2013 Tournament between Android Engines

[lucasart]

I am planning to run from both sides of Albert Silver's Silversuite positions, or perhaps the Silversuite gambit positions. Does that sound interesting enough?

Sure it will be interesting! The lines are not too long, and represent a varied overview of chess theory. I don't know if all opening lines are good (ie. tested by strong engine for any tactical mistake), but I'm guessing they must be allright as they are on chessbase (someone would have reported errors if any).

[jd1]

How did Critter 1.2 loose a game to DiscoCheck ? Did something go wrong (lost on time for example) ?

Here is the game, Lucas.

[pgn]
[Event "Chess for Android Tournament"]
[Site "United States"]
[Date "2013.02.09"]
[Round "2"]
[White "DiscoCheck 4.0.1"]
[Black "Critter 1.2 32-bit"]
[Result "1-0"]
[PlyCount "97"]
[TimeControl "1/1"]
[Device "Nexus 7"]

1. e4 c5
2. Nf3 e6
3. d4 cxd4
4. Nxd4 Nc6
5. Nb5 Nf6
6. N1c3 d6
7. Bf4 e5
8. Bg5 a6
9. Na3 b5
10. Bxf6 gxf6
11. Nd5 f5
12. c3 Bg7
13. exf5 Bxf5 {from PGN 0.008}
14. Bd3 {[12]17/ 1.76} e4 {[9/14]24/d3c2 1.081}
15. Be2 {[11]-17/ 1.918} O-O {[9/11]45/e1g1 1.366}
16. O-O {[11]-17/d8h4 1.633} Ne7 {[8/10]34/a3c2 1.434}
17. Nc2 {[11]7/f5d7 d1d2 g7e5 d 1.573} Be6 {[8/12]12/d5e7 1.463}
18. Nxe7+ {[11]29/d8e7 a2a4 b5a4 1.526} Qxe7 {[9/16]-4/a2a4 0.997}
19. a4 {[11]35/e7d7 c2d4 e6c4 e 1.007} Qc7 {[9/14]-12/c2d4 0.994}
20. Nd4 {[11]54/b5b4 c3b4 d6d5 a 2.009} b4 {[8/13]-19/c3b4 1.276}
21. cxb4 {[11]53/d6d5 a1c1 c7f4 0.72} Be5 {[9/16]-13/g2g3 1.004}
22. Rc1 {[13]98/ 1.996} Qb7 {[9/16]-51/d4e6 1.092}
23. Qd2 {[11]117/e6d7 1.91} Bd7 {[9/12]-46/b4b5 1.303}
24. b5 {[11]99/a6b5 d4b5 b7b6 b 1.696} axb5 {[8/13]-46/d4b5 1.022}
25. Nxb5 {[10]103/b7b6 b2b3 g8h8 d 0.973} Qb6 {[9/16]-53/c1c4 1.425}
26. b3 {[11]119/g8h8 1.58} Be6 {[8/10]-76/e2c4 1.785}
27. Bc4 {[10]139/f8c8 c4e6 f7e6 d 1.211} Qd8 {[8/14]-113/c4d5 1.777}
28. Bd5 {[10]181/a8a6 d5b7 a6a5 b 1.225} Ra6 {[9/13]-133/d5e6 1.751}
29. Bxe4 {[9]148/ 0.254} Bxh2+ {[8/14]-164/g1h2 1.235}
30. Kxh2 {[11]233/d8h4 0.76} Qh4+ {[10/21]-245/h2g1 1.614}
31. Kg1 {[]/ 0.002} Qxe4 {[10/17]-246/c1c3 1.048}
32. Rc3 {[11]289/e4e5 c3g3 g8h8 f 1.335} Bf5 {[9/15]-238/b5c7 1.476}
33. Nc7 {[11]310/a6a7 d2d6 a7a5 c 1.528} Ra7 {[9/15]-291/f1e1 2.112}
34. Qxd6 {[10]322/ 0.888} Bg6 {[8/10]-239/c7d5 1.432}
35. Rfc1 {[10]321/a7a5 c3e3 e4h4 c 1.565} Ra5 {[9/14]-235/c7b5 1.411}
36. Re3 {[11]330/e4h4 c7d5 h4d8 d 1.592} Qb7 {[9/19]-272/c7b5 1.221}
37. Ne8 {[11]354/b7a6 e8f6 g8g7 d 1.78} Ra6 {[9/16]-305/d6d4 1.469}
38. Qd4 {[12]362/ 1.527} f6 {[9/21]-375/c1c7 1.064}
39. Rc7 {[12]711/b7a8 c7g7 1.933} Qa8 {[8/15]-393/d4d7 1.399}
40. Rg7+ {[12]721/g8h8 g7g6 a8a7 g 1.606} Kh8 {[10/20]-547/d4d7 1.025}
41. Rxg6 {[14]744/a8a7 g6f6 f8f6 d 1.972} Rf7 {[7/15]-913/e8f6 1.007}
42. Rxf6 {[10]1227/a6f6 e8f6 h7h6 e 0.996} Raxf6 {[9/19]-1296/e8f6 1.028}
43. Nxf6 {[10]2129/h7h6 d4f4 f7h7 e 1.725} h6 {[9/21]-1323/b3b4 1.024}
44. Qf4 {[11]18848/f7h7 e3e7 a8g8 e 1.225} Qxg2+ {[9/19]-2065/g1g2 1.295}
45. Kxg2 {[]/ 0.002} Rg7+ {[12/19]mate:-4/g2h3 1.079}
46. Rg3 {[51]31993/g7g5 g3g5 h6g5 f 1.619} Rg5 {[25/6]mate:-3/f4c7 2.036}
47. Rxg5 {[49]31992/ 0.963} hxg5 {[56/4]mate:-2/f4c7 1.027}
48. Qc7 {[127]31997/g5g4 0.097} g4 {[56/1]mate:-1/ 0.067}
49. Qh7# 1-0
[/pgn]

