Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Deep Fritz 8 vs LCZero NN id 276 - A modern day miniature!
- - By Peter Grayson (****) [us] Date 2018-05-12 13:11
[Event "LCZero vs Deep Fritz 8"]
[Site "DEVO4"]
[Date "2018.05.11"]
[Round "4"]
[White "Deep Fritz 8"]
[Black "lczero v0.10"]
[Result "1-0"]
[ECO "B01"]
[Annotator "0.32;0.28"]
[PlyCount "43"]
[TimeControl "40/300:40/300:40/300"]

{Intel(R) Core(TM)2 Quad CPU    Q9550  @ 2.83GHz 3535 MHz  W=12.3 plies; 10,
076kN/s  B=16.0 plies; 7kN/s} 1. e4 d5 2. exd5 Qxd5 {Both last book move} 3.
Nc3 {[%eval 32,13] [%emt 0:00:06]} Qd8 {[%eval 28,16] [%emt 0:00:04] (Qe6+)} 4.
Nf3 {[%eval 54,14] [%emt 0:00:10]} Nf6 {[%eval 26,16] [%emt 0:00:03]} 5. d4 {
[%eval 56,14] [%emt 0:00:18]} e6 {[%eval 30,17] [%emt 0:00:07] (c6)} 6. Bd3 {
[%eval 57,13] [%emt 0:00:05]} c5 {[%eval 36,17] [%emt 0:00:07] (Bd6)} 7. O-O {
[%eval 64,12] [%emt 0:00:06] (Be3)} cxd4 {[%eval 23,15] [%emt 0:00:04] (Nc6)}
8. Nxd4 {[%eval 62,13] [%emt 0:00:06]} Be7 {[%eval 26,17] [%emt 0:00:07]} 9.
Be3 {[%eval 60,13] [%emt 0:00:09] (Nf3)} O-O {[%eval 22,16] [%emt 0:00:05]} 10.
Qf3 {[%eval 70,14] [%emt 0:00:16] (Qe2)} Nbd7 {[%eval 18,16] [%emt 0:00:04]}
11. Rad1 {[%eval 75,13] [%emt 0:00:10]} Ne5 {[%eval 3,16] [%emt 0:00:04]} 12.
Qg3 {[%eval 75,14] [%emt 0:00:09]} Nxd3 {[%eval -4,16] [%emt 0:00:03]} 13. Rxd3
{[%eval 68,12] [%emt 0:00:02]} Qa5 {[%eval 0,17] [%emt 0:00:08]} 14. Ndb5 {
[%eval 86,13] [%emt 0:00:16] (Rfd1)} Bd7 {[%eval -81,16] [%emt 0:00:05]} 15.
Bh6 {[%eval 166,14] [%emt 0:00:10] (Nd6)} Nh5 {[%eval -38,16] [%emt 0:00:05]}
16. Qe5 {[%eval 164,13] [%emt 0:00:04] (Qg4)} Bxb5 {[%eval 11,17] [%emt 0:00:
09]} 17. Rg3 {[%eval 182,12] [%emt 0:00:02] (Qxh5)} Bf6 {[%eval 117,15] [%emt
0:00:06] (Bxf1)} 18. Qxh5 {[%eval 302,12] [%emt 0:00:01]} Kh8 {[%eval 66,16]
[%emt 0:00:04] (Qb4)} 19. Ne4 {[%eval 863,12] [%emt 0:00:03] (Qxb5)} Bxb2 {
[%eval 23,16] [%emt 0:00:06] (Qd8)} 20. Rh3 {[%eval 32744,11] [%emt 0:00:01]
(Bxg7+)} g6 {[%eval 427,16] [%emt 0:00:08] (Qc3)} 21. Bg7+ {[%eval 32764,2]
[%emt 0:00:00] (Qh4)} Kxg7 {[%eval 12721,13] [%emt 0:00:07] (Bxg7)} 22. Qxh7# {
[%eval 32766,2] [%emt 0:00:00]} 1-0



I had set out 17 basic King's pawn opening lines to identify just how much LCZero "understands" chess. With current scores standing at
LCZero vs Deep HIARCS 14, +0/=2/-18, 1-19
LCZero vs Deep Fritz 8, +0/=0/-16, 0-16

clearly LCZero does not understand very much with the consequence I have not completed the 17 lines yet.

It became clear that currently LCzero has a massive gap in its knowledge, particularly how to finish off an opponent when in a winning position having lost at least two games it should have won.

This morning I updated the NN to id 280 and started a match against Deep Rybka 4.1. Surprisingly LCZero managed its first win in game 3 but currently stands 7-1 down.

