The Baron at the WCRCC

[Richard Pijl]

In the weekend the Baron was participating in the World Computer Rapid Chess Championship (WCRCC) organised by Charles Roberson. This year only 12 participants were prepared to spend their weekend on this tournament and as a consequence many of the stronger programs were missing. With a maximum of 15 scheduled rounds it was quickly decided to play a 12-player round robin.

Top favourite was Deep Sjeng on a 200+ core cluster where a couple of other programs, including Crafty, Tornado, Nightmare and the Baron were supposed to be competing for the remaining 'medals'. These others all ran on more modest hardware, ranging from 4 to 8 cores. The Baron used (again) an 8-core machine that was made available by the Hiarcs team and that has served well in previous tournaments.

The Baron was a slightly modified version 3.30a that is performing in the 'basement' tournaments recently, like in the Premier Division of the last WBEC tournament. The main remaining preparation task was in deciding on what search parameters to use. Many tests were performed by Rodolfo Leoni and Silvian Rucsandescu to find the optimal pruning/razoring/reduction/extension settings that resulted in the much improved Baron 3.30a version. They did not stop at that and continued, but in the end I chose the relative safe search parameters that were the default in 3.30a as the newer (more aggressive) settings seemed to be suffering from tunnel-vision due to the larger reductions used. Completely subjective and based on only a few games, but that is how decisions like this are made in the end. Also Sebastian Böhme delivered a new and up-to-date opening book so that we would not be surprised by home-cooked lines against the lines the Baron played in previous tournaments.

The first round the Baron was paired with white against Goldbar. By mistake I loaded the wrong opening book (last year's instead of the updated one) and by surprise the Baron opened with 1.d4 and a typical King's Indian appeared on the board. Goldbar was steered by the book to attack on the kingside, leading to a funny looking g-file where each side had a king, bishop, knight and a pawn. After deflecting the kingside attack of Goldbar, the Baron slowly managed into gaining space on the queenside. Goldbar attempted guerilla tactics by letting a lone knight penetrate the white position, but in the end the poor beast could
not do much harm.

The next round the proper book was loaded and the Baron played white against one of the main competitors: Tornado. After a long Sicilian Najdorf bookline a volatile but equal position was reached. The Baron appeared to have a nice supported knight on c6, but Tornado had a similar beastie on f4. Tornado obtained a strong pair of knights where the pair of bishops of the Baron were not finding the right squares. One of the bishops was tied to defending the worse pawn structure of the Baron. By that time I expected to get at most a draw from this game. Until Tornado started to complicate things:

The Baron - Tornado
[fen]r5k1/5r1p/p1Np1np1/B2Pp3/4Bn2/4bP2/PP5P/1K2R2R w - - 32 0[/fen]

32..N4xd5!?
Unexpected as it solves part of the Baron's problems. But it is hard to see in what other way Tornado can make progress.
33.Bxd5!?
Also a surprise. An alternative would have been 33.Nxe5 dxe5 34.Rxe3 Nxe4 (34..Nxe3 35.Bxa8 gives the advantage to the Baron) 35.Rxe4 and the position seems to be very equal.
33..Nxd5
34.Rd1 Rc8?
Sacrifices the exchange to keep the pawn, but now the Baron will get the advantage. 34..Nb6 or 34..Ne7 would have been better
35.Nd8! Rxd8
36.Bxd8 Bd4
37.Rd3 Rf5
38.h4 Kf7
39.Bg5 Ke6
40.Re1 Bc5
41.a3 Nf4
42.Bxf4 Rxf4
43.Re4 Bf2?!
Tornado should probably not allow further trade-downs as the queen-side pawn majority of the Baron will be decisive.
44.Rxf4 exf4
45.Rd1 Bxh4
Winning the pawn on h4 locks out the bishop from action.
46.Rh1 g5
Giving up the pawn on h7 is not an option as the rook will be snacking more pawns
47.Kc2 Kd5
48.Kd3 a5
49.Rc1
and the Baron won after 27 more moves.

In the Queen's Indian with black against Arasan the Baron got a nice position out of book. Arasan tried to gain space with a king-side pawn storm, but opening the position was not really an option as that would activate the pair of bishops of the Baron. And when the Baron managed to open the position on the queensside there were too many weaknesses in Arasan's position

Crafty-the Baron in the fourth round was a somewhat irregular Sicilian setup in which Crafty seemed to have most of the space. The Baron exchanged his good bishop for a somewhat crippled knight to ruin Crafty's pawn structure. When the Baron refused to play a4 to close the a3-f8 diagonal (and crafty played a4 instead) it became hard to materialize the advantage. At least the Baron needed many moves to find the correct way forward, but in the end it found the way to create a strong passer that won the game.

Parrot got a decent position out of book with black, but became greedy and missed a tactic that made it lose the exchange.

Jabba had to play against Baron's Gruenfeld, castled into trouble on the queenside, the Baron gradually activated his pieces. With Jabba's king and queen on the open c-file the Baron had sufficient targets for the attack with rook and pair of bishops.

Nebiyu, playing without an opening book, did not handle the Petrov as it should and ran quickly into trouble.

In the final game of the first day, Tinker had a very long Nimzo line in which it won a pawn. But the Baron had sufficient compensation with very active piece play and Tinker had to solve the problem of a bishop with pawn-aspirations. Tinker held on to that extra pawn for a long time, but when Baron started to collect it back it did so with interest.

So the Baron won all of his games on the first day. With only three more games to play it was clear that the game against Deep Sjeng would probably be decisive, and with Nightmare it still had another strong program to play.

The first round on the second day was a game with white against Telepath who managed to force the game into the unknown after just 4 moves. A quick look in the database told me that only 2 games were known with that line, both won by black. But Baron played a novelty on move 5 and managed to convert the initial Sicialian into something that resembled a King's Indian. Without a book guiding the program through that type of position it is hard to follow the correct plan. The Baron blew up the central pawn structure, created a passer in the centre and after the queen exchange this passer became a winning advantage.

Next was the game that was to be the decisive game for the tournament. By looking at the games Deep Sjeng played in previous rounds some adaptations to the book were made to avoid lines that proved to be successful for Deep Sjeng. But as we did not know what colour the Baron would have, both white and black lines needed to be examined. Initially in the game this seemed to have been successful. The Baron was happy with the position it got and Deep Sjeng apparently not so. Then the Baron tried to force things:

Deep Sjeng - The Baron
[fen]r1r3k1/pp1bppbp/n2p1np1/q2P4/3Q4/P1N1PN2/1P1BBPPP/2R1K2R w K - 1 1[/fen]
12.. Nc5!?
In the exchange that follows an unbalanced material position arises where The Baron has a queen and three pawns against three pieces.
13.b4 Qa3
14.bxc5 Rxc5
15.O-O
or 15.Qh4!? h6 16.O-O g5 17.Nxg5 hxg5 18.Qxg5 Rac8 19.Ra1 Qb2 20.Ra2 Qxa2 21.Nxa2 Rxd5 22.Qf4 Rxd2 23.Nc1 Rxc1 24.Rxc1 Rxe2 to mention another wild line
15..Nxd5
16.Qxd5 Rxd5
17.Nxd5 e6
18.Nc7 Rc8
Perhaps 18..Rb8 would have been better to prevent the rook exchange.
19.Nb5 Qa6
20.Nbd4
The rook exchange looks good as well
20..Qb6
21.Ra1 a6
22.Rfb1 Qa7
23.Ra2 h6
24.Ne1
Moving the knight to the black pawns on the queen side with the goal to block/eat them.
24..Kh7
25.Rab2!
Luring the pawns forward so that they will be easier to attack
25.. b5
Forced. Rb8 would be admitting that black is basically lost.
26.Ra2
Now the a-pawn is terribly weak. It becomes clear that having a queen for three pieces is really a disadvantage when your extra pawns are basically sitting ducks. No matter what the computer evaluation says here ...
26..Qb7
27.Bf3 d5
28.Bd1 Qb6
29.Rba1 Ra8
30.Nec2 Bc8
31.Ba5 Qb8
32.Rc1 Bb7
33.Nb3 Bf8
33..Be5 is not essentially different
34.Bd2 Bd6
35.g3 Qd8
36.Na5 Bc8
37.Bf3 Qf6
38.Ne1 Bd7
39.Nb7 Bf8
40.Nc5 Bxc5
41.Rxc5 Rc8
42.Bc3?
Exchanging rooks first on c8 is probably better
42..Qd8?
Missing the chance to get out of trouble with 42..e5!, e.g. 43.Rxc8 Bxc8 44.Bd2 (44.Bxd5 Qd6 45.Rd2 b4) d4 45.exd4 exd4 and the active queen may start to make the difference. 43.Rxd5 is another option, followed by 43..Rxc3 44.Rxd7 e4! 45.Bd1 Kg7 and the activity of the white pieces is reduced while the black queen is hyperactive
43.Nd3 Rxc5
what else?
44.Nxc5 Bc8
With the black bishop hemmed in by its own pawns, it is very clear that white is better now. 4 attackers against 2 defenders, although it took 50 more moves until the Baron resigned.

In the final round Baron played Nightmare with white, had to win and hope for Deep Sjeng to drop half a point or more against Crafty to get a tiebreak match. But Sjeng won rather quickly, and the Baron was 2nd, no matter what its result against Nightmare would be. With another long Najdorf bookline the Baron did get a small advantage which it slowly built out, entered a rook ending with an extra pawn and converted it to a win.

With 10 out of 11 a nice second place, 1.5 points ahead of number 3 was the Baron's reward and this has been the best result so far in any tournament it participated in.[/b]

[Bryan Whitby]

Hi Richard
Many thanks for the report.
Bryan