WCRCC Finals

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
CRoberson
Posts: 2005
Joined: Mon Mar 13, 2006 1:31 am
Location: North Carolina, USA
Contact:

WCRCC Finals

Post by CRoberson » Mon Jul 23, 2007 4:03 am

The web pages are updated with the Cross table and PGN.

http://www.taccl.org

Thanks to all the authors for their participation.

There were dissapointments and surprises.

My biggest dissappointment was not having an exciting tie breaker.
I expected that to be great to watch.

Surprises:
Zap/Erdo forfeits first round (happened in CCT, is this a trend)
At least 3 programs performed way above expectation.

Tony Thomas

Re: WCRCC Finals

Post by Tony Thomas » Mon Jul 23, 2007 4:23 am

Which engines performed better than expected?

swami
Posts: 6536
Joined: Thu Mar 09, 2006 3:21 am

Re: WCRCC Finals

Post by swami » Mon Jul 23, 2007 4:37 am

Tony Thomas wrote:Which engines performed better than expected?
Probably Dirty,Rascal and Mediocre and maybe Berta
Link:http://www.taccl.org/2007PresResults.html

User avatar
hgm
Posts: 23772
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: WCRCC Finals

Post by hgm » Mon Jul 23, 2007 7:42 am

Tony Thomas wrote:Which engines performed better than expected?
Well, microMax certainly performed better than I expected. Especially if you take into account that uMax doesn't ponder, which disadvantages it compared to the CCRL conditions under which its rating is measured.

It won against Roce and Vicki (normal), but also against Bird (an engines I did not know), and against Telepath (which had an ICC rating of 2343). It scored a draw against Parrot, (which is well above microMax' league). It was a bit lucky that it got a free point against Timea (we later played the game, and uMax lost...), while a game against Matilde, which would potentially have been very interesting, was interrupted by a failing connection, and lucky that the opponent later agreed to substitute a draw for the forfeit (there was no time left to finish the game before the next round). In addition it had very reasonable games against Tinker and Clarabit, where it both was a pawn ahead, but still lost in the end-game.

Especially the win against Telepath was remarkable:

Code: Select all

[Event "ICC tourney 772 (28 1 u)"]
[Site "Internet Chess Club"]
[Date "2007.07.22"]
[Round "13"]
[White "Telepath"]
[Black "microMaX"]
[Result "0-1"]
[WhiteElo "2366"]
[BlackElo "2000"]
[ICCResult "White resigns"]
[Opening "Ruy Lopez"]
[ECO "C70"]
[NIC "RL.09"]
[Time "13:17:55"]
[TimeControl "1680+1"]

1. e4 Nc6 2. Nf3 e5 3. Bb5 a6 4. Ba4 Bd6 5. O-O Nf6 6. Bxc6 dxc6 7. d4 Nxe4 8. dxe5 Be7 9. Qxd8+ Bxd8 10. Nd4 O-O 11. Re1 f5 12. Nd2 Nxd2 13. Bxd2 c5 14. Ne2 Re8 15. Nf4 Bg5 16. Rad1 c6 17. Re2 b6 18. Rde1 Ra7 19. Nh3 Bxd2 20. Rxd2 Rae7 21. f4 Be6 22. a3 h6 23. g3 a5 24. Rd6 Rc7 25. Nf2 a4 26. Nd1 Rcc8 27. Nc3 b5 28. Ne2 Rc7 29. Nc1 c4 30. Ne2 Bd7 31. Kf1 Re7 32. Rd2 Rb7 33. Nc3 Re8 34. Rd6 Re6 35. Red1 Re7 36. h3 Rc7 37. Re1 Re6 38. Kg1 Ra7 39. Re2 Kf7 40. Red2 Re7 41. Kg2 Rb7 42. Rd1 Rc7 43. Re1 Be6 44. Rd2 g6 45. Rde2 Rb7 46. Kg1 Re8 47. Kh2 Ra8 48. Rd1 Rh8 49. Rd6 Rc8 50. Red2 Rcc7 51. Na2 Rc8 52. Nb4 Rbc7 53. Na6 Re7 54. Nc5 Bd5 55. Re2 Rce8 56. c3 Rh8 57. h4 Kg7 58. Re3 Rg8 59. Kg1 Rc8 60. Kf2 Rce8 61. Re1 Rc8 62. Re2 Bf7 63. e6 Be8 64. Re5 Rb8 65. Rd2 Ra8 66. Kf1 h5 67. Re1 Rc7 68. Ree2 Re7 69. Rd1 Rb8 70. Re5 Ra8 71. Kg2 Rc7 72. Kf3 Re7 73. Kf2 Raa7 74. Rd8 Rac7 75. Kg1 Ra7 76. Re3 Kf8 77. Kf2 Rg7 78. Re1 Ke7 79. Rdd1 Rg8 80. Kg1 Rh8 81. Re2 Rg8 82. Rde1 Rh8 83. Rd1 Rg8 84. b4 axb3 85. Rde1 Rxa3 86. Kg2 b4 87. cxb4 c3 88. Nxb3 Rxb3 89. Re5 Rxb4 90. Kf3 Rc4 91. Ke3 c2 92. Kd2 c5 93. Kc1 Ba4 94. Rd5 Rb8 95. Rd7+ Bxd7 96. exd7+ Kxd7 97. Kd2 Rb3 98. Rc1 Rxg3 99. Ke2 Rd4 100. Kf2 Rh3 101. Kg2 Rc3 102. Kf2 Rd1 103. Ke2 Rxc1 104. Kd2 Rb1 105. Ke2 {result 0-1 white resigns}
It is interesting to see how the poor-man's substitute for handling passers works out in microMax. Recognizing real passers takes way too much code to ever pay off in terms of Elo per character, so uMax in stead attaches high scores to having Pawns on 6th and 7th rank (which get piece values 160 and 240 cP, respectively). A Pawn on the 7th is of course always a passer, but even if they are still blocked, Pawns on the 6th are sufficiently dangerous, and a sufficiently big annoyance for the opponent, to warrant the hefty bonus.

So although uMax still unwittingly allows passers to be created for the opponent, like on move 11 of the game above,
[d]r1bb1rk1/1pp2ppp/p1p5/4P3/3Nn3/8/PPP2PPP/RNB1R1K1 b - - 0 11
where it plays 11. ..., f7-f5, it from then on at least starts to direct all its efforts to prevent further advancing of this white e-pawn. E.g., after 21. ..., Be6 we have
[d]4r1k1/4r1pp/ppp1b3/2p1Pp2/5P2/7N/PPPR2PP/4R1K1 w - - 0 22

When white finally involves his Knight, uMax can no longer stop the advance of the passer to 6th rank, (63. e5-e6), despite its best efforts (with a resulting score drop):
[d]2r5/4rbk1/2pR2pp/1pN1Pp2/p1p2P1P/P1P3P1/1P2RK2/8 w - - 0 63

But then the fight focusses on e7, and as uMax really gives utmost priority to prevent this advance, it doesn't undertake anything else, and passively moves its pieces around to get maximum coverage and blocking of e7. This makes it very hard for Telepath to make any progress.

Finally Telepath gets so frustrated that it plays a fatal blunder in the following position (84. b2-b4?):
[d]4b1r1/r3k3/2p1P1p1/1pN2p1p/p1p2P1P/P1P3P1/1P2R3/3R2K1 w - - 0 84

I really suspect an e.p. bug here in Telepath. During the tournament chat it was suggested that uMax doesn't know e.p. capture, but as you can see here this is just false rumour! It immediately takes the b-pawn e.p., the score jumping from about -1 to +1.3. From there, it goes down-hill for Telepath fast, as uMax has no trouble finishing it off.

The 84. b4 by Telepath is inexplicable: there was not even 50-move pressure, as uMax had still moved a Pawn on move 66.

Tony Thomas

Re: WCRCC Finals

Post by Tony Thomas » Mon Jul 23, 2007 8:25 am

swami wrote:
Tony Thomas wrote:Which engines performed better than expected?
Probably Dirty,Rascal and Mediocre and maybe Berta
Link:http://www.taccl.org/2007PresResults.html
I was trying to get the answer from Charles, it is possible that he was talking about any of those engines.

User avatar
Roman Hartmann
Posts: 295
Joined: Wed Mar 08, 2006 7:29 pm
Contact:

Re: WCRCC Finals

Post by Roman Hartmann » Mon Jul 23, 2007 8:29 am

Thanks a lot for the tournament, Charles.
Although I didn't collect too many points (about what I could expect) it was still fun for me.
The only minor dissappointment and surprise to me was that there was not much chat going on in channel 64. There were almost 40 programmers participating and still channel 64 was often deserted. After all I had hoped to gather some information for my already ongoing rewrite of roce :wink:

best regards
Roman

PS: when I saw this position on the board I expected to win at least one full point but then roce didn't see Rxg2+. The older version of roce sees the win and plays that move after 3s but the new 'improved' version never even considered that move ...

[d]r5r1/3b4/p2b1k1p/3p1p2/3P3q/5R2/PP4P1/R1NQ1NK1 b - - 0 36

[Event "ICC tourney 772 (28 1 u)"]
[Site "Internet Chess Club"]
[Date "2007.07.21"]
[Round "5"]
[White "Matilde"]
[Black "roce"]
[Result "1/2-1/2"]
[ICCResult "Game drawn by repetition"]
[WhiteElo "2284"]
[BlackElo "2000"]
[Opening "Durkin's attack"]
[ECO "A00"]
[Time "13:39:19"]
[TimeControl "1680+1"]

1. Na3 e6 2. Nb5 a6 3. Nc3 d5 4. e3 Nc6 5. d4 Nf6 6. Bd3 e5 7. dxe5 Nxe5 8.
Nge2 c5 9. Ng3 Nxd3+ 10. cxd3 Be6 11. d4 cxd4 12. exd4 b5 13. Bg5 h6 14.
Bxf6 Qxf6 15. O-O Bd6 16. Nh5 Qg6 17. f4 f5 18. Qf3 Qf7 19. Rfe1 Kd7 20. Ng3
g5 21. Nce2 g4 22. Qb3 Rhe8 23. Nf1 b4 24. Qa4+ Ke7 25. Ne3 Kf6 26. Qc6 Red8
27. Nc1 Qe8 28. Qb6 Bxf4 29. Qxb4 g3 30. hxg3 Bxg3 31. Re2 Qh5 32. Nf1 Bd6
33. Qe1 Bd7 34. Qd1 Qh4 35. Re3 Rg8 36. Rf3 Qg5 37. Rf2 Rae8 38. Qb3 Bg3 39.
Nxg3 Qxg3 40. Qb6+ Re6 41. Qa5 Bb5 42. a4 Bc4 43. Na2 Bxa2 44. Rxa2 Qd3 45.
Ra1 Qxd4 46. Rf1 Rc8 47. Kh2 Qh4+ 48. Kg1 Qd4 49. Kh2 Qh4+ 50. Kg1 Qd4
{Game drawn by repetition}
1/2-1/2

Peter Fendrich

Re: WCRCC Finals

Post by Peter Fendrich » Mon Jul 23, 2007 10:12 am

I wouldn't expect Alaric to beat Hiarcs. I almost feel ashamed to take the tension out of the tourney :wink:
I think Alarics place as a single-CPU in the middle and ahead of all these multi-CPU engines is really good!
The only loss against single-CPU was Quark who outplayed Alaric with better mobility and that's not a shame. (I think Ktulu was multi-CPU to)

It can not be easy to be the TD for events like this. All these engines that didn't show up at time for each round or where the operator disappeared instead of issuing a match-command, all these questions about how to see the standings, the games etc with the same answer 300 times per day! All these questions about "why not start next round" while games were still going on during current round and so on.... When 40 engines are connected online 15 hours something will certainly happen and it did!

/Peter

Terry Giles

Re: WCRCC Finals

Post by Terry Giles » Mon Jul 23, 2007 11:51 am

Hi Mr Muller,

What version of microMax did you enter in the tournament, was it the current version 4 from your website?
I'm always amazed at how strong a game microMax plays bearing in mind its unbelievably small code.

Kind regards,

Terry :wink:

User avatar
hgm
Posts: 23772
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: WCRCC Finals

Post by hgm » Mon Jul 23, 2007 12:09 pm

It was the Winboard version of uMax 4.8 that is on my website. (Except that I added a few lines of code to make it Kibitz.)

I am indeed very happy with its performance, which also surpassed my expectation. The bonuses for Pawns on 6th and 7th really turned out a crucial improvement. I had intended it as a defensive measure, to put a stop to the many games it lost by naively letting the opponent advance its passers to the 7th before trying to stop them. But to my surprise it worked both ways, and most of the improvement actually comes from the fact that it now wins games through aggressively pushing passers himself, rather than from avoiding the unnecessary losses because it let the opponent do so.

Post Reply