[Event "LCZero Test, 5m/40+5m/40+5m/40"]
[Site "DEVO4"]
[Date "2018.05.12"]
[Round "3"]
[White "lczero v0.10"]
[Black "Deep Rybka 4.1 x64"]
[Result "1-0"]
[ECO "B01"]
[Annotator "0.26;0.42"]
[PlyCount "149"]
[TimeControl "40/300:40/300:40/300"]

{Intel(R) Core(TM)2 Quad CPU    Q9550  @ 2.83GHz 3536 MHz  W=16.2 plies; 0kN/s;
508 TBAs  B=15.6 plies; 365kN/s} 1. e4 d5 2. exd5 Qxd5 {Both last book move} 3.
Nc3 {[%eval 26,16] [%emt 0:00:06]} Qd6 {[%eval 42,15] [%emt 0:00:11] (Qd8)} 4.
Nf3 {[%eval 46,16] [%emt 0:00:07]} Nf6 {[%eval 41,16] [%emt 0:00:08]} 5. d4 {
[%eval 46,16] [%emt 0:00:03] (Bc4)} a6 {[%eval 38,16] [%emt 0:00:08] (g6)} 6.
g3 {[%eval 45,16] [%emt 0:00:05] (a3)} Bf5 {[%eval 37,15] [%emt 0:00:21] (Nc6)}
7. Bg2 {[%eval 49,16] [%emt 0:00:05] (a3)} Nc6 {[%eval 26,16] [%emt 0:00:06]}
8. O-O {[%eval 56,16] [%emt 0:00:04]} O-O-O {[%eval 32,15] [%emt 0:00:15] (e6)}
9. Be3 {[%eval 98,17] [%emt 0:00:08] (Bf4)} e6 {[%eval 11,15] [%emt 0:00:13]
(e5)} 10. Qc1 {[%eval 138,16] [%emt 0:00:08] (a3)} h6 {[%eval 0,14] [%emt 0:00:
07] (Ng4)} 11. Rd1 {[%eval 154,16] [%emt 0:00:08] (Re1)} Qe7 {[%eval 0,15]
[%emt 0:00:07] (Nd5)} 12. a3 {[%eval 160,16] [%emt 0:00:07] (Nh4)} g5 {[%eval
21,13] [%emt 0:00:06]} 13. b4 {[%eval 202,16] [%emt 0:00:05]} Nd5 {[%eval 20,
14] [%emt 0:00:04]} 14. Na4 {[%eval 179,17] [%emt 0:00:06] (Nxd5)} Qe8 {[%eval
0,14] [%emt 0:00:10] (Bg4)} 15. c4 {[%eval 257,16] [%emt 0:00:06] (Qb2)} Nxe3 {
[%eval -30,13] [%emt 0:00:04]} 16. fxe3 {[%eval 133,16] [%emt 0:00:03]} g4 {
[%eval -13,13] [%emt 0:00:03]} 17. Nh4 {[%eval 85,16] [%emt 0:00:03]} Nxd4 {
[%eval -13,14] [%emt 0:00:02]} 18. Rxd4 {[%eval 70,17] [%emt 0:00:07] (exd4)}
Rxd4 {[%eval -30,14] [%emt 0:00:02]} 19. exd4 {[%eval 67,16] [%emt 0:00:02]
(Nb6+)} Qxa4 {[%eval -20,15] [%emt 0:00:02]} 20. Nxf5 {[%eval 56,17] [%emt 0:
00:03]} exf5 {[%eval -14,15] [%emt 0:00:01]} 21. Qf4 {[%eval 26,17] [%emt 0:00:
07] (Qf1)} Qd7 {[%eval -25,17] [%emt 0:00:07]} 22. c5 {[%eval -1,17] [%emt 0:
00:10] (d5)} Bg7 {[%eval -30,16] [%emt 0:00:15]} 23. Rd1 {[%eval 14,16] [%emt
0:00:04]} h5 {[%eval -30,15] [%emt 0:00:08] (Re8)} 24. d5 {[%eval 92,16] [%emt
0:00:06]} h4 {[%eval -18,16] [%emt 0:00:04] (Re8)} 25. gxh4 {[%eval 76,16]
[%emt 0:00:10] (d6)} Rxh4 {[%eval 0,15] [%emt 0:00:09] (Kb8)} 26. Qg5 {[%eval
153,17] [%emt 0:00:11]} Rh7 {[%eval 0,16] [%emt 0:00:01]} 27. d6 {[%eval 191,
16] [%emt 0:00:08]} f6 {[%eval 0,16] [%emt 0:00:03] (c6)} 28. Qe3 {[%eval 136,
16] [%emt 0:00:10] (Qc1)} Bh6 {[%eval 0,16] [%emt 0:00:07]} 29. Qe2 {[%eval 93,
17] [%emt 0:00:10] (Qb3)} c6 {[%eval -45,15] [%emt 0:00:04]} 30. a4 {[%eval
148,16] [%emt 0:00:05] (Rf1)} Bf4 {[%eval -14,13] [%emt 0:00:10] (f4)} 31. b5 {
[%eval 55,16] [%emt 0:00:10]} cxb5 {[%eval 0,14] [%emt 0:00:06]} 32. Kf1 {
[%eval 3,17] [%emt 0:00:09] (Qa2)} Rxh2 {[%eval -17,12] [%emt 0:00:08] (Be5)}
33. Qe7 {[%eval -17,16] [%emt 0:00:10] (Rb1)} Rh7 {[%eval -26,13] [%emt 0:00:
07] (Rxg2)} 34. Qxf6 {[%eval 15,16] [%emt 0:00:08]} Kb8 {[%eval -27,13] [%emt
0:00:01]} 35. Re1 {[%eval 53,17] [%emt 0:00:08] (Rb1)} bxa4 {[%eval 135,13]
[%emt 0:00:13]} 36. Rb1 {[%eval 50,16] [%emt 0:00:07]} a3 {[%eval 135,13]
