Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / After this endgame example I rest my case :-)
- - By Ernst (***) Date 2008-08-12 12:33
Dead draw, but not to Rybka 3 mp.

2430: Cchess,R - Gcovn1,F, Rated game, 10m + 0s 2008
8/6P1/4k3/4p3/R7/4KP2/6r1/8 w - - 0 1


Analysis by Rybka 3:

48.Ra7 Rg1 49.Rc7
  +-  (2.19)   Depth: 6   00:00:00  0kN
48.Ke4 Re2+ 49.Kd3 Rg2 50.Ra7 Rg1
  +-  (2.32)   Depth: 6   00:00:00  1kN
48.Ke4 Re2+
  +-  (2.28)   Depth: 7   00:00:00  1kN
48.Rg4 Rxg4 49.fxg4[] Kf7 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4 Kg7 55.Kf5
  +-  (2.67)   Depth: 8   00:00:00  1kN
48.Rg4 Rxg4 49.fxg4[] Kf7 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4 Kg7 55.Kf5
  +-  (2.67)   Depth: 9   00:00:00  1kN
48.Rg4 Rxg4 49.fxg4[] Kf7 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4 Kg7 55.Kf5
  +-  (2.67)   Depth: 10   00:00:00  1kN
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4[] Kf6 53.g5+ Kg6 54.Kg4 Kg7 55.Kf5
  +-  (3.07)   Depth: 13   00:00:00  2kN
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4[] Kf6 53.g5+ Kg6 54.Kg4 Kg7 55.Kf5
  +-  (3.07)   Depth: 14   00:00:00  2kN
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4[] Kf6 53.g5+[] Kg6 54.Kg4[] Kg7 55.Kf5 Kf7 56.g6+ Kg8 57.Kf4 Kf8 58.Kg4 Kg8 59.Kf5 Kg7 60.Kg5
  +-  (3.07)   Depth: 15   00:00:00  13kN
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4[] Kf6 53.g5+[] Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 16   00:00:00  17kN
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 17   00:00:00  27kN, tb=2
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 18   00:00:00  37kN, tb=3
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 19   00:00:00  53kN, tb=7
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 20   00:00:00  76kN, tb=16
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 21   00:00:01  113kN, tb=38
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 22   00:00:01  166kN, tb=85
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 23   00:00:02  256kN, tb=182
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 24   00:00:03  377kN, tb=342
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 25   00:00:04  545kN, tb=627
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 26   00:00:07  900kN, tb=1187
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 27   00:00:10  1261kN, tb=1735
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 28   00:00:14  1822kN, tb=2754
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 29   00:00:21  2699kN, tb=4699
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 30   00:00:30  3865kN, tb=7611
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 31   00:00:44  5730kN, tb=11945
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 32   00:01:00  8125kN, tb=18148
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 33   00:01:37  14258kN, tb=29623
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 34   00:02:22  21219kN, tb=45865
48.Rg4 Rxg4[] 49.fxg4[] Kf7[] 50.Ke4 Kxg7 51.Kxe5 Kg6[] 52.Kf4 Kf6[] 53.g5+ Kg6 54.Kg4[] Kg7 55.Kf5
  +-  (3.07)   Depth: 35   00:03:45  35158kN, tb=71928

(,  12.08.2008)
Parent - - By Arkansaw (***) Date 2008-08-12 12:46
how is it a dead draw?
Parent - By Kappatoo (*****) Date 2008-08-12 12:54
Just follow the main line.
Parent - - By bnc (***) Date 2008-08-12 13:17 Edited 2008-08-12 13:53
8/6P1/4k3/4p3/R7/4KP2/6r1/8 w - - 0 1


The position after Rg4 Rxg4 2.fxg4 Kf7 is a simple draw

This is another example of the MP EGTB bug

On my quad QX6700 with 6 man EGTB's Rybka 3 (default) with all 4 cpus active gives this incorrect analysis :