OK, so it won fair and square. Critter seems to have accidently overlooked some tactics.

If I may suggest, could you cap the booklines to 7 or 8 moves let's say ? This will allow engines to develop their own strategy, rather than start the game from an already boring position with developpement finished and many pieces exchanged.

Very nice, Lucas!

Jerry

[abik]

Continued test results. Engine authors are free to appeal a disqualification with an updated binary before the qualifying tournament starts. Qualified engines can be updated anytime during the tournament.

Code: Select all

                     1   2  3  4  5
1   Toga II 1.4.1SE  ** 11 11 11 11   8.0/8
2   Sungorus 1.4 JA  00 ** 11 11 11   6.0/8
3   ZCT-0.3.2500     00 00 ** ½1 11   3.5/8
4   TJchess 1.01U    00 00 ½0 ** 11   2.5/8
5   VIRUTOR CHESS    00 00 00 00 **   0.0/8

Disqualified: rocinante-101-ja (hangs), Umko_1.2_arm (hangs).

                         1  2  3  4  
1   danasah-488-ja       ** 01 11 11   5.0/6  9.00
2   gk-090-ja            10 ** 11 11   5.0/6  9.00
3   exchess-054-beta-ja  00 00 ** 1½   1.5/6
4   fairymax48l-ab       00 00 0½ **   0.5/6

Disqualified: amundsen (illegal moves), bruja (hangs), buzz-os-11-27-08-ja (crashes), chess-one-201-ja (crashes), dirty-android (old compilation), gfc-08-rev5-ja (claims win without mating move).

                             1  2  3  4  5  6  
1   kmtchess-121-ja          ** 1½ 11 ½1 01 11   8.0/10
2   natwarlal-014-ja         0½ ** ½1 1½ 11 11   7.5/10
3   gullydeckel-215-pl10-ja  00 ½0 ** 01 11 11   5.5/10
4   mskcp-145-ja             ½0 0½ 10 ** 10 10   4.0/10  18.25
5   mizar-30-ja              10 00 00 01 ** 11   4.0/10  14.00
6   myrddin0.85j             00 00 00 01 00 **   1.0/10

Disqualified: gullydeckel-215-pl10-ja (uses too much time), knightcap-36-ja (crashes), komodo3 (old compilation), mizar-30-ja (uses too much time), mskcp-145-ja (uses too much time).

[abik]

Final set of test results.

Code: Select all

                      1  2  3  4  5  6  
1   scorpio-27-ja     ** 10 ½1 11 11 11   8.5/10
2   phalanx-xxii-ja   01 ** 01 ½1 11 11   7.5/10
3   sloppy-023-ja     ½0 10 ** 1½ 11 11   7.0/10
4   scidlet-261b2-ja  00 ½0 0½ ** 11 11   5.0/10
5   sjeng-112-ja      00 00 00 00 ** 11   2.0/10
6   olithink-531-ja   00 00 00 00 00 **   0.0/10

                              1  2  3  4  5  6  
1   tucano104-ab              ** 10 11 11 11 11   9.0/10
2   surprise-43-b13-ja        01 ** ½0 11 11 11   7.5/10
3   tscp181-ab                00 ½1 ** 0½ 01 11   5.0/10
4   umax48w                   00 00 1½ ** 1½ ½1   4.5/10
5   zzzzzz-351-ja             00 00 10 0½ ** 11   3.5/10
6   toledo-nanochess-2010-ja  00 00 00 ½0 00 **   0.5/10

Disqualified: surprise-43-b13-ja (uss too much time), zeta-dva-0202-ja (illegal moves).

