Hossa - The CCT2 tournament |
When I saw the pairing of the first round I thought "Ok, so I have some good chances to get a weaker opponent in the 2nd round." Although Hossa already managed in the past not to lose from Crafty in very few games I didn't expect anything else than a loss.
Bob started the game and it started with 1.d4 d5 2.c4 e6 ... Then there was some confusion because TD SJLIM hasn't already started the games. So we aborted the game and started a new one, when SJLIM gave his sign. This time Crafty played 1.e4 and Hossa played the Najdorf Sicilian. A very deep book line followed. Hossa's first own move was move #26. The resulting position was rather unbalanced.
In such unbalanced positions it happens that both programs think that they have a disadvantage. This often leads to a quick draw. And this was what happened in this game, too.
I didn't know much about Shrike before this game. Therefore I didn't know what I had to expect from this game. Again Hossa played a long book line. The first computed move was move #21.
Black castled short, white castled long. Some weeks ago I improved my pawn storm code. Hossa knows a bit about blocked storm pawns. Therefore he was happy that black's pawn a5 was blocked. On the other hand he managed to play 23.Nf6. After the exchange of the knight Hossa had very good attacking chances.
Now my son Lukas woke up and I had to leave my computer. When I returned the game was already over. :-)
After the good start from Hossa it was clear that I had to expect to face a very good opponent in the 3rd round. ZChess indeed was a tough opponent. About an hour before CCT2 started Franck Zibi and I let our engines play some warm up games. ZChess was clearly better in those games. In round 3 Hossa was out of book after move 13. It was an accepted queen gambit. Actually it should be very unlikely that Hossa plays this opening. It seems that there are some problems with the random move generator. Very soon ZChess had a good attacking position:
The pressure became stronger and stronger and I thought that Hossa would lose this game very soon. In move #34 Hossa already had a score of -3.59. But suddenly after 35.f4 Hossa was happy again because he saw some counter play with 35... Nf3+!
After a while there was an endgame on the board where Hossa down one pawn and white had a passed pawn on a7. Hossa knows that this is a disadvantage for white in such positions because the white rook is blocked if blacks king can stay at g7 or h7. This is a pattern I added about one week before. This position is a draw. If the white king walks to b6 black can give checks from behind. It also isn't possible for white to catch the g5 pawn because of side checks.
I was happy that my opponent in round 4 was Diep. I like this program because Vincent's philosophy is similar to mine. Hossa was out of book after move 5 in this game. In this game there were positions on the board where Hossa didn't have the right knowledge to find good plans. E.g. he never started a minority attack with b4 which was possible more than once. After some dubious moves the e-pawn became weak and in move #28 Diep was a pawn up. Nevertheless the position wasn't clearly won for black. Maybe Hossa even had some good draw chances in the resulting queen endgame. But then came a big blunder in the following position:
I have some special evaluation code so that Hossa would almost never exchange queens in such positions because the pawn endgame is very likely to be lost. Somehow I had a bad feeling and of course Hossa played 41.Qxf6. When I examined what happened some days later I found a black/white-bug (this term was invented by Chrilly Donninger ;-). The evaluation is done correctly for black. But because of stupid cut&paste I gave white a big bonus instead of malus for exchanging in the lost pawn endgame. After having fixed that bug (I only had to replace a b with a w Hossa never considers Qxf6??
Insomniac is a rather new program written by James Robertson. It has already shown in tournaments before that it is very strong. Also from the games at ICC between Hossa and JZInsomniac it was clear that Hossa is weaker than Insomniac. Therefore I was a bit surprised that Hossa never was in danger in round 5 and got an easy draw. Move #7 was the first move which Hossa had to compute. Both programs played weird bishop moves in the opening: 7.Bc4 Nc6 8.Bb5 Bd7 9.Be2 Bc8. After this brilliant maneuvers of both black was back in book! Unfortunately Hossa doesn't detect book transpositions.
Now Hossa played the nicely with 24.Be4 Rd6 25.f4 and was soon a pawn up. Unfortunately this pawn wasn't enough for a win.
In this game Sjeng's author was very unhappy because his program played without an opening book. Sjeng wasn't operated by himself. Sjeng managed to play very passively so that Hossa was out of book very soon, too (already in move #3). White wasted too much time with early queen moves (5.Qa4+, 6.Qb3) and Hossa developed very well. Already in move #12 Hossa showed an advantage of 1 pawn.
In move #15 Hossa missed a quick win with 15... d4! He rather took the exchange and lost some pressure. So it was a bit harder to win this game but in the end Hossa was successful.
So far I was very happy with Hossa's performance. He got 3.5/6 and I hoped to get at least 1/2 point from the last two games for a 50% result. In the game with Yace it was the only time when Hossa had a hardware advantage. Unfortunately so far Hossa played very bad vs. the Yace clone counter play at ICC.
It was played a long book line. Hossa was out of book after move #19. It was a gambit which Hossa didn't understand and which was dubious anyway. Therefore the pressure was gone only few moves later. After this the game was already lost although Hossa fought until the late endgame.
I am not happy at all with this game. Hossa played very bad. He was already out of book after move #4. After having left the book Hossa continued to play very strange. At least 5... Nh5, 9... Qd6, 14... Qg7, 15... g5 are moves which I must work on.
© Steffen A. Jakob