1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kg8 4.Kxe5 Kxg7 5.g5 Kg6 6.Kf4 Kh7 7.Kg3 Kg7 8.Kf3 Kf7 9.Kf4
  +-  (3.33)   Depth: 24   00:00:04  413kN, tb=238
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kg8 4.Kxe5 Kxg7 5.Kf5 Kf7 6.g5 Kg7 7.Ke5 Kg6
  +-  (3.33)   Depth: 25   00:00:06  614kN, tb=406
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kg8 4.Kxe5 Kxg7 5.Kf5 Kf7 6.g5 Kg7 7.Ke5 Kg6
  +-  (3.33)   Depth: 26   00:00:09  871kN, tb=635

Now if I change the Max CPU's engine parameter to 1 then Rybka 3's analysis looks much better :

1.Ra6+ Kf7 2.Ra7+ Kf6
  =  (0.00)   Depth: 21   00:00:02  95kN, tb=544
1.Ra6+ Kf7 2.Ra7+ Kf6
  =  (0.00)   Depth: 22   00:00:04  127kN, tb=837
1.Ra6+ Kf7 2.Ra7+ Kf6 3.Ke4 Re2+
  =  (0.00)   Depth: 23   00:00:06  156kN, tb=1136

This bug also exists in Rybka 2.3.2a. My impression is that EGTB bugs will not get fixed due to the view that they provide minimal if any ELO gain.  I hope I am wrong
Parent - By sylvain.renard.27 (**) Date 2008-08-12 14:02

> This is another example of the MP EGTB bug


You are right! Something I do not understand (Shredder Classic interface):

I load Rybka 3, Max CPU's = 2048, incorrect analysis
Now I set Max CPU's = 1, the analysis is now correct.
I set Max CPU's = 2048 again, I clear hash tables, and now I have a correct analysis. I have tried several times, clearing hash tables, now analysis always shows a draw 0.00.
  HELP!!!
Parent - - By M ANSARI (*****) Date 2008-08-12 13:01
This bug is only with the MP version ... the SP version sees it is a draw.  This is the same bug that has been considered the "blind bishop" bug where R3 MP somehow fumbles the data from EGTB's and its own data produced in hash.
Parent - By Juergen Faas (**) Date 2008-08-12 13:08

> ... the SP version sees it is a draw.


Well, not mine ;)
Parent - - By Charles (**) Date 2008-08-12 13:37 Edited 2008-08-12 13:39

> This bug is only with the MP version ... the SP version sees it is a draw.


If you are speaking of the Rybka 3 version how could the bug only be with the MP version since they are both the same. The only difference in them being that the SP version is hard coded for one cpu.

I ask the above as one very unknowledgeable of the inner workings of chess engines. No need to give an elaborate explanation. I'm just curious really if you are talking about the 3rd version. Thanks

Edit:

O.K., Forget my question above. I missed this!!

>Now if change the Max CPU's engine parameter to 1 then Rybka 3's analysis looks much better :

Parent - - By lkaufman (*****) Date 2008-08-12 13:41
I think he means only the SP version OR the MP version set to use only 1 processor. It's the number of processors used that seems to matter for this bug; the more cores used, the more often it happens.
Parent - - By Dragon Mist (****) Date 2008-08-12 13:43
This bug has 100% probability of occurence? (position and everything else set for this to happen?)
Parent - By lkaufman (*****) Date 2008-08-12 15:09
With MP, nothing is certain. My experience so far suggests that the bug just is more likely to show up in these situations with more cores used, and probably with eight cores it may happen often enough to round to 100%. Frequency of occurrence may also be position-dependent.
Parent - By M ANSARI (*****) Date 2008-08-12 18:02
This is analysis from both MP and SP ... obviously there is a bug in the MP version's handling of EGTB's.