[abik]

The test runs are over, and the qualified and disqualified engines are shown below. Next to follow is a qualification tournament which will place the top 25 engines in six groups, like I did in the 2011 tournament. However, I will wait a few days to see whether engine authors want to appeal the disqualification with an improved binary. Updates of qualified engines are of course also welcome.

Qualified:

• AdroitChess 0.3
  Arasan 14.0.1
  betsabe-II-130-android-ja
  BikJump v2.1P (32-bit)
  cheng3 1.07 JA
  crafty-235-android-ja
  Critter 1.6a 32-bit
  danasah-488-ja
  Daydreamer 1.75 JA
  Diablo 0.5.1
  DiscoCheck 4.0.1
  exchess-054-beta-ja
  fairymax48l-ab
  GarboChess 3 (32-bit)
  Gaviota v0.86
  gk-090-ja
  GNU Chess 6.0.2
  greko98-ab
  Jazz 6.40 JA
  kmtchess-121-ja
  myrddin0.85j
  natwarlal-014-ja
  olithink-531-ja
  Pepito v1.59
  phalanx-xxii-ja
  RedQueen 1.1.2
  Rodent 0.18.0
  Rotor 0.7
  scidlet-261b2-ja
  scorpio-27-ja
  Simplex 0.9.8
  sjeng-112-ja
  sloppy-023-ja
  Stockfish 2.3.1
  Sungorus 1.4 JA
  TJchess 1.01U
  Toga II 1.4.1SE
  toledo-nanochess-2010-ja
  tscp181-ab
  tucano104-ab
  typhoon-android-ja
  umax48w
  VIRUTOR CHESS
  ZCT-0.3.2500
  zzzzzz-351-ja

Disqualified:

• amundsen (illegal moves)
  apilchess-106-ja (hangs)
  bruja (hangs)
  buzz-os-11-27-08-ja (crashes)
  cassandre-026-ja (sends wrong format)
  chess-one-201-ja (crashes)
  cyrano-061-b17-ja (hangs)
  dirty-android (old compilation)
  gfc-08-rev5-ja (claims win without mating move)
  gullydeckel-215-pl10-ja (uses too much time)
  ifrit-m18-ja (illegal moves)
  iota (uses too much time)
  knightcap-36-ja (crashes)
  komodo3 (old compilation)
  mizar-30-ja (uses too much time)
  mskcp-145-ja (uses too much time)
  pawny-031-ja (illegal moves)
  rocinante-101-ja (hangs)
  sissa-20-ja (illegal moves)
  suff-03a-ja (uses too much time)
  umko_1.2_arm (hangs)

[kinderchocolate]

Aart, sounds interesting. Please update results when you have it. I'll watch this space closely.

[abik]

The first qualification round has started. It will be played from both sides of the second position found in the Silver Suite.

[d] rnbq1rk1/ppp1ppbp/3p1np1/8/4P3/3P1NP1/PPP2PBP/RNBQ1RK1 b - e3 0 6

[kinderchocolate]

The first qualification round has started. It will be played from both sides of the second position found in the Silver Suite.

[d] rnbq1rk1/ppp1ppbp/3p1np1/8/4P3/3P1NP1/PPP2PBP/RNBQ1RK1 b - e3 0 6

So, you aren't starting from the starting position? Any particular reason?

[abik]

So, you aren't starting from the starting position? Any particular reason?

Several reasons. First, many engines for Android don't ship with their own book, so this avoids putting engines with a book directly at an advantage. Second, playing from both sides of a fixed series of opening lines gives a nice symmetric, yet varied set of games. Third, it is the same format I used in 2011, and I like to keep the tradition. Last, call me a romantic, but I enjoy the fight between actual engines rather than pre-computed books, although I realize that here opinions may vary.

Note that test and qualify tournaments are played from both sides of a single opening line. The actual group A to F tournaments will be played from both sides of a full set of opening lines, also at longer time controls.

[lucasart]

So, you aren't starting from the starting position? Any particular reason?

Several reasons. First, many engines for Android don't ship with their own book, so this avoids putting engines with a book directly at an advantage. Second, playing from both sides of a fixed series of opening lines gives a nice symmetric, yet varied set of games.

Completely true. Another important reason is that w/o book, playing from the start pos, you'll observe the same games all the time. Engine A can only play engine B twice (with and black), after that it's going to be a repeat of the same game

Last, call me a romantic, but I enjoy the fight between actual engines rather than pre-computed books

Me too! There's nothing more annoying than these excessively long book lines that forces the engines into a drawish Karpov-like game with lots of pieces exchanged and symetric pawns structures (just to make sure it's going to be dead drawn and absolutely boring).
Why can't book authors limit the size of their opening books more ? Privilege variety (ie. higher branching factor in the book tree) over very long lines.