[%emt 0:00:01]} 37. Rxb7+ {[%eval 65,16] [%emt 0:00:06]} Qxb7 {[%eval 135,13]
[%emt 0:00:01]} 38. Bxb7 {[%eval 91,17] [%emt 0:00:07]} Rxb7 {[%eval 149,15]
[%emt 0:00:02]} 39. Qe6 {[%eval 79,19] [%emt 0:00:53] (Ke2)} Rb1+ {[%eval 149,
15] [%emt 0:00:23]} 40. Ke2 {[%eval 87,18] [%emt 0:00:02]} Rb5 {[%eval 234,17]
[%emt 0:00:17] (Ka7)} 41. d7 {[%eval 273,15] [%emt 0:00:08]} Rb2+ {[%eval 249,
17] [%emt 0:00:11]} 42. Kd3 {[%eval 377,16] [%emt 0:00:08]} Kc7 {[%eval 384,16]
[%emt 0:00:06] (Rd2+)} 43. Kc3 {[%eval 105,15] [%emt 0:00:09] (Kc4)} Rb8 {
[%eval 257,14] [%emt 0:00:02]} 44. Qxf5 {[%eval 188,17] [%emt 0:00:14]} Bg3 {
[%eval 257,15] [%emt 0:00:07] (Bh2)} 45. Qd5 {[%eval 274,16] [%emt 0:00:10]} a2
{[%eval 327,17] [%emt 0:00:08] (Bh2)} 46. Qxa2 {[%eval 413,15] [%emt 0:00:07]}
Kxd7 {[%eval 327,18] [%emt 0:00:03] (Be5+)} 47. Qd5+ {[%eval 360,16] [%emt 0:
00:11] (Qa4+)} Kc7 {[%eval 373,20] [%emt 0:00:09]} 48. Qf7+ {[%eval 327,15]
[%emt 0:00:07]} Kc8 {[%eval 373,20] [%emt 0:00:21] (Kc6)} 49. Kc4 {[%eval 332,
16] [%emt 0:00:13] (Qg6)} Rb7 {[%eval 581,16] [%emt 0:00:05] (Rb5)} 50. Qg8+ {
[%eval 543,15] [%emt 0:00:07]} Kc7 {[%eval 581,18] [%emt 0:00:01]} 51. Qxg4 {
[%eval 645,15] [%emt 0:00:04]} Bd6 {[%eval 581,19] [%emt 0:00:01] (Bh2)} 52.
cxd6+ {[%eval 2966,14] [%emt 0:00:06]} Kxd6 {[%eval 581,20] [%emt 0:00:01]} 53.
Qg6+ {[%eval 32722,1] [%emt 0:00:00]} Ke5 {[%eval 32724,1] [%emt 0:00:00]} 54.
Qg5+ {[%eval 32724,1] [%emt 0:00:00]} Kd6 {[%eval 32726,1] [%emt 0:00:00]} 55.
Qf6+ {[%eval 32726,1] [%emt 0:00:00]} Kc7 {[%eval 32728,1] [%emt 0:00:00]} 56.
Qxa6 {[%eval 32728,1] [%emt 0:00:00]} Rb6 {[%eval 32730,1] [%emt 0:00:00]} 57.
Qa7+ {[%eval 32730,1] [%emt 0:00:00]} Kc6 {[%eval 32732,0] [%emt 0:00:00]} 58.
Qe7 {[%eval 32732,1] [%emt 0:00:00]} Rb7 {[%eval 32734,1] [%emt 0:00:00]} 59.
Qe6+ {[%eval 32734,1] [%emt 0:00:00]} Kc7 {[%eval 32736,1] [%emt 0:00:00]} 60.
Kc5 {[%eval 32736,1] [%emt 0:00:00]} Kd8 {[%eval 32738,1] [%emt 0:00:00]} 61.
Qf5 {[%eval 32738,1] [%emt 0:00:00]} Kc7 {[%eval 32740,1] [%emt 0:00:00]} 62.
Qf3 {[%eval 32740,1] [%emt 0:00:00]} Ra7 {[%eval 32742,1] [%emt 0:00:00]} 63.
Qc6+ {[%eval 32742,1] [%emt 0:00:00]} Kb8 {[%eval 32744,1] [%emt 0:00:00]} 64.
Qe4 {[%eval 32744,1] [%emt 0:00:00]} Kc7 {[%eval 32746,1] [%emt 0:00:00]} 65.
Qe5+ {[%eval 32746,1] [%emt 0:00:00]} Kb7 {[%eval 32748,1] [%emt 0:00:00]} 66.
Kb5 {[%eval 32748,1] [%emt 0:00:00]} Ra2 {[%eval 32750,1] [%emt 0:00:00]} 67.
Qd5+ {[%eval 32750,1] [%emt 0:00:00]} Kc7 {[%eval 32752,1] [%emt 0:00:00]} 68.
Qxa2 {[%eval 32752,1] [%emt 0:00:00]} Kd6 {[%eval 32754,1] [%emt 0:00:00]} 69.
Qe2 {[%eval 32754,0] [%emt 0:00:00]} Kd5 {[%eval 32756,1] [%emt 0:00:00]} 70.
Qe3 {[%eval 32756,0] [%emt 0:00:00]} Kd6 {[%eval 32758,1] [%emt 0:00:00]} 71.
Qe4 {[%eval 32758,0] [%emt 0:00:00]} Kd7 {[%eval 32760,1] [%emt 0:00:00]} 72.
Qe5 {[%eval 32760,0] [%emt 0:00:00]} Kc8 {[%eval 32762,1] [%emt 0:00:00]} 73.
Kc6 {[%eval 32762,0] [%emt 0:00:00]} Kd8 {[%eval 32764,1] [%emt 0:00:00]} 74.
Qe4 {[%eval 32764,0] [%emt 0:00:00]} Kc8 {[%eval 32766,1] [%emt 0:00:00]} 75.
Qe8# {[%eval 32766,1] [%emt 0:00:00]} 1-0



Perhaps high end graphics cards would show LCZero in a better light but the economics of it currently makes no sense.

PeterG
Parent - - By shrapnel (***) [in] Date 2018-05-13 09:43 Edited 2018-05-13 11:49
Hi Peter
          Which Graphic Card are you using ?
Parent - - By Peter Grayson (****) [us] Date 2018-05-13 20:06

> Which Graphic Card are you using ?


I am not, the GPU version will not run on this 10 year old machine's nVidia GeForce 8600GT and does not look very fast on my dual Xeon machine's nVidia Quadro 600 card hence my comment regarding the impracticality of buying a high end GPU for a chess engine that does not look very good. The issue is how many times I am seeing LCZero get hit with a sucker punch that highlights a lack of basic knowledge regarding King safety. Just finished some games using two same build Quad PC's against the old Genius 7 engine via RS232 Autoplayer. Genius will not run in the CB GUI. Although LCZero registered most wins I have seen yet, it still lost overall to the 18 year old Genius 7 engine running on just 1 CPU and this 21 move defeat epitomises the naivety of the LCZero engine.

Genius 7 mated LCZero in 21 moves here! LCZero indicated a deeper ply depth than Genius 7 (and Fritz 5.32 for that matter) but was way off the money. Presently I see little merit in the AZ approach when a low end machine is sufficient for basic word processing and spreadsheets and Stockfish will see off most if not all human opponents therefore I am unable to identify any merit in chasing after the AZ dream with expensive hardware. LCZero seems to be somewhere around where chess engines were about 20 years ago in performance terms.

In this game Genius 7 ran with its supplied openings book and LCZero ran without book as was suggested elsewhere but the game score confirms LCZero needs an opening book. It also needs some chess knowledge!