New game
8/6P1/4k3/4p3/R7/4KP2/6r1/8 w - - 0 1


Analysis by Rybka 3 1-cpu :

1.Rg4
  +-  (2.43)   Depth: 3   00:00:00
1.Rg4
  +-  (1.90)   Depth: 3   00:00:00
1.Rg4
  +-  (2.10)   Depth: 4   00:00:00
1.Rg4
  +-  (2.18)   Depth: 4   00:00:00
1.Rg4 Rxg4
  +-  (2.06)   Depth: 5   00:00:00
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6
  +-  (1.74)   Depth: 6   00:00:00  1kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6
  +-  (1.74)   Depth: 6   00:00:00  1kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6
  +-  (1.74)   Depth: 6   00:00:00  1kN
1.Ra7 Kf6 2.Ke4
  ±  (0.94)   Depth: 7   00:00:00  5kN
1.Ra7 Kf6
  ²  (0.34)   Depth: 8   00:00:00  17kN, tb=1
1.Ra8 Rxg7 2.Ke4 Rg5 3.Ra6+ Kf7 4.Rc6
  ²  (0.37)   Depth: 8   00:00:00  22kN, tb=1
1.Ke4 Kf6
  ²  (0.53)   Depth: 8   00:00:00  24kN, tb=1
1.Ke4 Kf6 2.Ra2 Rxg7 3.Ra6+ Kg5 4.Re6 Rc7 5.Rxe5+ Kf6
  ²  (0.50)   Depth: 9   00:00:00  26kN, tb=1
1.Ke4 Kf6 2.Ra2 Rxg7 3.Ra6+ Kg5 4.Re6 Rc7 5.Rxe5+ Kf6 6.Rd5
  ²  (0.50)   Depth: 10   00:00:00  30kN, tb=6
1.Ke4 Kf6 2.Ra2 Rxg7 3.Ra6+ Kg5 4.Re6 Rc7 5.Re8 Kf6 6.Rxe5 Rc4+
  ²  (0.34)   Depth: 11   00:00:00  46kN, tb=18
1.Ke4 Kf6 2.Ra6+ Kxg7 3.Kf5 Re2 4.Ra3 Kh6 5.Rd3 Kh7 6.Rc3
  ²  (0.34)   Depth: 12   00:00:00  78kN, tb=46
1.Ke4 Re2+ 2.Kd3 Rg2
  ²  (0.26)   Depth: 13   00:00:00  105kN, tb=77
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6
  =  (0.21)   Depth: 14   00:00:01  187kN, tb=160
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6
  =  (0.16)   Depth: 15   00:00:01  254kN, tb=258
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.10)   Depth: 16   00:00:02  393kN, tb=449
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ra2 Rxg7 5.Ke4 Rg1 6.Ra6+ Kg5
  =  (0.08)   Depth: 17   00:00:03  585kN, tb=805
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.07)   Depth: 18   00:00:05  833kN, tb=1319
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.04)   Depth: 19   00:00:10  1209kN, tb=2119
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.04)   Depth: 20   00:00:13  1473kN, tb=2956
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.00)   Depth: 21   00:00:22  2273kN, tb=4656
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.00)   Depth: 22   00:00:31  2969kN, tb=7124
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.00)   Depth: 23   00:00:39  3652kN, tb=10111
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.00)   Depth: 24   00:00:55  4813kN, tb=15231
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.00)   Depth: 25   00:01:15  6404kN, tb=21850
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6
  =  (0.00)   Depth: 26   00:01:40  8453kN, tb=30868
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+
  =  (0.00)   Depth: 27   00:02:09  10906kN, tb=40818

(,  12.08.2008)

This is with the MP version ... eventually it will see a draw ... but only by brute force and Depth 40

New game
8/6P1/4k3/4p3/R7/4KP2/6r1/8 w - - 0 1


Analysis by Rybka 3 :

