1n1ppp/2p1pn2/1p6/3P4/2NBPN2/PP3PPP/R1BQ1RK1 b kq - 0 9

This is one of the positions from the Silver Suite. There are several acceptable continuations. Each engine plays it as both white and black.

9...a6 10. e4 c5 11. d5 c4 12. Bc2 Qc7 13. Nd4 Nc5 14. Qe2 b4 15. Na4 O-O-O 16. Nc6 Bxc6 17. dxc6 Qxc6 18. Qxc4 Qb5

[D]2kr1b1r/5ppp/p3pn2/1qn5/NpQ1P3/8/PPB2PPP/R1B2RK1 w - - 0 19

Although I can find the attacking moves that followed the queen exchange with other engines, none of them chose to exchange queens here as Rybka 3 did.

This kind of brought to mind a comment by the late Rudolph Spielman, a great tactician, who said that he was perfectly capable of finding every single combination Alekhine found, but as to setting up those positions that led to the combinations....

19. Qxb5 axb5 20. Nxc5 Bxc5

[D]2kr3r/5ppp/4pn2/1pb5/1p2P3/8/PPB2PPP/R1B2RK1 w - - 0 21


[D]3r3r/5ppp/1k2pn2/1P6/1b6/8/2B2PPP/R1B2RK1 w - - 0 27

27. Ra6+! Kxb5 28. Ra7! Kc4 29. Rc7+ Bc5

[D]3r3r/2R2ppp/4pn2/2b5/2k5/8/2B2PPP/2B2RK1 w - - 0 30

and here is a challenging (winning) move for engines that Rybka 3 finds in about 1 second

30. Bd1! Rd7 31. Be2+ Kb4 32. Rc6 Ne4 33. Be3 Ka3 34. Bxc5+ Nxc5 35. Rb6 Nb3 36. Bc4 Kb2 37. Bxb3 1-0

When I first played thru this game I was astounded that there was a absolute mating attack after the Queen exchange.And the move that interested me the most was the quiet 30.Bd1!And shortly afterwards Fritz had to give up a piece to delay mate.
But closer inspection shows that Fritz 30...Rd7 is mistake but understandable considering the time control of 3 min +2 secs for game.But 30...e5! seems to hold out a bit more resistance and my computer with Toga only shows a modest +0.83 at D=15.
But still 30.Bd1 is very good move and is considered best by Toga too but just does not lead to forced mate.

[Jeroen]

Hi Albert,

Thanks! Yesterday I also tried a small match between F11 and R3 32-bit single (3 mins blitz) and after 50 games F11 had won only 1 game.

Kind regards, Jeroen

[ernst]

Yes very good question and I will try again. I wonder how much hash was used by the others.

I wonder why your Rybka 3 insists on finding this in ply 19, while Rybka 3 of ANSARI and Albert find it in 17 ply.

I mostly wonder about why Rybka 3 single CPU on Albert's 2.2GHz Athlon, finds this in about 7.5 minutes while your Rybka 3 MP on 4 cores on Q6600 makes more than 10 minutes to find.

Perhaps you are using different versions or else i can't find another reason.Have you used for this analysis the latest version of Rybka 3? Or one older beta?
The same question to Albert, if he used the latest Rybka 3 for this analysis?

Since your results if you use same Rybka 3 make no sense.

[ernst]

OK it is clear to me that hash size is the cause of the discrepancy. With 128MB hash I get 17 ply as well.

Code: Select all

New game - Rybka 3, Blitz:16'
1B1r4/rp2npkp/2b1pbp1/1qp5/nPN1R3/1P1P1QP1/2P2PBP/5R1K w - - 0 1

Analysis by Rybka 3:

1.Bxa7 
  +/-  (0.93 !)   Depth: 6   00:00:00  5kN
1.Bxa7 Bxe4 2.Qxe4 Nc3 3.Qxb7 Qxb7 4.Bxb7 Rd7 5.Na5 cxb4 6.Bc5 Ned5 7.Kg2 Rc7 8.Bd6 
  +/-  (0.86)   Depth: 6   00:00:00  10kN
1.Bxa7 Bxe4 2.Qxe4 Nc3 3.Qxb7 Qxb7 4.Bxb7 Rd7 5.Na5 cxb4 6.Bc5 Ned5 7.Kg2 Rc7 8.Bd6 
  +/-  (0.86)   Depth: 7   00:00:00  13kN
1.Bxa7 Bxe4 2.Qxe4 Nc3 3.Qxb7 Qxb7 4.Bxb7 Rd7 5.Na5 cxb4 6.Bc5 Ned5 7.Bxd5 Nxd5 8.Nc4 Rc7 9.Be3 
  +/-  (0.77)   Depth: 8   00:00:00  27kN
1.Bxa7 
  +/-  (0.97 !)   Depth: 9   00:00:00  60kN
1.Bxa7 
  +/-  (1.17 !)   Depth: 9   00:00:00  65kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.Ne3 Nb6 4.bxc5 
  +-  (1.51)   Depth: 9   00:00:00  98kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.Ne3 Nb6 4.bxc5 Qxc5 5.Ng4 Ng8 6.Qe3 Qxe3 7.Nxe3 Nc8 8.Bb8 Rd2 
  +/-  (1.31)   Depth: 10   00:00:00  225kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.Ne3 Nb6 4.bxc5 Qxc5 5.Ng4 Ng8 6.Qe3 Qxe3 7.Nxe3 Nc8 8.Bb8 Rd2 
  +/-  (1.31)   Depth: 11   00:00:00  243kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.bxc5 Nxc5 4.Qe3 Ne6 5.Bb6 Ra8 6.Qd2 Nd4 
  +/-  (1.21)   Depth: 12   00:00:01  504kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.bxc5 Nxc5 4.Qe3 Ne6 5.Bb6 Ra8 6.Qd2 Ra2 7.Be3 Nd4 
  +/-  (1.14)   Depth: 13   00:00:03  1000kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.bxc5 Nxc5 4.Qc3 Ne6 5.Bb6 Rd7 6.Ra1 Nc6 7.Bh3 Ncd4 8.Kg2 Qc6 9.Qd3 
  +/-  (1.20)   Depth: 14   00:00:07  2195kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.bxc5 Nxc5 4.Rd1 Rxd1+ 5.Qxd1 Nc8 6.Bb8 Na6 7.Nd6 Nxd6 8.Bxd6 Nc5 9.Kg1 Qb4 10.h4 Qc3 
  +/-  (1.17)   Depth: 15   00:00:17  5232kN
1.Bxa7 Bxe4 2.dxe4[] e5[] 3.bxc5 Nxc5 4.Qc3 Ne6 5.Bb6 Rc8 6.Be3 Rd8 7.Ra1 Nc8 8.Qa5 Qxa5 9.Nxa5 Rd7 10.Kg1 Rc7 11.Ra2 
  +/-  (1.18)   Depth: 16   00:00:40  12324kN
1.Bxa7 Bxe4 2.dxe4[] e5[] 3.bxc5 Nxc5 4.Qc3 Ne6 5.Bb6 Rc8 6.Be3 Rd8 7.Ra1 Nc8 8.Qa5 Qxa5 9.Nxa5 Rd7 10.Kg1 Rc7 11.Ra2 
  +/-  (1.18)   Depth: 17   00:01:13  24170kN
1.Qxf6+ 
  +/-  (1.38 !)   Depth: 17   00:02:04  40564kN
1.Qxf6+ 
  +-  (1.58 !)   Depth: 17   00:02:14  42431kN
1.Qxf6+ 
  +-  (1.98 !)   Depth: 17   00:02:25  44543kN
1.Qxf6+ 
  +-  (2.78 !)   Depth: 17   00:02:44  48291kN
1.Qxf6+ 
  +-  (4.38 !)   Depth: 17   00:04:39  75252kN

(,  31.07.2008)

[George Tsavdaris]

OK it is clear to me that hash size is the cause of the discrepancy. With 128MB hash I get 17 ply as well.

Code: Select all

New game - Rybka 3, Blitz:16'
1B1r4/rp2npkp/2b1pbp1/1qp5/nPN1R3/1P1P1QP1/2P2PBP/5R1K w - - 0 1

Analysis by Rybka 3:

