Since the rescheduling of the IPCCC in Paderborn to the week after Christmas, the CCT tournament has become a main tournament for which I really prepare. It is no longer the last rehearsal of the Paderborn version, but an important tournament on its own. This meant that I did not like the bad performance shown by the development versions in the weeks before the tournament. As my testers can confirm I promised them test versions, but did not deliver because I could determine quite easily for myself that they were no good. But the location of the problem remained hidden until a few days before the tournament.
The 9th edition of the Computer Chess Tournament at the ICC server had a novelty. The start times were changed to accomodate the poor people down-under who usually had to stay up all night to participate. Instead, the Americans had to stay up all night, and the European programmers had to start their first round at a time that some of them usually go to bed. In my case that was halfway my normal weekend nights. I put my alarm clock at 5:30 so that I could still doze a little before I really had to get up. Fortunately the roosters next door tried to help my alarm clock a little in getting me awake so I could have a proper cup of coffee, prepare the machine and be ready and waiting for the first games.
There was one little problem though. I received a new book from Arturo Ochoa, and I found it in my mailbox that morning. Importing that book in my own book format would certainly take a while, so I decided to run the task on the machine where Crashtest Dummy would play its game on. That is a dual machine, and CTD would be using only one processor. For the first round I just used an older book. Unfortunately, the import took a bit longer than expected and I had to use the older book also during the second round.
Anyway. I mentioned CTD. Like last year it was entered as an odd/even number or participants even maker. As 53 programs was registered, CTD was entered to get rid of potential byes. CTD is a very simple engine. Evaluation is limited to piece square tables (i.e. move your pieces to the centre of the board, keep your king in a corner until the endgame) and a few basic pawn structure (isolated pawns, doubled pawns, passed pawns) things. The only thing sophisticated in its evaluation is the passer evaluation in pawn endings. Its search is also basic. Plain PVS. No nullmove, no other pruning tricks. It does have pretty decent move ordering though, and a reasonably fast move generator and make move. Finally, its book is a PGN book with very strict parameters, so it will not play many moves from the book. The hardware it used was an AMD MP 2000+, using a single CPU.
The Baron was playing with version 2.11 and used four processors on the Dual Opteron 270 (dual core processors, so 2x2 equals 4 cores) with 4Gb of memory and Windows x64. The most relevant 6-men tablebases were also present, most of them uncompressed to save on decompression table memory. The Baron uses 4 processes and has to allocate the EGTB decompression tables 4 times. One day I am going to fix that by having a fifth process taking care of the EGTB’s, but I have been too lazy to do that so far.
In the first round the Baron was playing Telepath. When Telepath decided to launch a king-side pawn storm while its own king was also at the king-side the Baron made use of the holes that were left in the position and quickly gained a decisive advantage. CTD meanwhile was demonstrating what a lack of king-safety evaluation can do. Symbolic launched a king attack and CTD did not know what hit it until it was too late.
n the second round Symbolic had to face me again, this time with the big brother. I assured Stephen that he was not facing the same engine twice and that in fact the engines had nothing in common. Well, that was not entirely true. They share the book code, but that is it. Unless you count the experience I gained during the 6 years I have been developing the Baron. In the game Symbolic fell for a bookline that must have been prepared for these kind of tournaments and did not have much choice but to let the Baron gain control of the game. Meanwhile CTD was demolished by Rascal. Again in a king attack.
The third round brought the first big opponent for the Baron: Chess Tiger 2007.
Chess Tiger 2007 - The Baron 2.11
1.e4 c6 2.d4 d5 3.e5 c5?!
4.dxc5 e6 5.Be3 Nd7 6.c4
Not so common. More regular is 6.Bb5 after which 6..Qc7 7.Nf3 Bxc5 8.Bxc5 Qxc5 9.Nc3 Ne7 10.O-O completes white’s development
6..Bxc5 7.Bxc5 Nxc5 8.cxd5 exd5 9.Nc3 Ne7 10.Bb5 Bd7 11.Bxd7 Qxd7 12.Nge2 O-O 13.O-O
Baron was quite happy with his passer. Perhaps a little too happy as it is a weak pawn as well and can easily be blocked with a knight.
13..Rad8 14.Nd4 Nc6 15.Re1 Rfe8 16.f4 Ne6
Was this really the best move? At the next move Tiger allows a pawn supporting the passer, but this duo is isolated on open files. Not the best precondition for advancing them. On the other hand a blockading piece is exchanged. Suggestions are welcome.
18.g3 Qb7 19.b3 Qb6 20.Kg2 c5 21.Rc1 Qb7 22.Qf3 Rd7 23.Red1 Red8 24.Na4 Rc7 25.Kg1 g6
26.h3 h5 27.Nc3 Rcd7 28.Rd3 Qc6 29.Kh2 a6 30.Ne2 Ng7 31.Rdc3 Ne6
It is clear that the black centre pawns are more a liability than an asset. All pieces are tied into defending the pawns. And then Tiger starts to launch small attacks to the pieces:
32.f5 Ng5 33.Qf4
Tough position. Both the knight on g5 and the pawn on c5 are hanging. The most logical defensive move fails: 33..Ne4 34.Qh6! with the threat of f6. and mate on g7. The Baron finds a different way out:
Better to attack than to defend!
34.Rxc5 Nf3+ 35.Kh1 Qa8
The idea behind 33..d4 is clear now. The Baron will win the pawn back by a revealed check with the knight.
An alternative is 36.Rc6 with 36..Nxe5! 37.Qxe5 d3 38.fxg6 d2! 39.Kh2 dxc1Q 40.gxf7+ Rxf7 41.Rg6+ Kf8 42.Qh8+ with a perpetual check
36..Nxe5+ 37.Kg1 Nxg6
Baron’s troubles are over. The hanging pieces are resolved and it can think of how to make the best use of the passer on d4
38.Qf2 Qe4 39.Kh2 Re8 40.Ng1 d3
41.Rd1 Re5 42.Rc4 Rf5 43.Qg2 Qe6 44.Rd2 Ne5 45.Re4 Qf6 46.Kh1 Kh8
47.Re1 Rc7 48.Qe4 Kg8 49.Red1 Kg7 50.Kg2 Re7 51.Qd4 Qc6 52.Kh2 Rd7 53.Qa1 Kg8 54.Rf1?
54.. Qh6 55.Rxf5
Moving the rook back to d1 does not work: 55.Rfd1 Ng4+!! 56.hxg4 hxg4+ 57.Kg2 Qc6+ and an inevitable mate
55..Qxd2 56.Kh1 Ng6 57.Rf1 h4 58.Qc1 Qc2
59.g4 Qxc1 60.Rxc1 d2 61.Rd1 Rd3 62.Kg2 Nf4 63.Kf2 a5
64.a3 Rxb3 65.Rxd2 Nxh3 66.Nxh3 Rxh3 67.Rd8 Kg7 68.a4 Ra3 69.Rd4 Ra2 70.Kg1 Kg6 71.g5 h3 72.Rf4 Kxg5 73.Rxf7 Rxa4 74.Rc7 Rf4 75.Rc5 Rf5 76.Rc4 Kf6 77.Re4 Rh5 78.Kh2 Kf5 79.Rc4 Ke6 80.Rd4 Ke5 81.Rd3 Ke4 82.Rd7 Ke3 83.Rc7 a4 84.Ra7 Rh4 85.Ra6 Kd3 86.Ra5 Kc3 87.Ra7 Kb3 88.Re7 a3 89.Re3 Kb4 90.Re1 a2 91.Kg3 Rh5 92.Kh2 Ra5 93.Ra1 Ra3 94.Kh1 Kc3 95.Kh2 Kb2 96.Rg1 a1=Q 97.Re1 Qc1 98.Re2 Kc3 99.Rd2 Qxd2 100.Kxh3 Qf4 101.Kg2 Ra2 102.Kg1 Qc1# 0-1
CTD got its first score with a GM like draw against Noonianchess, the little brother of Telepath. Noonian’s queen was trapped and chased hin and forth between two squares, a plan with which CTD bigger brother agrees.
The last round of day one the Baron had to play another commercial engine: Hiarcs. The Baron chose a somewhat passive position and got outplayed by Hiarcs. Congrats to Mark Uniacke for that.
LearningLemming tried to follow the example of Symbolic and Rascal by attacking at the kingside. This time CTD managed to defend its position and created a passer in the centre with which LearningLemming had some problems. When CTD managed to trade the good bishop of LearningLemming for a knight it went downhill for LearningLemming. CTD scored its first win in this tournament!
That was the end of day one. I felt quite tired which was of course partly due to the unusual time for getting out of bed. So besides chatting a bit, doing weekend shopping, getting dinner from the local ‘Frituur’, watching some M.A.S.H. reruns I did not do much else and got to bed.
The next morning the rooster saved me again as I seemed to have turned off the alarmclock more or less in an unconscious mode. But I was awake enough to realize that when roosters start to sing it was time to get out of bed for the next round.
The opponent for the Baron was npmess, the latest incarnation of the mess-series of Mridul Muralidharan. The Baron started slowly with a Caro-Kann, and slowly gained space, occupied a nice d5 square with the knight while npmess’ bishop was degraded to a pawn. Baron got the initiative in the endgame and won it.
Meanwhile, CTD survived the first moves against Neurosis and managed to get a space advantage with a firm pawn centre and the pair of bishops. Neurosis defended for quite a few moves but could not prevent CTD from winning its second game.
Against Naum the game was going for a draw as soon as the opening phase was over. Any human would have seen that. Unfortunately the computers did not and the game went on for over a 100 moves. CTD also played a long game against PostModernist. PM was very optimistic during the game, and probably had all rights to be so. However, CTD is a very secure defender, as long as king safety has no role in the game. Nevertheless, PM won in the endgame.
With the 5th place after 6 rounds I could be happy. I already calculated that the top 4 would be playing amongst eachother, so the Baron would be playing a program that was ranked below it. If someone would have asked me which program I rather not played the last round, it would be exactly the one that I got: Zap!Chess. Erdo, the bookauthor of Zap had been watching the Baron’s games, and saw that it would play the Caro-Kann, making it hard for Zap to win. So it opened with the d-pawn, getting into a sort of London system. By doing that he managed to get the Baron out of book quickly. In the closed position that followed, the Baron did not choose the correct continuation, and lost.
CTD meanwhile got another clear signal that king-safety evaluation is not luxury. Now sacrificed a piece, put up a cannon in the centre of the board and CTD got shot.
All considered, the Baron put up a nice show, and with the shared 9/15 th place (11th after tie breakers) it did end above its seeding.
Both the Baron and CTD also participated in the blitz tournament and were paired against eachother in the first round. No sweat for the Baron and it won its game easily.
In the second round the Baron played Berta. A neo-Grünfeld and a draw. The Baron tried, got a nice position, but could not bring in the win. CTD had a strong opponent in Petir and lost without much of a chance.
The Baron sacrificed a piece in the openingbook in the third round and managed to keep it a draw against Naum. CTD was lost quite soon after the opening and Buzz took the point.
Danasah did underestimate the Baron’s chances in a king attack and had to accept that the Baron’s piece ruled the board. CTD managed to beat the point donor Tohno.
In the fifth round the Baron drew Spike. Both programs tried for more, but did not succeed. CTD made a mess out of Noonianchess’ pawn structure and won.
Natwarlal was the surprise of the blitz tournament. It managed to stay with the leaders for quite a while. But the Baron had no problems with it and won. It was time for CTD to lose again and it did it against Prophet.
Also Hiarcs could not penetrate the Sämish-Nimzo that the Baron was using, following the game with Spike for quite a few moves. With the same result: Draw. CTD played Neurosis and won again.
While I was thinking of a top result, Loop was the opponent in the 8th round. Loop did penetrate Baron’s Nimzo and won. CTD did not lost against npmess and added half a point to its total.
The last round for CTD was another demonstration on the importance of king safety. The two ladies Eve and Ann beat up the dummy until it surrendered. Meanwhile the Baron had the not so easy task to play Scorpio, an engine with a sting:
Scorpio - The Baron
CCT-9 Blitz round 9
1.e4 e6 2.d4 d5 3.Nc3 Nf6 4.Bg5 dxe4 5.Nxe4 Nbd7 6.Nf3 Be7 7.Nxf6 Bxf6 8.h4 h6 9.Bxf6 Qxf6 10.Bc4 Nb6 11.Bd3 Bd7 12.Ne5 O-O-O 13.Qg4 Ba4 14.Rh3 h5 15.Qe4 Be8 16.Rg3 Kb8 17.c3 Nd5 18.O-O-O Rd6 19.Qf3 Qh6 20.Kb1 f6 21.Nc4 Rd8 22.Bc2 g5 23.Qe4
At this important position the Baron decides to sacrifice a pawn. The weak pawn on e6 is sacrificed for rapid development, and immobilizing the white rook on g3.
23...g4 24.Qxe6 Bg6 25.Qe2 Rhe8 26.Qd2 Qxd2
The Baron trades down to enter a favorable endgame. It is in fact almost a rook up as the one on g3 will not join in the conflict for a long while.
27.Nxd2 Bxc2 28.Kxc2 Re2 29.Rf1 Nf4 30.Kc1 Rde8
31.Kd1 b5 32.a3 c5 33.Kc1 c4 34.Kd1 Nd5 35.b3 cxb3 36.Nxb3 Ra2 37.Re1 Rxe1 38.Kxe1 Rxa3
The Baron has its pawn back, and gets another one soon. This releases the rook on g3, but the ending is now rather easily won.
39.Nc1 Rxc3 40.Rxc3 Nxc3 41.Kd2 Nd5 42.Ne2 a5 43.Ng3 Nf4 44.Kc3 Nxg2 45.Nxh5 f5 46.d5 Nxh4 47.d6 Kb7 48.Kd4 a4 49.Ng3 f4 50.d7 Kc7 51.Ne4 Nf5 52.Kd3 g3 53.fxg3 fxg3 54.Kc3 g2 55.d8=N Kxd8 56.Nd2 g1=Q 57.Nb3 axb3 58.Kxb3 Qc5 59.Ka2 Nd4 60.Ka1 Qc3 61.Ka2 Qb3 62.Ka1 Nc2# 0-1
3rd in the Blitz tournament is much more than I would have expected beforehand. And CTD? Well, its crappy hardware and lack of chess knowledge did rule it out for a top position to begin with. 29th out of 34 (when not counting the withdrawals) is probably about where it belongs.