1.Rg4 Rxg4
  +-  (2.06)   Depth: 5   00:00:00
1.Rg4 Rxg4
  +-  (2.06)   Depth: 5   00:00:00
1.Rg4 Rxg4
  +-  (2.06)   Depth: 5   00:00:00
1.Rg4 Rxg4
  +-  (2.06)   Depth: 5   00:00:00
1.Rg4 Rxg4
  +-  (2.06)   Depth: 5   00:00:00
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4
  +-  (2.54)   Depth: 9   00:00:00  5kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4
  +-  (2.54)   Depth: 9   00:00:00  5kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4
  +-  (2.54)   Depth: 9   00:00:00  5kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4
  +-  (2.54)   Depth: 9   00:00:00  5kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4
  +-  (2.54)   Depth: 9   00:00:00  5kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4
  +-  (2.54)   Depth: 9   00:00:00  5kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4
  +-  (2.54)   Depth: 9   00:00:00  7kN
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 10   00:00:00  11kN, tb=2
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 10   00:00:00  11kN, tb=2
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 11   00:00:00  11kN, tb=2
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 12   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 19   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 19   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 19   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 19   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 19   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 19   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 19   00:00:00  73kN, tb=26
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 20   00:00:01  352kN, tb=270
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 21   00:00:01  352kN, tb=270
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 22   00:00:01  392kN, tb=295
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 23   00:00:01  436kN, tb=330
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 24   00:00:01  499kN, tb=387
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 25   00:00:02  600kN, tb=467
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 26   00:00:03  803kN, tb=648
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 27   00:00:05  1087kN, tb=916
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 28   00:00:06  1477kN, tb=1282
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 29   00:00:10  2109kN, tb=2001
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 30   00:00:15  3055kN, tb=3060
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 31   00:00:24  4738kN, tb=4865
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 32   00:00:34  6797kN, tb=6913
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 33   00:00:53  10657kN, tb=10394
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 34   00:01:19  16009kN, tb=14969
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 35   00:02:12  27164kN, tb=23351
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 36   00:03:33  45019kN, tb=34849
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 37   00:06:01  77467kN, tb=56339
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5 Kg6 5.Kf4 Kf6 6.g5+ Kg6 7.Kg4 Kg7 8.Kf5
  +-  (2.74)   Depth: 38   00:09:00  115mN, tb=81822
1.Rg4 Rxg4
  +-  (2.74)   Depth: 39   00:28:48  423mN, tb=253055
1.Rg4 Rxg4 2.fxg4 Kf7
  =  (0.00)   Depth: 40   02:30:57  3547mN, tb=2042331

(,  12.08.2008)
Parent - By sylvain.renard.27 (**) Date 2008-08-13 12:35

> This bug is only with the MP version ... the SP version sees it is a draw.  This is the same bug that has been considered the "blind bishop" bug where R3 MP somehow fumbles the data from EGTB's and its own data produced in hash.


When I use Aquarium with Rybka 3, I cannot reproduce the bug with MP version!!!
As I already said, with Aquarium, infinite analysis does not start after initialiazing many TB's. You stop and run again, everything is ok.
With Shredder and Fritz GUI's, analysis starts after initialiazing many TB's, but it is wrong!!!  (it seems that Rybka cannot collect all TB's informations)
So I am wondering: does the bug comes from Microsoft Windows???
Parent - By sylvain.renard.27 (**) Date 2008-08-12 13:33

> Dead draw, but not to Rybka 3 mp.


Hello,
are you using only 3-5 men Tablebases?