  +/-  (1.21)   Depth: 12   00:00:01  504kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.bxc5 Nxc5 4.Qe3 Ne6 5.Bb6 Ra8 6.Qd2 Ra2 7.Be3 Nd4 
  +/-  (1.14)   Depth: 13   00:00:03  1000kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.bxc5 Nxc5 4.Qc3 Ne6 5.Bb6 Rd7 6.Ra1 Nc6 7.Bh3 Ncd4 8.Kg2 Qc6 9.Qd3 
  +/-  (1.20)   Depth: 14   00:00:07  2195kN
1.Bxa7 Bxe4 2.dxe4[] e5 3.bxc5 Nxc5 4.Rd1 Rxd1+ 5.Qxd1 Nc8 6.Bb8 Na6 7.Nd6 Nxd6 8.Bxd6 Nc5 9.Kg1 Qb4 10.h4 Qc3 
  +/-  (1.17)   Depth: 15   00:00:17  5232kN
1.Bxa7 Bxe4 2.dxe4[] e5[] 3.bxc5 Nxc5 4.Qc3 Ne6 5.Bb6 Rc8 6.Be3 Rd8 7.Ra1 Nc8 8.Qa5 Qxa5 9.Nxa5 Rd7 10.Kg1 Rc7 11.Ra2 
  +/-  (1.18)   Depth: 16   00:00:40  12324kN
1.Bxa7 Bxe4 2.dxe4[] e5[] 3.bxc5 Nxc5 4.Qc3 Ne6 5.Bb6 Rc8 6.Be3 Rd8 7.Ra1 Nc8 8.Qa5 Qxa5 9.Nxa5 Rd7 10.Kg1 Rc7 11.Ra2 
  +/-  (1.18)   Depth: 17   00:01:13  24170kN
1.Qxf6+ 
  +/-  (1.38 !)   Depth: 17   00:02:04  40564kN
1.Qxf6+ 
  +-  (1.58 !)   Depth: 17   00:02:14  42431kN
1.Qxf6+ 
  +-  (1.98 !)   Depth: 17   00:02:25  44543kN
1.Qxf6+ 
  +-  (2.78 !)   Depth: 17   00:02:44  48291kN
1.Qxf6+ 
  +-  (4.38 !)   Depth: 17   00:04:39  75252kN

OK many thanks!! Your solution time also makes much more sense now.

Although i wonder now:
Does someone should use low hash for testpositions with R3?

Also are you sure that this result does not come because of a learning function of Rybka? I heard something about persistent hash or whatever. So is there any chance Rybka 3 to remember the position from your previous analysis so to solve it more quickly now?
I guess then would solve it instantly so actually you didn't use any kind of learning, right?

[ernst]

OK many thanks!! Your solution time also makes much more sense now.

Although i wonder now:
Does someone should use low hash for testpositions with R3?

Also are you sure that this result does not come because of a learning function of Rybka? I heard something about persistent hash or whatever. So is there any chance Rybka 3 to remember the position from your previous analysis so to solve it more quickly now?
I guess then would solve it instantly so actually you didn't use any kind of learning, right?

Don't worry, persistent hash is disabled and I can reproduce the 19 ply when setting the hash size to 512MB again.

[Albert Silver]

OK many thanks!! Your solution time also makes much more sense now.

Although i wonder now:
Does someone should use low hash for testpositions with R3?

Also are you sure that this result does not come because of a learning function of Rybka? I heard something about persistent hash or whatever. So is there any chance Rybka 3 to remember the position from your previous analysis so to solve it more quickly now?
I guess then would solve it instantly so actually you didn't use any kind of learning, right?

Don't worry, persistent hash is disabled and I can reproduce the 19 ply when setting the hash size to 512MB again.

Yes, I re-checked and the solution was not with 512MB as in the post (sorry) but 256.

Albert

[Albert Silver]

Hi Albert,

Thanks! Yesterday I also tried a small match between F11 and R3 32-bit single (3 mins blitz) and after 50 games F11 had won only 1 game.

Kind regards, Jeroen

See? It's not just me...

Albert

[Nimzovik]

YAWN!! Yes.......... A new and improved Rybka. I laud the the beast! TRULY I do... a new analysis tool for me........ BUT!!!!!!!!!! Does it PLAY CLOSED POSITIONS!! At LEAST have a pre-progrmmed setting that does so...... There are positional players in the chess world that feel somewhat left out in the cold..........

[TSP]

Yes, chess consists in a number of parts ( Opening theory, tactics, endgame, etc.), which includes closed positions. Along with ground held within the endgame, this may be mankind's last bastion of territory held over the cold blooded, cerebral shredding villain.

Unfortunately for us carbon based creatures, time is not on our side either. The tiny island we build our mettle on is dwindling steadily, while the beast's ocean roar rages ever louder at the door.