[Event "5'/40+5'/40+5'/40"]
[Site "Newport, South Wales"]
[Date "2018.05.13"]
[Round "4"]
[White "Genius 7"]
[Black "LCZero V0.10"]
[Result "1-0"]
[ECO "D23"]
[PlyCount "41"]
[TimeControl "40/300:40/300:40/300"]

{1024MB, MT.ctg, DEVO4} 1. d4 {[%emt 0:00:00]} Nf6 {[%eval 12,16] [%emt 0:00:
07]} 2. c4 {[%emt 0:00:01]} e6 {[%eval 8,15] [%emt 0:00:04]} 3. Nf3 {[%emt 0:
00:01]} d5 {[%eval 4,16] [%emt 0:00:06]} 4. Nc3 {[%emt 0:00:01] (cxd5)} dxc4 {
[%eval 6,16] [%emt 0:00:07]} 5. Qa4+ {[%emt 0:00:01] (e4)} Nbd7 {[%eval -14,15]
[%emt 0:00:05]} 6. e4 {[%emt 0:00:01] (Qxc4)} c5 {[%eval 5,16] [%emt 0:00:07]}
7. d5 {[%emt 0:00:01]} exd5 {[%eval 50,15] [%emt 0:00:02]} 8. e5 {[%emt 0:00:
01]} b5 {[%eval 84,16] [%emt 0:00:05]} 9. Qxb5 {[%emt 0:00:01]} Rb8 {[%eval 74,
16] [%emt 0:00:03]} 10. Qa4 {[%emt 0:00:01]} d4 {[%eval 68,16] [%emt 0:00:03]}
11. exf6 {[%emt 0:00:01]} dxc3 {[%eval 38,16] [%emt 0:00:02]} 12. Bxc4 {[%emt
0:00:01] (bxc3)} cxb2 {[%eval -8,17] [%emt 0:00:08]} 13. Bxf7+ {[%emt 0:00:10]
(Bxb2)} Kxf7 {[%eval -188,6] [%emt 0:00:00]} 14. Ng5+ {[%emt 0:00:14] (Qc4+)}
Kg6 {[%eval 150,16] [%emt 0:00:07]} 15. Qe4+ {[%emt 0:00:01]} Kh5 {[%eval 150,
15] [%emt 0:00:02]} 16. g4+ {[%emt 0:00:01]} Kh4 {[%eval 93,16] [%emt 0:00:03]}
17. Nf3+ {[%emt 0:00:01]} Kh3 {[%eval 97,16] [%emt 0:00:00]} 18. Ng1+ {[%emt 0:
00:01]} Kh4 {[%eval 102,14] [%emt 0:00:00]} 19. Bg5+ {[%emt 0:00:01] (Nf3+)}
Kxg5 {[%eval -129,6] [%emt 0:00:00]} 20. Qf5+ {[%emt 0:00:01] (Nf3+)} Kh6 {
[%eval 11149,15] [%emt 0:00:10]} 21. Qh5# {[%emt 0:00:01]} 1-0



                  
1   Genius 7      2148  +17/=7/-7 66.13%   20.5/31
2   lczero v0.10  2066  +7/=7/-17 33.87%   10.5/31

PeterG
Parent - By Peter Grayson (****) [us] Date 2018-05-13 20:37
The games. The scores shown above tell their own story.
Attachment: LCZeroTest.7z - The Games (109k)
Parent - - By shrapnel (***) [in] Date 2018-05-14 03:18
Have it on good authority that the GPU version is much, much stronger than the CPU version.
On the other hand, its true that LCZero is going through a rough patch in its development and making a lot of mistakes.
But I have a strong hunch that these are just hiccups and that LCZero is just a glimpse of the Future ; its just a matter of time before it overtakes every single  alpha-beta Engine out there !
Parent - By Peter Grayson (****) [us] Date 2018-05-14 14:38
I just compared the GPU engine using the graphics card on my dual Xeon machine and output looks similar to the 4 cores of my 10 year old quad. The engine was hitting 2 to 3 extra ply using the full 16 cores of the dual Xeon and looked about the same with 32 HT threads of the dual Xeon machine but in analysis would still be blind to the sucker punch to which it fell foul in the 21 move defeat to Genius 7 posted above.

It has a problem in that it cannot see 12.Bxc4 and misses 12..cxb2 13.Bxf7+ when it scores 14..Kg6 as 0.13. The latter move gives White mate in 7. It also misses 16.g4+, 18.Ng1 and 19.Bg5 with the conclusion it fails assessing material sacrifices as potential risks. Maybe that is the problem, it is too oriented towards bean counter characteristics. Whatever, at its present stage of development, it is certainly susceptible to being caught out by King hunts.

PeterG

Using the 32 HT Threads it needed over two minutes to change to 12..Rb4, in contrast Stockfish 9 required a fraction of a second at 16 ply to select 12..Rb4. LCZero showed a curious jump from 20 to 24 ply with a gap in any output. Also curious is the 1.24 score on the last 19 ply and initial two 20 ply output lines when it saw 25.Qg3#. Surely its scoring of the move should be significantly more adverse if it was to be mated? Very odd!

118: Genius 7 - lczero v0.10, 5'/40+5'/40+5'/40
1rbqkb1r/p2n1ppp/5P2/2p5/Q1B5/2p2N2/PP3PPP/R1B1K2R b KQk - 0 1


Analysis by lczero v0.10:

12...Qxf6 13.0-0
  +/-  (1.12)   Depth: 7   00:00:00  0kN
12...Qxf6 13.0-0 Be7
  +/-  (0.89)   Depth: 8   00:00:00  0kN
12...Qxf6 13.0-0 Be7 14.Bg5
  +/-  (1.19)   Depth: 8   00:00:00  0kN
12...Qxf6 13.0-0 Be7 14.Bg5 Qg6 15.Bxe7
  +/-  (0.85)   Depth: 11   00:00:00  0kN
12...Qxf6 13.0-0 Be7 14.Bg5 Qg6 15.Bxe7
  +/-  (1.02)   Depth: 12   00:00:00  0kN
12...Qxf6 13.0-0 Be7 14.Bg5 Qd6 15.Rad1 Qc7
  +/-  (1.00)   Depth: 12   00:00:00  0kN
12...Qxf6 13.0-0 Be7 14.Bg5 Qg6 15.Rae1 0-0
  +/-  (0.72)   Depth: 14   00:00:00  0kN
12...Qxf6 13.0-0 Be7 14.Bg5 Qg6 15.Rae1 0-0
  +/=  (0.69)   Depth: 14   00:00:00  0kN
12...Qxf6 13.0-0 Be7 14.Bg5 Qd6 15.Rad1 Qc7 16.Rfe1 f6
  +/=  (0.52)   Depth: 15   00:00:01  0kN
12...Qxf6 13.0-0 Rb4 14.Re1+ Be7 15.Qc2 cxb2 16.Bxb2 Rxb2 17.Qe4
  +/=  (0.42)   Depth: 16   00:00:01  0kN
12...Qxf6 13.0-0 Rb4 14.Re1+ Be7 15.Qc2 Rxc4 16.Bg5 Qd6 17.Rxe7+
  +/=  (0.36)   Depth: 16   00:00:01  0kN
12...Rb4 13.Qc2 Rxc4 14.Qe2+ Be7 15.Qxc4 Bxf6 16.bxc3 0-0 17.0-0 Qa5 18.Bg5 Bxc3
  +/=  (0.33)   Depth: 16   00:00:01  0kN
12...Qxf6 13.0-0 cxb2 14.Bxb2 Rxb2 15.Rfe1+ Be7 16.Rad1 0-0 17.Rxd7 Bxd7 18.Qxd7 Bd6 19.Rd1
  =  (0.13)   Depth: 17   00:00:02  1kN
12...Qxf6 13.0-0 cxb2 14.Bxb2 Rxb2 15.Rfe1+ Be7 16.Rad1 0-0 17.Rxd7 Bxd7 18.Qxd7 Bd6 19.Rd1
  =  (0.10)   Depth: 17   00:00:02  1kN
12...cxb2 13.Bxb2 Rxb2 14.0-0 Qxf6 15.Rae1+ Be7 16.Ne5 Rb4 17.Bxf7+ Kf8 18.Nxd7+ Bxd7 19.Qxd7 Kxf7
  =  (-0.22)   Depth: 18   00:00:03  1kN
12...cxb2 13.Bxb2 Rxb2 14.0-0 Qxf6 15.Rae1+ Be7 16.Ne5 Rb4 17.Bxf7+ Kf8 18.Nxd7+ Bxd7 19.Qxd7 Kxf7 20.Qd5+ Kg6
  =/+  (-0.34)   Depth: 18   00:00:03  1kN
12...cxb2 13.Bxb2 Rxb2 14.0-0 Qxf6 15.Rae1+ Be7 16.Ne5 Rb4 17.Bxf7+ Kf8 18.Nxd7+ Bxd7 19.Qxd7 Kxf7 20.Re3 Rd8 21.Qxa7
  =/+  (-0.47)   Depth: 18   00:00:04  2kN
12...cxb2 13.Bxb2 Rxb2 14.0-0 Qxf6 15.Rae1+ Be7 16.Ne5 Rb4 17.Bxf7+ Kf8 18.Nxd7+ Bxd7 19.Qxd7 Kxf7 20.Qd5+ Kf8 21.Qa8+ Kf7
  =/+  (-0.51)   Depth: 18   00:00:04  2kN
12...cxb2 13.Bxb2 Rxb2 14.0-0 Qxf6 15.Rae1+ Be7 16.Ne5 Rb4 17.Bxf7+ Kf8 18.Nxd7+ Bxd7 19.Qxd7 Kxf7 20.Qd5+ Kg6 21.Re6 Kf7 22.Rxf6+
  =/+  (-0.50)   Depth: 19   00:00:05  2kN
12...cxb2 13.Bxb2 Rxb2 14.0-0 Qxf6 15.Rae1+ Be7 16.Ne5 Rb4 17.Bxf7+ Kf8 18.Nxd7+ Bxd7 19.Qxd7 Kxf7 20.Qd5+ Kg6 21.Re6 Rd8 22.Rxf6+ Bxf6
  =  (0.00)   Depth: 19   00:00:06  3kN
12...cxb2 13.Bxf7+ Kxf7 14.Ng5+ Kg6 15.Qe4+ Kh5 16.g4+ Kh4 17.Nf3+ Kh3 18.Rg1 bxc1Q+ 19.Rxc1 Qa5+ 20.Kd1 Nxf6 21.Rg3#
  =  (0.03)   Depth: 19   00:00:06  3kN
12...cxb2 13.Bxf7+ Kxf7 14.Ng5+ Kg6 15.Qe4+ Kh5 16.g4+ Kh4 17.Nf3+ Kh3 18.Rg1 bxc1Q+ 19.Rxc1 Bd6 20.Ng5+ Kh4 21.Nf3+ Kh3 22.Ng5+ Kh4 23.Nf3+ Kh3