Here is the Analysis by Rybka 3, with some 6 men TBs, obvious draw for her with the help of these TBs :

16.00   0:01    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (64.971) 47  TB:97
17.00   0:02    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (72.224) 31  TB:124
18.01   0:02    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (84.930) 31  TB:162
19.01   0:03    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (99.042) 28  TB:207
20.01   0:04    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (118.380) 26  TB:257
21.01   0:06    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (141.700) 22  TB:372
22.01   0:09    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (165.369) 18  TB:529
23.01   0:13    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (213.149) 16  TB:763
24.01   0:16    0.00   1.Ke4 Kf6 2.Ra7 Re2+ (254.959) 15  TB:1.143
25.01   0:21    0.00   1.Ke4 Kf6 2.Ra7 Re2+ 3.Kd5 Rd2+ 4.Kc6 Rg2 5.Kd6 Rd2+ 6.Kc5 Rg2 (366.923) 17  TB:1.584
26.01   0:27    0.00   1.Ke4 Kf6 2.Ra7 Re2+ 3.Kd5 Rd2+ 4.Kc6 Rg2 5.Kd6 Rd2+ 6.Kc5 Rg2 (450.614) 16  TB:2.195
27.01   0:43    0.00   1.Ke4 Kf6 2.Ra7 Re2+ 3.Kd5 Rd2+ 4.Kc6 Rg2 5.Kd6 Rd2+ 6.Kc5 Rg2 (612.266) 14  TB:3.123
28.01   3:53    0.00   1.Ke4 Kf6 2.Ra7 Re2+ 3.Kd5 Rd2+ 4.Kc6 Rg2 5.Kd6 Rd2+ 6.Kc5 Rg2 (792.922) 3  TB:3.824

If you don't want to have many 6 men TBs, I think these ones are very important, because rooks and pawns endgames occur frequently:
krpkrp; krppkr, kppkpp; kpppkp.
I put krpkrp, krppkr and all 3-5 men TBs on a 16 GB USB key, other ones on hard disk.
  Best regards,
    Sylvain
Parent - By Permanent Brain (*****) Date 2008-08-12 17:33
For comparison, results from older versions. These were with the related rook vs. rook 5-piece tablebases (but not will all 5p tables), D945 3.4 GHz:

Analysis by Rybka 2.2n2 mp 32-bit an-ftb:

1.Ra7 Kf6 2.Ke4 Re2+ 3.Kd5 Rg2 4.Rd7 Rxg7 5.Rxg7
  +/-  (0.71)   Depth: 2   00:00:00
  +/-  (0.71)   Depth: 3   00:00:00
1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7
  +/-  (0.83)   Depth: 3   00:00:00
  =  (0.00)   Depth: 4   00:00:00
1.Ra7 Kf6 2.Ke4 Re2+
  +/-  (0.71)   Depth: 4   00:00:00
  +/=  (0.38)   Depth: 7   00:00:00  3kN
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra6+ Kf5 4.Ra7 Rg3 5.Kc4
  +/=  (0.54)   Depth: 7   00:00:00  5kN
  =  (0.00)   Depth: 25   00:00:58  9978kN

This is from a setting with frequent tbs. accesses (from USB stick). I stopped after 0.00 appeared. The tbs. numbers are missing here because 2.2n2 doesn't display them.

Analysis by Rybka 2.3.2a 32-bit (single/default):

1.Rg4 Rxg4 2.fxg4 Kf7 3.Ke4 Kxg7 4.Kxe5
  +-  (3.11)   Depth: 5   00:00:00
  =  (0.00)   Depth: 9   00:00:00  10kN, tb=1
1.Ra7 Kf6 2.Ke4 Re2+ 3.Kd5 Rd2+ 4.Kc6 Rg2
  +/=  (0.37)   Depth: 9   00:00:00  12kN, tb=1
1.Ke4 Re2+ 2.Kd3 Rg2 3.Ra7 Kf6 4.Ke4 Re2+ 5.Kd5 Rd2+ 6.Kc6 Rg2 7.Ra8 Kxg7
  +/=  (0.49)   Depth: 9   00:00:00  19kN, tb=2
  =  (0.00)   Depth: 21   00:00:15  1597kN, tb=1450

(was already 0.01 some plies earlier)
Up Topic Rybka Support & Discussion / Rybka Discussion / After this endgame example I rest my case :-)

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill