On-line engine blitz tourney June

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
Greg Strong
Posts: 388
Joined: Sun Dec 21, 2008 6:57 pm
Location: Washington, DC

Re: On-line engine blitz tourney June

Post by Greg Strong »

xr_a_y wrote: Sat Jun 23, 2018 5:22 pm
Visual Studio is amazing! It's the one program that I miss on Linux and will still switch to Windows for.
Me too, but QtCreator with some clang/LLMV based plugins is quite amazing too...
I meant to check out QtCreator but never got around to it. Heard it was nice. Now I've just started writing a new engine and I've decided to use D instead of C++. Really, really impressed with the D language. It was rough in the beginning but has matured a lot recently. Anyway, Code::Blocks supports D reasonably well and there aren't too many options.
jdart
Posts: 4366
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: On-line engine blitz tourney June

Post by jdart »

Microsoft went to auto-updates in Windows 10 for security reasons. It is actually a good thing. There is no point issuing security patches if no one is going to install them promptly. You can though turn it off if you set the network connection to "metered" (limited bandwidth).

Re development platforms I was a Windows developer in the 1990s but was an emacs/CLI guy before then and now that is still what I use. Once in a while I will use Visual Studio for debugging on Windows, but Linux has been my primary development platform for the past decade or so.

--Jon
Henk
Posts: 7216
Joined: Mon May 27, 2013 10:31 am

Re: On-line engine blitz tourney June

Post by Henk »

Nice (not). Last windows update cost me another 8 hours to install. Thank you Microsoft for reminding me that my computer is slow and totally useless. Now I finally found time to re-read my old chess books instead of useless programming.
tomitank
Posts: 276
Joined: Sat Mar 04, 2017 12:24 pm
Location: Hungary

Re: On-line engine blitz tourney June

Post by tomitank »

Sorry about the first round, i did not test new version of tomitankChess in Winboard.
User avatar
hgm
Posts: 27790
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: On-line engine blitz tourney June

Post by hgm »

