|
< Earlier Kibitzing · PAGE 14 OF 79 ·
Later Kibitzing> |
| May-10-06 | | RandomVisitor: 17.d5 cxd5 18.Rad1 Rxe2 19.Rxd5 is likely best play +0.46 |
|
| May-11-06 | | WillC21: <notyetagm> I understand what you meant. If you feel, intuitively, that White has some better moves than in the lines given by <randomvisitor> and me then just let us know. |
|
| May-11-06 | | monad: <RandomVisitor: Version 1.2 d/???/f> I was offered the chance to download one of the beta versions. Actually I think it was the 1.2d. So far I have read that it is very buggy, and seeing that I paid for the 1.1 version, I don't want to overwrite this and beta test. When do you think it is a good time to replace the original commercial version? And, more importantly, why? I would appreciate your opinion and advice. All this programming is way above my head and I can't figure out for myself what to do, if anything. |
|
| May-11-06 | | monad: <Vasik will probably announce when version 1.2 has become stable and working to full potential> I am not so sure. He advised me per email to download 1.2d, and I have a feeling that .d doesn't have all the bugs fixed yet. He had been rather busy with the Leiden tourney I suppose. Besides, why would there be an -f- version if -d- was OK? Confusing. I reckon I leave well alone for a bit. |
|
| May-11-06 | | RandomVisitor: There appear to be problems with the present release of Rybka - 1.2f. It repeatedly crashes my graphical user interface chessprogram9. It has also come up with some weird moves. Apparantly, when Vasik patched the early release he caused some problems in other areas - this is common. He should have delayed the release date. I have had Rybka1.2f crash in arena. So I would just use version 1.1 until version 1.2 stabilizes. |
|
| May-11-06 | | RandomVisitor: Thisis from Talkchess.com:
A number of Chessbase server engine room players have seen a dramatic drop in Rybka's performance. This is totally a consequence of upgrading the GUI to the latest GUI9 vers 5 as indicated on the update notification when logging in but the C:\Windows\Chssbase.ini file shows it as GUI9 vers 8.
This update causes a 5s time delay (AMD 64+ 4000) before the engine starts during position analysis and seems to cause Rybka's time control to go haywire during game play causing a previously rare occurence of time loss to become quite common. This may affect all UCI engines, not just Rybka.
I tried this update and had an unprecedented string of six consecutive defeats against random opposition. Reverting to GUI9 version 4 restored playing strength.
Let us hope this is not Chessbase's response to Rybka dominating the field. It would be better to focus on improving their own commercial releases to be competitive! It would be sad if it was deliberate dirty tricks to prevent UCI engines performing through their interface.
To avoid the problem, make sure no higher than GUI9 version 4 is installed. If you made mistake of updating then uninstalling all GUI9 software and reinstalling disk version may be only recourse.
This affects all Rybka 1.2 versions tested equally. This does not occcur when running through the non Chessbase Shredder interface. |
|
| May-11-06 | | RandomVisitor: Someone at the Talkchess.com forum has run an experiment where he has played many "1 minute" per game games among the various Rybka's. Here are some of his results: Program Elo + - Games Score Av.Op. Draws 1 Rybka v1.2 test 4.w32 : 2962 103 98 33 72.7 % 2792 36.4 % 2 Rybka v1.2c.w32 : 2899 98 95 32 64.1 % 2799 40.6 % 3 Rybka v1.01 Beta 13b.w32 : 2889 105 103 33 63.6 % 2792 30.3 % 4 Rybka v1.01 Beta 11.w32 : 2885 88 86 34 61.8 % 2802 47.1 % 5 Rybka v1.2f.w32 : 2885 106 103 32 62.5 % 2796 31.2 % While this can hardly be considered scientific, it seems that there are some performance issues with the current release of Rybka. |
|
| May-11-06 | | notyetagm: <RandomVisitor: ... Let us hope this is not Chessbase's response to Rybka dominating the field.> If you can't beat them, break them.
:-)
|
|
| May-12-06 | | babakova: Anyone with rybka who could maybe point out some mistakes in this correspondence game of mine? 1. e4 c5
2. Nf3 d6
3. d4 cxd4
4. Nxd4 Nf6
5. Nc3 a6
6. Be3 Ng4
7. Bg5 h6
8. Bh4 g5
9. Bg3 Bg7
10. h3 Ne5
11. f3 Nbc6
12. Bf2 Be6
13. Qd2 Nxd4
14. Bxd4 Qa5
15. a3 O-O
16. h4 Ng6
17. hxg5 Qxg5
18. Bxg7 Qxd2+
19. Kxd2 Kxg7
20. g4 d5
21. Nxd5 Rad8
22. c4 Bxd5
23. exd5 e6
24. Bd3 Ne5
25. f4 exd5
26. c5 Nxg4
27. Rag1 f5
28. Bxf5 Rxf5
29. Rxg4+ Kh7
30. Rgh4 Rf6
31. Kd3 a5
32. b4 axb4
33. axb4 Re8
34. b5 Re4
35. f5 Rxh4
36. Rxh4 Rxf5
37. c6 bxc6
38. bxc6 Kg6
39. c7 Rf8
40. Kd4 Kg5
41. Rh1 h5
42. Rg1+ Kh6
43. Kxd5 h4
44. Kd6 Kh5
45. Rb1 Rc8
46. Kd7 Rxc7+
47. Kxc7 1/2-1/2
|
|
| May-13-06 | | RandomVisitor: <babakova>(#41): 39.Rg4+ Kh7 40.Kd4 h5 41.Rg1 Rf8 42.Rg5 Kh6 43.Re5 h4 44.Kxd5 wins for white (better was 38...Rf6=). 32.Kd4 would preserve a nice advantage for white -- 32.b4 equalizes for Black. |
|
| May-13-06 | | babakova: <RandomVisitor> 32.Kd4! yes that was the move I was afraid of. The last few moves might be mixed up ( I think white played 37.Kd4 not c6) because I wouldnt miss a move like Rf6. But even with the given position--- Cant I still draw with : 39.Rg4+ Kh7 40.Kd4 h5 41.Rg1 and now instead of the computerish Rf8 I go for my drawing idea 41...Kh6 securing important squares. 42.c7 Rf8 43.Rc1 Rc8 44.Kxd5 h4 and white has to sac his rook for the pawn later? Thx for your time, much appreciated. |
|
| May-13-06 | | RandomVisitor: <babakova>Rybka says that The position on the board after Black's 38th move is a mate in 41 (!) beginning with 39.Rg4+. After your idea 41...Kh6 it is mate in 37 with 42.Rg8. This analysis is assisted by tablebases. |
|
| May-13-06 | | babakova: Scary... One should not mess with tablebases... |
|
| May-13-06 | | RandomVisitor: <babakova>play might continue: 39. Rg4+ Kh7 40. Kd4 h5 41. Rg1 Kh6 42. Rg8 Rf7 43. Kxd5 h4 44. Ke6 Ra7 45. Kf6 Kh5 46. Rg5+ Kh6 47. Rc5 Rc7 48. Ke5 h3 49. Kd6 h2 50. Rc1 Rc8 51. Rh1 Rd8+ 52. Ke7 Rd2 53. c7 Re2+ |
|
| May-13-06 | | babakova: <RandomVisitor> I guess I was lucky that the position we are discussing didnt come up, but in fact he might have been able to transpose into it.. Chess is mighty complex sometimes... thx again. |
|
| May-13-06 | | tjoffy: Here it is! The great book series from the all-time strongest chessplayer: Hydra: "My great preprocessors"
The critics say:
Fritz: "I enjoyed every Bit"
Rybka: " +2.08" |
|
| May-14-06 | | MagnaPsygnosis: I dont know.....
Put Rybka in Hydra's shoes (64 CPU cluster).
Then Hydra will be annihalated..
(Same goes for Fruit, Fritz, Shredder) |
|
| May-15-06 | | RandomVisitor: I received this e-mail from Vasik concerning the changes from Rybka 1.2f to Rybka 1.2i: Hi John,
ok. This means that the earlier problem was caused by the GUI sending illegal positions to the engine. The only difference between 1.2i and 1.2f is that 1.2i is robust to this while 1.2f is not. Thanks for the information.
Best regards,
Vas
|
|
| May-15-06 | | oneguy: "Put Rybka in Hydra's shoes (64 CPU cluster)."
Hydra is not just 64 CPUs ... it has special FPGA cards that are programmed to play chess (like Deep Blue had special chess hardware). So maybe even 64 cpu Rybka won't stand a chance against Hydra. |
|
| May-15-06 | | MrMojoRisin: <Here it is! The great book series from the all-time strongest chessplayer: Hydra: "My great preprocessors"
The critics say:
Fritz: "I enjoyed every Bit"
Rybka: " +2.08">
*chuckle, chuckle* |
|
| May-16-06 | | MrMojoRisin: 
click for larger viewI came across that position the other day in some thread here. It is White to move, mate in 9. Fritz8 running on 200MB RAM, 2.7GHz CPU, took a full minute to find the mate in 9. I just ran it through the free download version of Rybka (same RAM and CPU). It found the mate in 9 in a microsecond. I can't believe Rybka found that combo in a hundreth of a second, while it took Fritz an entire minute. |
|
| May-16-06 | | LoFarkas: Something is seriously wrong with your fritz... CM 9000 spots the mate in 9 within 1 second, as expected. Engines always do. They flick through a good couple of million positions per minute. Of course, they will start with looking at forcing lines, giving checks etc. Look at how many legal moves black has in that forced line, adding them up after each move. One to 5 at each move, so it'll be under a hundred in all. Of course a comp will go through them in no time. Anyone care to have their fritz look at the position? |
|
| May-16-06 | | MrMojoRisin: Nope nothing is wrong with my Fritz. The person who posted that position (can't remember who it is) said that it takes Fritz5 forever to find the mate. It doesn't take Fritz8 forever, but it does take a lot longer than it should. |
|
| May-16-06 | | MrMojoRisin: here it is Rotlewi vs Rubinstein, 1907 |
|
| May-16-06 | | LoFarkas: Ok, then probably Fritz is programmed in a silly way and it doesn't start analysis with the forced lines. |
|
 |
|
|
< Earlier Kibitzing · PAGE 14 OF 79 ·
Later Kibitzing> |
Kibitzing
