Computer Chess Club Archives


Search

Terms

Messages

Subject: Hossa@CCT5, my notes

Author: Steffen Jakob

Date: 01:43:00 01/21/03


Here are my notes about Hossa playing the CCT5. I have not yet
analysed Hossa's game deeply but I want to give you a fast overview of
my impressions. Please give me some feedback if I made mistakes in my
quick analysis.

I haven't worked a lot on the "old" Hossa since CCT4 because again I
started to rewrite my engine (It sometimes runs as "Jockel" at
ICC). Unfortunately this is a very time consuming and also sometimes
boring job. This new version isn't by far as strong as the stable
Hossa. Therefore I first didn't want to join CCT5. Also hardware was
an aspect against joining because it seemed as if I had to play with
my Athlon 900. Having a look at the participants list I saw that most
engines will run on much faster machines. Hossa already is a slow
searcher and gets outsearched most of the time. I feared that Hossa
would not be competitive with a big hardware disadvantage. On the
other hand I played all CCTs so far and didn't want to miss the fun I
always had in those tournament. Then in the last month I began to
enjoy to modify the old Hossa again. I think that the changes were
significant compared to the CCT4 version so I decided to join CCT5 -
encouraged by people like Alex Kure and Bob Hyatt - and not to miss
the fun even if I had to play on a slow hardware. I was very happy
that some days before CCT5 the hardware problem was solved. Alex Kure
organized a P4 2.8 GHz with 2 GB RAM for me which I could use for
CCT5! This made me much more optimistic. On thursday we installed
RedHat 8.0 without problems on the P4. Hossa was ready to go.

Then finally the tournament started.

----------------------------------------
Round 1: Monsoon - Hossa (0-1)
----------------------------------------

After the tournament I read that people think that rank 6 for Monsoon
was a surprise. I wasn't suprised about that at all because I had
played many games with Monsoon before and knew how strong it
was. Hossa had a very bad score against Scott's creature. So I wasn't
too optimistic for round #1. Well, if I had lost I could declare it as
having played a swiss gambit. ;-) One of Hossa's biggest problem is
his opening book. I tried to compile a very small opening book and
hoped that Hossa would survive this phase. It wasn't successful all
the time. Right in the first round Hossa allowed his opponent to
destroy blacks king pawn shield:

[D]r1bq1rk1/pppp1p2/1bn2p1p/1B2P3/3p4/2P2N2/PP3PPP/RN1Q1RK1 w - - 0 10
after 9... gxf6

Later Hossa gave away a pawn (c6) and got a rather strong passer on
the d file:

[D]r4r2/p1p2pk1/1bN5/3pP1p1/6b1/2N5/PP3PPP/2R2RK1 b - - 0 18
after 18... d4

I was more confident now. Monsoon was very happy with its position though.

In move 20 Monsoon offered the exchange:

[D]r4r2/p1p2pk1/1bN5/4P1N1/3p4/8/PP2bPPP/2R2RK1 b - - 0 20
after 20... d3

I am a bit proud that Hossa refused to take the rook but instead
played 20... d3 and kept the pressure. Later Hossa opened the f file
and all his pieces very involved in active play. Monsoon even gave his
e pawn so that black got a pawn on the e file which could support the
strong pawn d3.

This is the position one move before Monsoon resigned:

[D]8/N1p4k/8/P3Rr2/1P2pr2/3p3P/3Rb1PK/8 w - - 1 33
after 32... R8f5

Here Monsoon played 33.Kg3 and resigned after 33... Rxe5 because of
34.Kxf4 e3 35.Ra2 d2. But what if Monsoon had played 33.Re6 ? This was
the move which Hossa expected during the game with a draw score. Can
anybody find an improvement for black in after this move??

----------------------------------------
Round 2: Hossa - Crafty (0-1)
----------------------------------------

This game was a disaster. I could not watch most parts of the game
because some minutes after its beginning my wife had a circulatory
collapse. Fortunately she was ok very soon. Hossa showed solidarity
with Regina and collapsed, too.

I saw Crafty computing more than 2600000 nps in that game! Crafty was
able to build a king attack and with that computing power it was over
very soon.

[D]r4r1k/1p4p1/2pp2np/2b1p2q/p3Q1bP/P1PP1NP1/BP1N1PK1/R5R1 b - - 0 20
before 20... Rxf3

I resigned in move 26.

----------------------------------------
Round 3: Amateur - Hossa (1/2-1/2)
----------------------------------------

Games between Amateur and Hossa are most of the time interesting. IMHO
they are about equally strong. Amateur ran on the same CPU as Hossa
did, so I expected an open fight. Unfortunately the opening was bad
again. Hossa was down a pawn at move 15:

[D]b2q1rk1/2p1bppp/3p1n2/1N6/4P3/1PN5/1P3PPP/2BQR1K1 b - - 0 15
after 15.Ndxb5

Kingway said: "few humans would want that pawn"

Soon lots of pieces were exchanged. Both sides had a rook, a queen,
and a bishop of opposite color. My hope was that Amateur would allow to
trade even more pieces. Kingway comment was: " trading rooks and
queens is trival draw, trading queens is easy draw, trading rooks is
still work...". I was right: in move 31 Hossa could exchange the
rooks. The position was already very drawish. Amateur was still very
happy with its position though. In move 33. Amateur took Hossas queen
and the position was a dead draw:

[D]6k1/3q1ppp/2p5/2B2b2/1P6/5P2/1P4PP/3Q2K1 w - - 1 33
before 33.Qxd7

Although it was a easy draw the game continued to move 144. Hossa even
almost blundered and I started to get nervous but in the end the game
was drawn. It also was interesting that Amateur showed a score of >+6
in the last moves. Will could not explain this. I hope he can find and
fix this bug.

----------------------------------------
Round 4: Hossa - Averno (1-0)
----------------------------------------

At the beginning of this game I had some massive lag. Hossa got
disconnected but luckily I could establish the internet connection
soon again. This game was the cleanest victory from Hossa. Averno
didn't have much chance. Already in move 17 white had an impressive center:

[D]3rnrk1/1pqnppbp/p5p1/P1pPP3/5Pb1/2NBBN2/1P2Q1PP/R4RK1 b - - 1 17

It was only a question of time when something tactically was possible:

[D]3rnr1k/1pqnp2p/p4p1b/P1pPP1p1/2B2PbN/2NQB3/1P4PP/R4RK1 w - - 0 21

In this position Hossa found the nice 21.Ng6+! Of course other moves
were winning, too (e.g. 21.d6). I am interested if your favourite
engine would play Ng6+ here, too. Black couldn't stop whites king
attack and resigned at move 28.

----------------------------------------
Round 5: Chezzz - Hossa (0-1)
----------------------------------------

I was already very exhausted - this is something btw. which my wife
could not understand. She asked me why I was exhausted because I had
nothing else to do than to start the games. Everything else was done
by the engine. Technically speaking she is right, of course ;-). I
could hardly follow the game and expected to fall asleep during the
game.

[D]r2qk2r/pp2bppp/2npbn2/2p1p3/2B1P3/2NPNP2/PPP3PP/R1BQK2R b KQkq - 2 9
before 9... Bxc4

After 9... Bxc4 black had already a suspicious position because his d
pawn became weak. Usually Hossa took about 1-2 minutes per move in the
tournament. Now something strange happened in this position:

[D]3b4/p2k1qp1/3Pn3/p2Np2p/2P1P3/2P1B1Q1/6PP/6K1 b - - 0 32

White had just taken d6 and Hossa started to think and just didn't
want to move. Suddenly I was awake. Already while Hossa was still
thinking I could find with the help of the log file and having a look
at the source what was going wrong. In certain circumstances Hossa
extends the search time by small pieces. The pieces should get smaller
each time the time gets extended. Here it was different! The pieces
got larger, much larger. As Hossa usually plays blitz most of the time
this bug was not visible because with shorter time control the time
doesnt be extended as much there. Luckily Hossa wasn't in danger to
extend its search time over the remaining time. It was also a lot of
luck that this bug occured in the last round of the day so I had time
to implement a workaround until day 2.

When Hossa moved 32... Kxd6 he had spent more than 17 minutes for this
move!! He had about 5 minutes left on the clock. Chezzz had more than
20 minutes IIRC. And the position wasn't very well, neither. Then
Hossa had a lot of ponder hits which helped him to recover a bit on
the clock. But also the position on the board changed. Chezzz traded
the queens and suddenly the two a-file passers became dangerous. This
game needs some deeper analysis which I haven't done so far. In the
following position I think white is already lost:

[D]8/2k5/8/p2P4/2P1p1p1/pKP3P1/Nn6/8 w - - 0 59

White played until he got checkmated in move 93. David was very
unhappy with that game which I can understand very well. It's the
opposite here of course. I already thought that Hossa would lose this
one for sure. It was a nice turnaround for me.

----------------------------------------
Round 6: Hossa - Bringer (1/2-1/2)
----------------------------------------

This was the first round of the second day. I hoped that the
workaround for the bug from game 5 was working (spoiler: it did!). So
far Hossa had performed much better than I had expected it. But I knew
that I would get only very strong opponents on the 2nd day (well, the
opponents from day 1 were strong, too).

Hossa played a strange opening again in this game with 1.f4 d5 2.d4 e6
3.Bd2 and tried to play actively with 11. f5 and later with 25.g4.

[D]2r2rk1/pp1b2pp/1q2p1n1/3p4/3P2P1/P1N1Q1BP/1PP5/1R2R2K w - - 3 30

Here Hossa played 30. b4 which I could not understand. Bringer
improved the pressure on the c file:

[D]5rk1/pp1b2pp/2q3n1/3pp3/1PrP2P1/P1N1Q1BP/2P5/1R1R2K1 w - - 2 33

White already loses material. Hossa played 33.b5 which gives the best
counterplay IMHO. Hossa won back the pawn later and it even looked as
he had good chances to win the game.

[D]6k1/3K2p1/3P3p/8/8/5R2/8/6r1 b - - 0 74

As Peter already stated tablebases helped to fix the draw.

----------------------------------------
Round 7: Terra - Hossa (0-1)
----------------------------------------

In the opening white moved his queen in an inactive position:

[D]1r1qk2r/pp3pbp/2ppbnp1/n3p3/Q1PP4/1PN1P1P1/P3NPBP/R1B1K2R b KQk - 0 11

Hossa has some expensive knowledge about queen mobility which should
pay in this game. Nevertheless white had the better position because
he managed to build some pressure on Hossa's center pawns:

[D]2r1nrk1/pp4qp/2nppbp1/2p5/Q1P5/1PN1P1PB/PB1R1P1P/2R3K1 b - - 13 21

Here a move like 21... Qf7 seems obvious to protect e6. Hossa instead
played for positional reasons 21... Nc7 which I like very much. This
gives away the pawn d6 but removes a lot of pressure from his
position. Look at the position after move 27:

[D]3q2k1/pp5p/n1n1pbp1/2p5/Q1P5/PPN1P1P1/1B3P1P/5BK1 w - - 0 28

Whites queen is still under control which Hossa knew. White has the d
file. Later Hossa forced Terra to give a knight in order to save his
queen:

[D]1n1b1k2/p7/1pn1B1p1/1Qp4p/N1P5/PP2P1P1/1B1q1P1P/6K1 w - - 3 36

The threat was a6 therefore Terra played 36. Nxb6. But even after that
white was not able to free his queen for some time. This time Hossa
used to grab some pawns. Finally when white's queen escaped from her
prison it was already too late.

----------------------------------------
Round 8: Hossa - Ferret (0-1)
----------------------------------------

This is the 2nd loss from Hossa in CCT5. Hossa was as chanceless as in
the game with Crafty although Bruce meant "Hossa is still kicking"
when I thought it was already over. Hossa lost time for his
development by playing Ne5 and Nb5 too early. After move 17 there was
a triple pawn on the board:

[D]3rk2r/p3ppbp/q3b1p1/1pp1P3/8/1P2P1P1/PBQ1P1BP/R4RK1 b k - 0 17

Ferret soon got two connected advanced passed pawns on the b and c
file and had no problems to get the full point and become leader of
the table.

----------------------------------------
Round 9: Diep - Hossa (1/2-1/2)
----------------------------------------

The first move after having left the opening book was 8... g6 and
allowed Diep a nice tactical manoever:

[D]rnbqkb1r/p3pp1p/2p3p1/1p2N3/P1pPP3/2P5/5PPP/R1BQKB1R w KQkq - 0 9

After 9. axb5 cxb5 Diep played 10. Nxf7! Kxf7 11. Qf3+ Kg7 12. e5 Nc6
13. Qxc6. Later Diep was a pawn up by winning c4.

[D]r6r/4pkbp/6p1/p7/2RP1B2/2P5/5PPP/4K2R b K - 0 21

Hossa managed to push the a-file passed pawn to a2. After a while the
following position was on the board:

[D]4k3/6b1/4p3/2B5/2PP1R2/2K5/8/7r b - - 2 50

I am not qualified to comment this endgame. After the game Vincent
said that he thinks that white should win. In the game Diep could not
find a winning plan and it was a draw after move 74:

[D]8/R7/4pbk1/2K5/2PP4/8/3r1B2/8 w - - 51 75


----------------------------------------
Conclusions
----------------------------------------

It was a lot of fun to play this very well organized tournament. The
time control is too fast for my taste but on the other hand it's more
family-friendly to play on only one weekend. For the next CCT5 I would
vote for not having a winner decision by blitz games. Crafty, YACE and
Ruffian all deserve to be named as winners of this tournament.

Some conclusions:

- Hossa had some comebacks in bad positions. I think my latest changes
  were rather good.
- Hossa overlooked some tactical shots, but not as many as it usually
  does. The 2.8 GHz machine helped a lot here.
- I badly need a better opening book!
- Now I see lots of ways how to improve the old Hossa. Motivation came
  back. I think I will freeze the reimplementation and continue to work
  on Hossa-1.151.
- My time management needs a redesign.
- TD is the abbrevation for "touch down", BOOC means "bishops of
  opposite colors" ;)

I am looking forward for CCT6.

Best wishes,
Steffen.



This page took 0.05 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.