Code: Select all

:Tourney Players: Round 9 of 9 
:
:     Name              Rating Score Perfrm Upset  Results 
:     ----------------- ------ ----- ------ ------ ------- 
:  1 +Texel             [2252]  13.0  [2299] [   0] +02w +07w =03b +04w =05b +11w +09b +08b +12w 
:  2 +LeelaZero         [1883]  12.5  [2247] [ 752] -01b +20b +12w +07w +03b =04w +05w +11b +09w 
:  3 +ArasanX           [2092]  11.5  [2171] [  81] +13b +09w =01w =11b -02w +08w =04b +05w +07b 
:  4 -nemorino          [2094]  10.5  [2082] [   0] +10w =11w +05b -01b +07w =02b =03w =09w =08b 
:  5 -WaspX             [2044]  10.0  [1977] [ 104] +14b =08w -04w +09b =01w +06w -02b -03b +11w 
:  6 +rpiArminius       [1581]  10.0  [1837] [ 743] =08b +12w -11b =13w +14w -05b -10w +20b +18b 
:  7 -NightmareX        [2025]  9.5  [1893] [   0] +15w -01b +13w -02b -04b +20w =11b +10b -03w 
:  8 -WaDuuttie         [1974]  9.5  [1872] [  95] =06w =05b =14w +19b =11w -03b +15b -01w =04w
:  9 -Pedone            [1915]  9.5  [1884] [ 102] +16w -03b +15w -05w +20b +12b -01w =04b -02b 
: 10 +RuyDos            [1819]  9.5  [1746] [   0] -04b -19w +18b +14b -12w +13w +06b -07w =15w 
: 11 +rpiStockfish      [1941]  9.0  [1902] [ 209] +19w =04b +06w =03w =08b -01b =07w -02w -05b 
: 12 -Goldbar           [1928]  9.0  [1741] [   0] =20w -06b -02b +15w +10b -09w =14w +13b -01b 
: 13 -Marvin            [1815]  9.0  [1700] [   0] -03w +22w -07b =06b +19w -10b +20w -12w =14b 
: 14 +Schooner          [1664]  9.0  [1689] [ 362] -05w +16b =08b -10w -06b +21w =12b =18w =13w 
: 15 -tomitankChess     [1612]  9.0  [1688] [ 103] -07b +23w -09b -12b +22w +18b -08w =19w =10b 
: 16 +Spartacus         [1517]  9.0  [1523] [  57] -09b -14w -22b -21b =27w +26b +24w +23w +19b
: 17 +Embla             [1223]  9.0  [1630] [1013] +24b +25b +26w +30w -21b =23w +27b +22w +28w 
: 18 -RookieMonster     [1421]  8.5  [1643] [ 338] +25w +21b -10w +27w +23b -15w +22b =14b -06w 
: 19 +Floyd             [1574]  8.0  [1555] [ 264] -11b +10b =20w -08w -13b -22b +21w =15b -16w 
: 20 +Bliep             [1535]  8.0  [1611] [ 215] =12b -02w =19b +22w -09w -07b -13b -06w +21b 
: 21 +rpiMinkoChess     [1502]  8.0  [1431] [  15] -22b -18w +23b +16w +17w -14b -19b +26w -20w 
: 22 +rpiTogaII         [1483]  8.0  [1444] [ 144] +21w -13b +16w -20b -15b +19w -18w -17b +29w 
: 23 +rpiFruit          [1454]  8.0  [1346] [   8] +27b -15b -21w +25b -18w =17b +31w -16b +34w 
: 24 +microMax          [1245]  8.0  [1409] [ 255] -17w +28b -27b +26w +31b +32w -16b +33w +30b
: 25 -Isa               [1414]  7.5  [1275] [   0] -18b -17w +30b -23w -26b +34w +29b +32w +33b 
: 26 +KingSlayer        [1409]  7.5  [1340] [  58] +33b +27w -17b -24b +25w -16w +34b -21b +31w 
: 27 +Joker             [1462]  7.0  [1207] [  27] -23w -26b +24w -18b =16b +29w -17w -28b +32b 
: 28 +rpiEmbla          [1128]  6.5  [1319] [ 633] +31b -24w +34w +32b -29w +30b +33w +27w -17b 
: 29 +Weini             [1205]  5.5  [1225] [ 131] -30w =32w +33b +31w +28b -27b -25w +34b -22b 
: 30 +rpiPhalanx        [1175]  5.0  [1131] [ 191] +29b +33w -25w -17b +34w -28w =32b =31b -24w 
: 31 +Skiull            [1172]  3.5  [1067] [ 165] -28w +34b =32w -29b -24w +33b -23b =30w -26b 
: 32 +rpiBikjump        [1061]  3.5  [1126] [ 459] +34w =29b =31b -28w +33b -24b =30w -25b -27w
: 33 +Teki              [1336]  2.0  [ 778] [   0] -26w -30b -29w -34b -32w -31w -28b -24b -25w 
: 34 +POS               [ 797]  1.0  [ 950] [ 539] -32b -31w -28b +33w -30b -25b -26w -29w -23b 
:
:     Average Rating    1580.9 
:
Joost Buijs
Posts: 1563
Joined: Thu Jul 16, 2009 10:47 am
Location: Almere, The Netherlands

Re: On-line engine blitz tourney June

Post by Joost Buijs »

jdart wrote: Sat Jun 23, 2018 7:00 pm Microsoft went to auto-updates in Windows 10 for security reasons. It is actually a good thing. There is no point issuing security patches if no one is going to install them promptly. You can though turn it off if you set the network connection to "metered" (limited bandwidth).
Of course, patching security issues is a necessity, but I don't like the way they force this upon you. When a system is idle, ok, but when a system is running at 100% CPU usage it is a stupidity to forcefully reboot. The metered connection only applies to WiFi and not to Ethernet connections, in the past it was possible to disable the reboot service, in the latest (1803) version though they automatically re-enable it when they feel it is necessary.
Joost Buijs
Posts: 1563
Joined: Thu Jul 16, 2009 10:47 am
Location: Almere, The Netherlands

Re: On-line engine blitz tourney June

Post by Joost Buijs »

hgm wrote: Sat Jun 23, 2018 11:11 pm

Code: Select all

:Tourney Players: Round 9 of 9 
:
:     Name              Rating Score Perfrm Upset  Results 
:     ----------------- ------ ----- ------ ------ ------- 
:  1 +Texel             [2252]  13.0  [2299] [   0] +02w +07w =03b +04w =05b +11w +09b +08b +12w 
:  2 +LeelaZero         [1883]  12.5  [2247] [ 752] -01b +20b +12w +07w +03b =04w +05w +11b +09w 
:  3 +ArasanX           [2092]  11.5  [2171] [  81] +13b +09w =01w =11b -02w +08w =04b +05w +07b 
:  4 -nemorino          [2094]  10.5  [2082] [   0] +10w =11w +05b -01b +07w =02b =03w =09w =08b 
:  5 -WaspX             [2044]  10.0  [1977] [ 104] +14b =08w -04w +09b =01w +06w -02b -03b +11w 
:  6 +rpiArminius       [1581]  10.0  [1837] [ 743] =08b +12w -11b =13w +14w -05b -10w +20b +18b 
:  7 -NightmareX        [2025]  9.5  [1893] [   0] +15w -01b +13w -02b -04b +20w =11b +10b -03w 
:  8 -WaDuuttie         [1974]  9.5  [1872] [  95] =06w =05b =14w +19b =11w -03b +15b -01w =04w
:  9 -Pedone            [1915]  9.5  [1884] [ 102] +16w -03b +15w -05w +20b +12b -01w =04b -02b 
: 10 +RuyDos            [1819]  9.5  [1746] [   0] -04b -19w +18b +14b -12w +13w +06b -07w =15w 
: 11 +rpiStockfish      [1941]  9.0  [1902] [ 209] +19w =04b +06w =03w =08b -01b =07w -02w -05b 
: 12 -Goldbar           [1928]  9.0  [1741] [   0] =20w -06b -02b +15w +10b -09w =14w +13b -01b 
: 13 -Marvin            [1815]  9.0  [1700] [   0] -03w +22w -07b =06b +19w -10b +20w -12w =14b 
: 14 +Schooner          [1664]  9.0  [1689] [ 362] -05w +16b =08b -10w -06b +21w =12b =18w =13w 
: 15 -tomitankChess     [1612]  9.0  [1688] [ 103] -07b +23w -09b -12b +22w +18b -08w =19w =10b 
: 16 +Spartacus         [1517]  9.0  [1523] [  57] -09b -14w -22b -21b =27w +26b +24w +23w +19b
: 17 +Embla             [1223]  9.0  [1630] [1013] +24b +25b +26w +30w -21b =23w +27b +22w +28w 
: 18 -RookieMonster     [1421]  8.5  [1643] [ 338] +25w +21b -10w +27w +23b -15w +22b =14b -06w 
: 19 +Floyd             [1574]  8.0  [1555] [ 264] -11b +10b =20w -08w -13b -22b +21w =15b -16w 
: 20 +Bliep             [1535]  8.0  [1611] [ 215] =12b -02w =19b +22w -09w -07b -13b -06w +21b 
: 21 +rpiMinkoChess     [1502]  8.0  [1431] [  15] -22b -18w +23b +16w +17w -14b -19b +26w -20w 
: 22 +rpiTogaII         [1483]  8.0  [1444] [ 144] +21w -13b +16w -20b -15b +19w -18w -17b +29w 
: 23 +rpiFruit          [1454]  8.0  [1346] [   8] +27b -15b -21w +25b -18w =17b +31w -16b +34w 
: 24 +microMax          [1245]  8.0  [1409] [ 255] -17w +28b -27b +26w +31b +32w -16b +33w +30b
: 25 -Isa               [1414]  7.5  [1275] [   0] -18b -17w +30b -23w -26b +34w +29b +32w +33b 
: 26 +KingSlayer        [1409]  7.5  [1340] [  58] +33b +27w -17b -24b +25w -16w +34b -21b +31w 
: 27 +Joker             [1462]  7.0  [1207] [  27] -23w -26b +24w -18b =16b +29w -17w -28b +32b 
: 28 +rpiEmbla          [1128]  6.5  [1319] [ 633] +31b -24w +34w +32b -29w +30b +33w +27w -17b 
: 29 +Weini             [1205]  5.5  [1225] [ 131] -30w =32w +33b +31w +28b -27b -25w +34b -22b 
: 30 +rpiPhalanx        [1175]  5.0  [1131] [ 191] +29b +33w -25w -17b +34w -28w =32b =31b -24w 
: 31 +Skiull            [1172]  3.5  [1067] [ 165] -28w +34b =32w -29b -24w +33b -23b =30w -26b 
: 32 +rpiBikjump        [1061]  3.5  [1126] [ 459] +34w =29b =31b -28w +33b -24b =30w -25b -27w
: 33 +Teki              [1336]  2.0  [ 778] [   0] -26w -30b -29w -34b -32w -31w -28b -24b -25w 
: 34 +POS               [ 797]  1.0  [ 950] [ 539] -32b -31w -28b +33w -30b -25b -26w -29w -23b 
:
:     Average Rating    1580.9 
:
Leela (LC0-20180622 with network 395) did surpricingly well, it was running on my old i7-5960x with GTX-1080ti. I've been watching most of the games, although Leela is tactically very weak it was able to positionally outmaneuver most other engines, only Texel with his (100 core ???) cluster was a little bit too much for it, against Nemorino it also had a big plus, but in the end it turned out to be no more than a draw.

It is clear to me that most a-b style engines have a weak evaluation function, maybe comparable to a human player with 1800 rating, combined with a very deep tactical search. For Leela it is the other way around, and it shows that with a better evaluation function tactics are getting of less importance. Now I wonder if there is a way to combine these two features (deep tactics with NN evaluation), and what it would do for playing strength.
petero2
Posts: 686
Joined: Mon Apr 19, 2010 7:07 pm
Location: Sweden
Full name: Peter Osterlund

Re: On-line engine blitz tourney June

Post by petero2 »

Joost Buijs wrote: Sun Jun 24, 2018 8:58 am Leela (LC0-20180622 with network 395) did surpricingly well, it was running on my old i7-5960x with GTX-1080ti. I've been watching most of the games, although Leela is tactically very weak it was able to positionally outmaneuver most other engines, only Texel with his (100 core ???) cluster was a little bit too much for it, against Nemorino it also had a big plus, but in the end it turned out to be no more than a draw.

It is clear to me that most a-b style engines have a weak evaluation function, maybe comparable to a human player with 1800 rating, combined with a very deep tactical search. For Leela it is the other way around, and it shows that with a better evaluation function tactics are getting of less importance. Now I wonder if there is a way to combine these two features (deep tactics with NN evaluation), and what it would do for playing strength.
Texel was using 180 cores (5 computers, 36 cores each) and searched almost 300MN/s in the early middle game, but still Leela won one bullet game (1m+1s/move) I played before the tourney began.

[Event "ICS rated lightning match"]
[Site "winboard.nl"]
[Date "2018.06.23"]
[Round "-"]
[White "LeelaZero"]
[Black "Texel"]
[Result "1-0"]
[WhiteElo "2178"]
[BlackElo "2273"]
[TimeControl "60+1"]
[Annotator "68. +24.75 8... -0.23"]

1. e4 c6 {+0.18/2} 2. Nf3 d5 {+0.20/2} 3. Nc3 dxe4 {+0.22/2} 4. Nxe4 Nf6
{+0.25/2} 5. Nxf6+ exf6 {+0.25/2} 6. Bc4 Bd6 {+0.28/2} 7. Qe2+ Qe7
{+0.26/2} 8. Qxe7+ {+0.25/2} Bxe7 {-0.23/23 1.1} 9. O-O {+0.30/2} O-O
{-0.19/24 0.4} 10. Re1 {+0.39/2} Bd6 {-0.14/23 1.6} 11. d4 {+0.39/2 5} Bf5
{-0.18/25 1.1} 12. c3 {+0.52/2 3} h6 {-0.10/22 1.9} 13. Nh4 {+0.65/2 3} Bh7
{-0.03/24 2.0} 14. g4 {+0.58/2 4} g5 {+0.00/25 3} 15. Nf5 Bxf5 {+0.85/2}
16. gxf5 {+0.85/2 4} Kg7 {-0.12/26 3} 17. a4 {+0.93/2 2.2} a5 {-0.03/25 3}
18. Bb3 {+0.88/2 2.5} Nd7 {+0.00/25 3} 19. Bd2 Rfd8 {+0.81/2} 20. h4
{+0.77/2 5} c5 {+0.00/23 4} 21. Be3 {+0.63/2 0.6} Rac8 {-0.02/23 2.8} 22.
Rac1 {+0.67/2 1.8} c4 {+0.00/24 4} 23. Bc2 {+0.66/2 0.9} Nb6 {+0.11/22 1.5}
24. Rb1 {+0.65/2 2.7} Nd5 {+0.00/23 2.6} 25. h5 {+0.58/2 4} Re8
{+0.31/25 2.9} 26. Kf1 {+0.52/2 3} Bf4 {+0.17/25 2.8} 27. Bxf4 {+0.72/2}
Nxf4 {+0.00/28 1.3} 28. Rxe8 Rxe8 {+0.78/2} 29. Bd1 {+0.86/2} Rd8
{-0.85/29 9} 30. Bf3 {+1.42/2} b6 {-0.74/28} 31. Ke1 {+1.69/2 0.7} Nd3+
{-0.90/27 1.9} 32. Kd2 {+1.83/2} Nxf2 {-1.21/27 2.5} 33. b3 {+1.98/2 0.4}
cxb3 {-1.33/27 2.0} 34. Ke3 {+2.08/2 0.6} Nh3 {-0.96/24 2.2} 35. Rxb3
{+1.99/2} Re8+ {-1.41/26 6} 36. Kd2 {+2.20/2 0.1} Nf2 {-1.44/25 2.3} 37.
Rxb6 {+2.30/2} g4 {-1.61/24 0.4} 38. Bc6 {+2.14/2 1.2} Rd8 {-2.26/26 1.5}
39. Rb1 {+3.58/2} Nh3 {-2.65/26 1.4} 40. Ke3 {+3.82/2 1.6} Rc8 {-3.31/26 4}
41. Rb6 {+3.97/2 2.1} Kh7 {-1.46/24 2.1} 42. d5 {+4.71/2} g3 {-3.30/26 2.6}
43. Rb2 {+5.11/2} Rg8 {-3.88/27 1.4} 44. Rg2 {+5.69/2} Nf2 {-4.20/27 1.0}
45. d6 {+5.74/2 1.5} Rd8 {-3.61/25 1.9} 46. d7 {+6.73/2 0.8} Ng4+
{-5.16/26 2.0} 47. Kd4 {+6.74/2} Ne5 {-5.64/28 2.2} 48. Kc5 {+6.64/2} Nxd7+
{-5.63/29 0.9} 49. Bxd7 {+6.53/2 1.3} Rxd7 {-6.16/29 0.6} 50. Rxg3
{+6.53/2 0.8} Rb7 {-6.99/28 4} 51. c4 {+6.49/2} Rb4 {-5.21/28 1.6} 52. Kd5
{+7.80/2 1.2} Rxa4 {-9.10/25 1.5} 53. c5 {+7.97/2 1.6} Ra1
{-1006.57/27 1.5} 54. c6 {+8.05/2} a4 {-1005.67/27} 55. Rc3 {+9.35/2 5} a3
{-1006.53/21 1.7} 56. Kc4 {+10.18/2 2.4} a2 {-1006.37/18 1.5} 57. Kb3
{+11.15/2 2.3} Rh1 {-1004.74/18 0.9} 58. Kxa2 {+15.50/2 1.7} Rxh5
{-1001.71/18 0.3} 59. c7 Rxf5 {+18.28/2} 60. c8=Q {+20.29/2 2.0} Ra5+
{-1000.63/16 1.6} 61. Kb3 Rb5+ {+20.12/3} 62. Kc2 Kg7 {+21.13/3} 63. Qd7
{+26.13/2 0.5} Ra5 {-1000.12/17 1.6} 64. Rc7 Kg6 {+29.11/2} 65. Qxf7+ Kg5
{+22.29/2} 66. Qg7+ {+20.71/2} Kf5 {-1000.08/16 0.4} 67. Qxh6 {+22.38/2}
Ra2+ {-1000.09/15 1.2} 68. Kd3 {+24.75/3} Ra3+ {-1000.11/14 0.5} 69. Kd4
Ra4+ {+25.97/3} 70. Rc4 Rxc4+ {+21.75/3} 71. Kxc4 Ke5 {+22.13/2} 72. Qh8
{+23.05/2} Kf5 {-1000.08/17 0.5} 73. Kd5 {+25.12/2} Kf4 {-1000.07/15 0.1}
74. Qxf6+ Ke3 {+28.22/3} 75. Qd4+ {+28.83/2} Kf3 {-1000.06/13} 76. Ke5
{+32.66/2 1.7} Kg3 {-1000.05/11} 77. Ke4 Kg2 {+39.57/2} 78. Ke3
{+45.29/2 0.2} Kg3 {-1000.03/7} 79. Qg7+ Kh2 {+43.78/2} 80. Kf3 Kh1
{+65.03/2} 81. Qg2#
{Texel checkmated} 1-0
Joost Buijs
Posts: 1563
Joined: Thu Jul 16, 2009 10:47 am
Location: Almere, The Netherlands

Re: On-line engine blitz tourney June

Post by Joost Buijs »

petero2 wrote: Sun Jun 24, 2018 11:51 am Texel was using 180 cores (5 computers, 36 cores each) and searched almost 300MN/s in the early middle game, but still Leela won one bullet game (1m+1s/move) I played before the tourney began.
Very nice hardware!

Maybe bullet is not the strongest point of such a large cluster, I assume you have to deal with large latencies.

I'm still using the i7-6950x that I build 1.5 years ago, by the end of this year I want to upgrade to somewhat largers, it depends on what Intel will deliver for the high end consumer market at that time. Intel has to compete with AMD threadripper (32 cores), I assume Intel is going to deliver something comparable. AMD processors I don't like very much, overclocked factory wise with almost no headroom, the bad performance of PDEP/PEXT is also a drawback.

First I have to rework my engine, after 9 years of experiments the code became a mess, the evaluation function is at least 7 years old and still manually tuned. I'm working on a new engine, it goes slowly, each time I change my mind about the data structures, there are so many ways to solve things, sometimes it makes me crazy. LOL