Our Best of the Web includes the 2018 Canadian Junior, the preliminary rounds of the world's top computer chess engine tournament, and a rapid and blitz tournament featuring 9 of the world's top 14 players.... as well as two games!
The diagram comes from the round 2 game bewteen Hiran Liang and Brandon Zhu at the 2018 Canadian Junior Chess Championship. White to play (solution(s) in game viewer below).
The second game in the viewer is Leela Zero's smooth win in the TCEC Division 4 RR.
2018 Canadian Junior Championship
The Canadian Junior Chess Championship runs August 8-12, 2018 at Humber College, Toronto.
It features four sections: a 10-player invitiational RR, Open Swiss, U1800, and U1300. Male and female winners of the 10-player RR represent Canada at the World Junior in Turkey (Sept 4-16), and get $1500 for travel.
The diagram at the top of the page comes from the round 2 game bewteen Hiran Liang and Brandon Zhu at the Canadian Junior. White to play (solution(s) in game viewer below).
homepage
https://elevatemychess.com/canadian-junior-chess-championship/
live games (top 3 boards in Invitational Section)
http://view.livechesscloud.com/a3a3c64b-2331-456c-af1b-6761974d5b15
Standings
http://chess-results.com/tnr370772.aspx?lan=1&art=0
TCEC Season 13
The only chess engine competition that matters is underway. It features 32 chess engines competing on identical hardware (almost, see below) with identical short opening books. The top engines, including the perennial top 3 Stockfish, Komodo, and Houdini -- are all seeded directly into the Premier Division. Other engines have to qualify via round-robins from one of the four lower divisions, which compete in 4xRRs with a TC of 30 minutes + 10 seconds per game.
This year the hardware cannot be identical, since, for the first time Neural Net engines will be in the competition. Neural Nets are highly parallelized programs, and are typically run on specialized hardware, or on the GPUs in high-end graphics cards. For Season 13, standard chess engines will run on a 44 core CPU, while the two Neural Net engines run on a remote machine with 2x GTX 1080ti GPUs.
Leela
An online group of programmers and chess enthusiasts is attempting to reproduce the success of Google's AlphaZero by creating an open-source self-taught neural net program, called Leela Zero (L0), based (to the extent possible on the limited informtion released by Google Mind) on what is know about AlphaZero.
Leela uses distributed processing (including free processing time via Google Colab) to create millions of self-play games, asesses positions based on their winning probability as shown by Monte Carlo Tree Searches, and revises itself constantly, with a newer version being posted on their site approximately every 8 hours.
Leela's progress has been almost constant -- there were setbacks, e.g. when the programmers found errors in the 50-move rule implementation -- but uneven: Leela frequently played wonderful chess, only to allow a mate-in-three! This is because, unlike the typical alpha-beta engines, Leela doesn't analyze all possible moves at a short distance (and so misses some tactics). This uneven progress has revealed a split among Leela's supporters: the ones I'll call the "science types" think of Leela as an experiment to test the ideas revealed by the AlphaZero developers, while the "chess nuts" want Leela to get better faster, and think the undirected self-play method used by AZ and L0 could be sped up by training it on tactics and table-base positions.
A result of this split is that there was some controversy when a second NN engine was added to the tournament, DeusX, which turns out to be a derivative of Leela, using some guided training chosen by Albert Silver.
As it turns out, both Leela and DeusX qualified from the Division 4 group and move a step closer to the Premier Division:
TCEC 13: Division 4 Results
- LCZero 20/28(+14 =12 -2)
- DeusX 1.0 18.5(+13 =11 -4)
- Wasp 3.2 18.0(+11 =17 -2)
- Rodent III12.5(+4 =17 -7)
- Senpai 2.012.5(+3 =19 -6)
- Chess22k 1.1011
- Tucano 7.0510
- Ivanhoe 999946h 9.5
With new versions of Leela released every day, the next round-robin might feature a significantly stronger version: the current Leela Network seems to be approximately 50 Elo points stronger than the one that entered TCEC last week. All developers are able to update their engines each time they advance to the next level, with the caveat that any glitches introduced by the newer version cannot be fixed or replaced once that round-robin has begun.
The game viewer below shows one of Leela's wins. It is a remarkable example of spatial control, as well as a move (Bxa7, that almost all humans would automatically discount).
TCEC
http://www.chessdom.com/tcec-season-13-the-advance-of-the-nns/
live games (games run 24/7)
http://tcec.chessdom.com/season13/live.php
Leela
http://lczero.org/
St.Louis Rapid & Blitz
The third stage in the 2018 Grand Chess Tour hits St.Louis August 11-15, 2018. It features two tournaments back-to-back:
- Rapid: a round robin played at 25 min + a 10 second delay.
- Blitz: a double round-robin at 5 min + 3 sec delay.
Players, by FIDE world ranking, are:
- 2. Caruana
- 3. Mamedyarov
- 6. So
- 8. MVL
- 9. Nakamura
- 10. Karjakin
- 12. Anand
- 13. Aronian
- 14. Grischuk
- 25. Lenier Dominguez
No Magnus? No worries: he bumps out Dominguez and joins the others to play in the Sinquefield Cup: August 18-28, 2018.
live games
StLCC (Seirawan, Ashley, etc.)
https://grandchesstour.org/2018-grand-chess-tour/2018-saint-louis-rapid-blitz
GMs Hansen and Hambleton
https://www.twitch.tv/chessbrah
Click the arrow button next to the players' names to choose game:
..
[Event "TCEC Season 13 - Division 4"]
[Site "http://tcec.chessdom.com"]
[Date "2018.08.08"]
[Round "25.3"]
[White "LCZero 16.10161"]
[Black "Chess22k 1.10"]
[Result "1-0"]
[ECO "B56"]
[WhiteElo "3219"]
[BlackElo "3072"]
[Annotator "John Upper"]
[PlyCount "144"]
[EventDate "2018.??.??"]
1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 Nc6 6. f3 e5 7. Nb3 Be7 8. Be3
O-O 9. Qd2 a5 10. Bb5 Na7 {[#]There are over 300 human-played games in
MegaBase from here -- many featuring top-level players -- almost all of which
have continued by retreating the Bb5 to either d3 or e2.} 11. Bxa7 $5 {When I
first saw it I thought this was a bad move -- giving up the DSB looks
completely anti-positional -- but the move has its positonal pluses too: 1)
keeping control over b5 means Black can get play against White's K after 000
only by playing an exchange sac on c3, which is expensive; 2) Black won't be
able to force through ...d5, so the only square his minor pieces might look to
use is c5, and they can't all use it.} (11. Be2 Nc6 12. a3 a4 13. Nc1 Be6 14.
Nd5 Bxd5 $1 15. exd5 Nd4 16. Bxd4 exd4 17. Qxd4 Qa5+ 18. Qd2 Qb6 $5 (18... Nxd5
$15) 19. c4 Nd7 20. Nd3 Nc5 (20... Bf6 $142) 21. Nxc5 Qxc5 $11 22. Qb4 Qe3 23.
Qd2 Qc5 24. Qb4 Qe3 25. Qd2 {½-½ Piorun,K (2655)-Hracek,Z (2599) Jurmala,
2016.}) (11. Ba4 $6 {(the actual game had 11.Be2 Nc6 12.Bb5 Na7 inserted)} Be6
12. Bxa7 Rxa7 13. Qf2 Ra8 14. O-O-O Qb8 $15 15. Bb5 $2 Rc8 (15... a4 $1) 16.
Na4 (16. a4 $5 Rxc3 $5) 16... d5 17. Nb6 a4 $1 18. Bxa4 Qa7 $1 $19 19. exd5 Bd8
(19... Nxd5 $1) 20. dxe6 Bxb6 21. exf7+ Kxf7 22. Nc5 Qxa4 {0-1 Anand,V (2781)
-Kramnik,V (2751) Monte Carlo (blindfold), 1999.}) 11... Rxa7 12. O-O-O Qb6 13.
a4 {At 26 ply, Stockfish 9 rates this position as 0.00} Ra8 (13... Be6 14. Kb1
Rd8 15. g4 Raa8 16. g5 Nh5 17. h4 Nf4 18. Nd5 Nxd5 19. exd5 Bd7 20. Bxd7 (20.
c4 $5) 20... Rxd7 21. Rhe1 Qa6 22. Re4 f5 23. gxf6 Bxf6 24. c4 $14 {0-1 (54)
Tomczak,J (2596)-Czarnota,P (2557) Warsaw 2017}) 14. Qe2 Be6 15. Nd2 Rfc8 16.
Nc4 Qc5 17. Ne3 Rc7 18. Ned5 (18. g4 $5) 18... Nxd5 19. Nxd5 Bxd5 20. Rxd5 Qb4
21. Kb1 Rac8 22. c4 Rc5 23. Rd3 $1 R5c7 24. g3 Kf8 {Black has nothing
constructive to do, while White can triple on the d-file and/or prepare to
open a file with f3-f4. Any file opening on the kingside will doubly benefit
White, who has more space to swing major pieces that way, and because Black's
K will be a target.} 25. Qc2 Kg8 26. Rb3 Qc5 27. Qg2 Bg5 28. Re1 Rd8 29. h4 Bh6
30. Ra3 {I don't know if I would have to be a better or a worse player to
understand this move. It doesn't LOOK like zugzwang....} Qd4 31. Qf1 Qc5 32.
Rd3 Qb6 {[#]} 33. f4 $1 exf4 34. gxf4 (34. g4 $2 g5) 34... Rcc8 35. h5 Qc7 36.
Qf3 Qe7 37. Red1 Qh4 38. e5 $1 dxe5 39. Qg4 $1 {remember Adams - Torre? Leela
doesn't.} Qe7 (39... Qxg4 $4 40. Rxd8+ Rxd8 41. Rxd8#) 40. fxe5 Rf8 (40... Rxd3
$4 41. Qxc8+ $18) 41. Rd5 Rc7 42. Qe4 Rc5 43. Re1 b6 44. Rd7 Qe6 45. Rd6 Qh3
46. Bc6 {Continuing to dominate the Bh6. Leela's piece coordination to control
space is just teriffic.} (46. Rxb6 Bd2 47. Re2 Qf1+ 48. Ka2 Bb4 49. h6 $18)
46... Qxh5 47. Bd5 Rc7 (47... Bd2 $2 48. Rh1 $18) 48. e6 fxe6 49. Rxe6 $18 Kh8
50. Rxb6 Rcc8 51. Rh1 Qg5 52. Qd4 Rf4 53. Qc3 Qf5+ 54. Ka2 Bg5 55. Be6 Qf6 56.
Qxa5 Rd8 57. Qb5 Qe7 58. Bd5 Rf2 59. Rb7 Qd6 60. Be4 Bh6 61. Qb6 Qxb6 62. Rxb6
Re2 63. Bd5 Bd2 64. Rb5 h6 65. a5 Rd7 66. a6 Ra7 67. Bb7 Bf4 68. c5 Re6 69. Kb3
Kh7 70. Rb6 Re2 71. Kc4 Bc7 72. b4 Bf4 {game was adjudicated a win for White
based on the TCEC win rules.} 1-0
[Event "Canadian Junior"]
[Site "Humber College"]
[Date "2018.08.09"]
[Round "2.1"]
[White "Liang, Hairan"]
[Black "Zhu, Brandon"]
[Result "1-0"]
[ECO "B11"]
[Annotator "John Upper"]
[PlyCount "49"]
1. e4 c6 2. Nf3 d5 3. Nc3 Bg4 4. h3 Bxf3 5. Qxf3 e6 6. Be2 Nf6 7. O-O Be7 8.
Rd1 O-O (8... b5 $2 9. e5 Nfd7 10. d4 a6 11. Nb1 $5 (11. Qg3 {White attacks
before Black gets play against d4.}) 11... c5 12. c3 Nc6 13. Qg4 g6 14. Bh6 $14
{(1-0, 36) Vachier Lagrave,M (2796)-Topalov,V (2749) Paris (blitz) 2017.}) 9.
d4 Kh8 $6 (9... Nbd7 10. Be3 dxe4 11. Nxe4 Nxe4 12. Qxe4 Nf6 13. Qf3 Nd5 14.
Bd2 Bf6 15. c3 Qb6 16. Rab1 Rfd8 17. Bd3 $14 {½-½ (41) Trifunovic,P-Filip,M
Oberhausen, 1961.}) 10. a3 a5 11. Rb1 a4 12. Be3 b5 13. Bd3 Qa5 $2 (13... dxe4
$142 14. Nxe4) 14. e5 $1 $16 Ng8 15. Ne2 g6 16. h4 $1 Bxh4 (16... h5 17. Nf4 {
and White sacs a piece on one of the light squares.}) 17. g3 $18 Be7 18. Kg2
Qd8 19. Rh1 Bg5 20. Nf4 (20. Qg4 {also wins} Bxe3 21. fxe3 Nd7 22. Rxh7+ Kxh7
23. Rh1+ Kg7 24. Nf4 Qe7 (24... Re8 25. Bxg6 $18) 25. Nh5+) 20... Kg7 {Black
has expanded on the queenside, but White has 5 of his 6 pieces attacking
Black's K.} (20... Ra7 21. Bxg6 (21. Rxh7+ $1 {also wins.}) 21... fxg6 22.
Nxg6+ Kg7 23. Nxf8 Rf7 24. Rxh7+ $4) 21. Nxe6+ $1 {White is spoiled for choice,
as all plausible sacs win:} (21. Rxh7+ {also wins} Kxh7 22. Nxg6 Nh6 23. Ne7+
Kh8 24. Bxg5 $18) (21. Bxg6 $1 {also wins} hxg6 (21... Bxf4 22. Rxh7+ Kxg6 23.
Qh5#) (21... fxg6 22. Nxe6+ Kh8 23. Nxf8 $18 (23. Rxh7+ $18)) 22. Nh5+ $1 gxh5
23. Qxh5 $18) 21... fxe6 22. Rxh7+ $1 Kxh7 23. Qh5+ Bh6 24. Qxg6+ Kh8 25. Qh7#
1-0
merida
46
..