  =  (0.13)   Depth: 19   00:00:07  3kN
12...cxb2 13.Bxf7+ Kxf7 14.Ng5+ Kg6 15.Qe4+ Kh5 16.g4+ Kh4 17.Nf3+ Kh3 18.Rg1 bxc1Q+ 19.Rxc1 Bd6 20.Ng5+ Kh4 21.Nf3+ Kh3 22.Ng5+ Kh4 23.Nf3+ Kh3
  =  (0.23)   Depth: 19   00:00:07  3kN
12...cxb2 13.Bxf7+ Kxf7 14.Ng5+ Kg6 15.Qe4+ Kh5 16.g4+ Kh4 17.Nf3+ Kh3 18.Rg1 bxc1Q+ 19.Rxc1 Bd6 20.Ng5+ Kh4 21.Nf3+ Kh3 22.Ng5+ Kh4 23.Nf3+ Kh3
  =  (0.23)   Depth: 19   00:00:07  4kN
12...cxb2 13.Bxf7+ Kxf7 14.Ng5+ Kg6 15.Qe4+ Kh5 16.Nf7 Qe8 17.Qxe8 bxa1Q 18.g4+ Kh4 19.Qe3 Qxc1+ 20.Qxc1 Rg8 21.Qg5+ Kh3 22.Qh5+ Kg2 23.Qd5+ Kh3 24.Qf3+ Kh4 25.Qg3#
  +/-  (1.14)   Depth: 20   00:00:16  8kN
12...cxb2 13.Bxf7+ Kxf7 14.Ng5+ Kg6 15.Qe4+ Kh5 16.Nf7 Qe8 17.Qxe8 bxa1Q 18.g4+ Kh4 19.Qe3 Qxc1+ 20.Qxc1 Rg8 21.Qg5+ Kh3 22.Qh5+ Kg2 23.Qd5+ Kh3 24.Qf3+ Kh4 25.Qg3#
  +/-  (1.23)   Depth: 20   00:00:17  9kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 h6 20.Bh4 Qa8 21.Bg3 Bxg3 22.hxg3 Bd5 23.Rad1 Bxc4 24.Qxc4 Qb7 25.Rd3 Rc8 26.Ne5 Re6 27.Rde3 Rce8
  +/-  (0.77)   Depth: 24   00:02:07  56kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 h6 20.Bh4 Qa8 21.Bg3 Bxg3 22.hxg3 Bd5 23.Rad1 Bxc4 24.Qxc4 Qb7 25.Rd3 Rc8 26.Ne5 Re6 27.Rde3 Rce8
  +/-  (0.77)   Depth: 24   00:02:07  57kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 h6 20.Bh4 Qa8 21.Bg3 Bxg3 22.hxg3 Bd5 23.Rad1 Bxc4 24.Qxc4 Qb7 25.Rd3 Rc8 26.Ne5 Re6 27.Rde3 Rce8 28.Qxc5
  +/-  (0.77)   Depth: 24   00:02:09  58kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 h6 20.Bh4 Qa8 21.Bg3 Bxg3 22.hxg3 Bd5 23.Rad1 Bxc4 24.Qxc4 Qb7 25.Rd3 Rc8 26.Ne5 Re6 27.Rde3 Rce8 28.Qxc5
  +/-  (0.77)   Depth: 24   00:02:10  58kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 h6 20.Bh4 Qa8 21.Bg3 Bxg3 22.hxg3 Bd5 23.Rad1 Bxc4 24.Qxc4 Qb7 25.Rd3 Qa6 26.Qxa6 Rxa6 27.a4 c4 28.bxc4 Rxa4 29.Rc1 Rc8
  +/-  (0.76)   Depth: 25   00:02:28  66kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 h6 20.Bh4 Qa8 21.Bg3 Bxg3 22.hxg3 Bd5 23.Rad1 Bxc4 24.Qxc4 Qb7 25.Rd3 Qa6 26.Qxa6 Rxa6 27.a4 c4 28.bxc4 Rxa4 29.Rc1 Rc8
  +/-  (0.76)   Depth: 25   00:02:29  66kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 Rd6 20.Qc2 h6 21.Bf4 Bxf3 22.gxf3 Rd4 23.Bg3 Rd2 24.Qf5 Bxg3 25.hxg3 Qd4 26.Rf1 Re8 27.Rac1 Re5 28.Qc8+ Re8 29.Qf5 Re5 30.Qc8+
  +/-  (0.76)   Depth: 25   00:02:29  66kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 Rd6 20.Qc2 h6 21.Bf4 Bxf3 22.gxf3 Rd4 23.Bg3 Rd2 24.Qf5 Bxg3 25.hxg3 Qd4 26.Rf1 Re8 27.Rac1 Re5 28.Qc8+ Re8 29.Qf5 Re5 30.Qc8+
  +/-  (0.76)   Depth: 25   00:02:29  66kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 Rd6 20.Qc2 h6 21.Bf4 Bxf3 22.gxf3 Rd4 23.Bg3 Rd2 24.Qf5 Bxg3 25.hxg3 Qd4 26.Rf1 Re8 27.Rac1 Re5 28.Qc8+ Re8 29.Qf5 Re5 30.Qc8+
  +/-  (0.76)   Depth: 25   00:02:30  66kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 Rd6 20.Qc2 h6 21.Bf4 Bxf3 22.gxf3 Rd4 23.Bg3 Rd2 24.Qf5 Bxg3 25.hxg3 Qd4 26.Rf1 Re8 27.Rac1 Re5 28.Qc8+ Re8 29.Qf5 Re5 30.Qc8+
  +/-  (0.76)   Depth: 25   00:02:30  67kN
12...Rb4 13.Qc2 Nxf6 14.b3 Bd6 15.0-0 0-0 16.a3 Rb6 17.Qxc3 Bc7 18.Re1 Bb7 19.Bg5 Rd6 20.Qc2 h6 21.Bf4 Bxf3 22.gxf3 Rd4 23.Bg3 Rd2 24.Qf5 Bxg3 25.hxg3 Qd4 26.Rf1 Re8 27.Rac1 Re5 28.Qc8+ Re8 29.Qf5 Re5 30.Qc8+
  +/-  (0.76)   Depth: 25   00:02:31  67kN

(Grayson, Newport, South Wales 14.05.2018)
Parent - - By MarshallArts (***) [us] Date 2018-05-13 22:37
All LCZ networks after n241 are thought to be based on buggy versions/data.

:neutral:
Parent - - By Peter Grayson (****) [us] Date 2018-05-14 00:00

> All LCZ networks after n241 are thought to be based on buggy versions/data.


Thanks for info. That may explain some of the games when it failed to win what should have been straight forward victories involving a queen. I used NN id 276 and subsequently id 280 for the games.

I'll check it out again in a month or two when hopefully the problems will have been addressed.

Peter
Parent - By gsgs (***) [de] Date 2018-05-14 03:39
https://docs.google.com/spreadsheets/d/18UWR4FVhPi0vNwwPreu_avd9ycujGQ5ayR2LzJOWP4s/edit#gid=0
id276=2713,id278=2632,id283=2669,sf1.01=2767

https://docs.google.com/spreadsheets/d/1XSJiCcQpCLv0fNwrUn7jXjdkZFU63YFEWpdXv6dSSg0/edit#gid=0
id270=2805,id275=2660,id280=2652,Rybka1.0=2792

they use a strong GPU, subtract ~100 for LC0 for fair hardware comparison  = same Power (Watts)
Up Topic The Rybka Lounge / Computer Chess / Deep Fritz 8 vs LCZero NN id 276 - A modern day miniature!

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill