TCEC Season 1 - Stage 2a lineup and start date!

[Martin Thoresen]

.
TCEC Season 1 – Stage 2a starts Friday, February 22nd at 18:00 CET.

Stage 2a is a double Round Robin consisting of 8 chess engines, with a total of 56 games.
Each pair will play both sides of the openings.

The top 4 will move on to Stage 3, while those finishing 5-8 are out of TCEC for the current season.
Stage 2b will start 2-3 days after Stage 2a is finished.

The list of participating chess engines for Stage 2a is as follows:
Houdini, Stockfish, Hiarcs, Rybka, Critter, Junior, Gaviota, Spike

The list of participants for Stage 2b:
Vitruvius, Gull, Chiron, Komodo, Shredder, Scorpio, Quazar, Hannibal

New on the website since Stage 1 are the double PV boards that have been
added to the Live page. So now you can see what kind of position the engines are expecting.

In other news, the new TCEC computer has been built, a dual CPU Intel Xeon with a total of 16 cores.
It will not be used yet since there are still single threaded engines left in the current Season.
If there are no single threaded engines left for Stage 3 this new hardware will be used from that point on.

TCEC is using 150m + 60s per move for the whole game, running on an Intel Core i7 3770k @ 4500 MHz.
Up to 3 cores can be used for engines capable of SMP.

Read the full TCEC rules here:
http://www.tcec-chess.net/info.php

The broadcast will be running 24/7 with 1 game at a time, and can be watched here, also with chat:
http://www.tcec-chess.net/live.php

Finished games from Stage 2a will be moved here:
http://www.tcec-chess.net/stage_2a.php

Read the latest TCEC news here:
https://www.facebook.com/tcec.chess (you can read without having a FB account)

Hope to see everyone there!

Best Regards,
Martin Thoresen
http://www.tcec-chess.net

[Martin Thoresen]

Now live!

[BB+]

Hiarcs played a couple of interesting draws, firstly against Junior (who was the protagonist in making it exciting), when Black managed to generate enough counterplay to hold. I don't know if passive defense, rather than Ng4 and particularly b4 on move 32, is better, but it worked for Junior here. Hiarcs put way too much confidence in the b6/c7 pawns, when they were immobile.

Hiarcs also was pressing against Houdini. Somewhere around moves 30-35 there might be improvements, particularly on the timing of h3 (played on move 32). Again Hiarcs ended up in a drawn rook endgame after a long forced sequence. A better chance for Hiarcs at the end was to give up the c-pawn and hope to win something like R+gh versus R+e, though this seems not quite to work.

I must say that it looked like Houdini was going to get a draw against Critter, but 44. b6 and 46. Kd2 tore up this picture. Main points are that 44. b6 Kf8 45. Qxf6! and 46. Kd2 Qg7 47. Qf5 Rxb6 48. Rh3 Bc8 49. Bd3 Nf6 50. Rf3 Ne8 51. Rg3! both win for White. This game had Black suffering for a long time, with an e5/e6 Benoni breakthru, but then in converting to endgame White had to keep enough wood on the board to make e7 feasible, while not allowing Black to give up a piece for pawns.

Rybka has had two rather boring draws, the first against Stockfish in a QID that was 0.00 for 50+ moves, and just now against Spike with a blocked board. Rybka did defeat Critter, it seems that Black essentially missed c4 on move 21.

[Martin Thoresen]

Standings after Round 3:

Code: Select all

N Engine           Pts   SB S R J H C H G S 

1 Stockfish 210213 2.5 1.50   =         1 1 
2 Rybka 4.1        2.0 3.00 =       1   =   
3 Junior 13.3      2.0 2.25       = =   1   
4 Hiarcs 14        2.0 1.75     =     =   1 
5 Critter 1.6a     1.5 2.50   0 =     1     
6 Houdini 3        1.5 1.00       = 0     1 
7 Gaviota 0.8602   0.5 1.00 0 = 0           
8 Spike 1.4        0.0 0.00 0     0   0     

All games can be found here:
http://www.tcec-chess.net/stage_2a.php

[Martin Thoresen]

Standings after Round 6 of 14:

Code: Select all

N Engine           Pts    SB S R C H J H S G 

1 Stockfish 210213 4.5 10.50   =   = 1 = 1 1 
2 Rybka 4.1        4.0 11.50 =   1   = 1 = = 
3 Critter 1.6a     4.0  9.25   0   1 = 1 = 1 
4 Hiarcs 14        3.5  7.50 =   0   = = 1 1 
5 Junior 13.3      3.0  7.50 0 = = =     = 1 
6 Houdini 3        2.5  6.00 = 0 0 =     1 = 
7 Spike 1.4        1.5  5.50 0 = = 0 = 0     
8 Gaviota 0.8602   1.0  3.25 0 = 0 0 0 =     

All games can be found here:
http://www.tcec-chess.net/stage_2a.php

[Martin Thoresen]

Standings after Round 9 of 14:

Code: Select all

N Engine           Pts    SB St Ry Cr Hi Ho Ju Sp Ga 

1 Stockfish 210213 6.5 23.00    =  =  =  =  1  1= 11 
2 Rybka 4.1        5.5 24.25 =     1= =  1  =  =  == 
3 Critter 1.6a     5.5 23.25 =  0=    1  1  == =  1  
4 Hiarcs 14        5.0 19.50 =  =  0     == == 1  1  
5 Houdini 3        4.5 17.25 =  0  0  ==    1  1= =  
6 Junior 13.3      4.0 17.00 0  =  == == 0     =  1  
7 Spike 1.4        2.5 13.00 0= =  =  0  0= =     0  
8 Gaviota 0.8602   2.5 10.25 00 == 0  0  =  0  1     

Current live game can be viewed here:
http://www.tcec-chess.net/live.php

Finished games can be found here:
http://www.tcec-chess.net/stage_2a.php

[BB+]

Round 10:
Spike-Hiarcs had enough missteps from White to give Black a victory. This sort of technical endgame seems difficult to play well. It seems though the main error by White occurred somewhere between moves 40-45 (particularly moving the King away from the g-pawn, rather than bringing the Knight to bear on e5).

Houdini-Critter showed the former's prowess in king attacks with major pieces. Houdini saw the score jump before Critter did (as indeed also occurred when Houdini lost the previous matchup). The endgame had some chances to get blocked, but a5 was the can opener that led to victory.

Gaviota-Junior was a throwback game, when one side (here White) manages to think it has an advantage for many moves of shuffling, just to find that the opponent suddenly does, who then shuffles to victory after another heap of moves.

Stockfish-Rybka seems to be a drawn opening that went quickly to a repetition.

Round 11:
Hiarcs-Critter. White thought it has a plus in a dry endgame, but the extra pawn was meaningless, and Black then managed to take control. The real action was at move 55, where White's Re2 seems to lose to Bh5 (not played by Critter), while Rf5 appears to draw. After Critter's Rxb3, I think it is a draw, though Black might still have some chances. One problem is that the RPPP vs RN endgames are often draws by giving up the knight and reducing to the a-c pawns (note that this requires special knowledge or 6 piece TBs).

Houdini-Gaviota, the exchange sac at move 20 makes it unbalanced, though White is not really that much better. The Queen trade by Gaviota (move 34) might not be best, as White appears to have superior chances in the endgame. As with other games, defending (and playing) the ensuing technical endgame is tricky. Gaviota was unable to keep pace in the end, though both sides can probably see some improvements. About the last chance for Black was move 65, where Rxb5 leads to an endgame with Q vs RN (maybe with an extra pawn for White after a bit), where Black has some fortress possibilities.

Spike-Rybka, it seems to me that White played aimlessly and succumbed.

Stockfish-Junior shows that you can get unrealistic evals when playing without tablebases. After 65. fxg5 it is definitely a draw, but fiddling around with Kd4 should not win either.

Round 7 had a good example of outsearching in Houdini-Junior. Rounds 8 and 9 had a bunch of draws, some of them rather lifeless. Critter-Rybka never diverged much from equality, but was a bit more interesting than some. Houdini-Hiarcs had White with a useless extra pawn in a drawn endgame.

[Martin Thoresen]

Wow, what a finish!

Houdini, Stockfish, Rybka and Hiarcs (!) qualifies for Stage 3, while the rest is out. Stage 2b will start in a few days, I will make a new thread for that.

Final standings after 14 of 14 rounds:

Code: Select all

N Engine           Pts    SB Ho St Ry Hi Cr Ju Sp Ga 

1 Houdini 3        9.0 56.25    == 01 == 01 11 1= =1 
2 Stockfish 210213 9.0 54.75 ==    == == == 1= 1= 11 
3 Rybka 4.1        8.5 56.75 10 ==    == 1= =1 =1 == 
4 Hiarcs 14        8.5 51.00 == == ==    0= == 11 11 
5 Critter 1.6a     8.0 50.50 10 == 0= 1=    == == 11 
6 Junior 13.3      5.5 32.75 00 0= =0 == ==    =0 11 
7 Spike 1.4        5.0 32.00 0= 0= =0 00 == =1    01 
8 Gaviota 0.8602   2.5 18.00 =0 00 == 00 00 00 10    

All games: http://www.tcec-chess.net/stage_2a.php

[BB+]

Round 12: Spike-Critter was pretty lifeless. Spike thought it had a slight advantage for awhile, but I don't see it. Rybka-Junior turned into a KID-like attack (w/o Bg7), but Junior fiddled with its Bc8 rather than push on the K-side. Black was summarily stomped. Hiarcs-Gaviota was a Siclian where Black's King was never comfortable. After Rg8 on move 17, it might already be over. Stockfish-Houdini was a tense game, with a passive Nb5 sacrifice (unaccepted) to get passed pawns. White ended up a pawn ahead, with it then become equal material with White having an a-file passed pawn, but Black seemed to defend without too much effort (even if it took awhile).

Round 13: Spike-Junior, nice R1xg5 exchange sac, and it led to a victory once the c-pawn got going. Not sure whether Black can defend or not. Critter-Gaviota, again Gaviota had a technical endgame to defend. Although "no progress" appeared a possible outcome, Critter made the win look rather easy, especially once a7 and f4 were played. Rybka-Houdini was a payback game, with Black winning both points in this matchup. Here White appeared to overestimate the value of pawns when giving up a piece for them (move 20). Then it was largely over, as the evals tended more to Black almost every move (however, most of the next 10 moves were nonobvious for Black, and Houdini did well to choose the best path). Finally the attack broke through, which Houdini seems to see before other engines. Hiarcs-Stockfish also saw a piece for pawns trade, but here Black seemed to expect a repetition to ensue, while White played on. The endgame offered marginal chances, and Hiarcs ran out of pawns in the end, allowing the draw.

Round 14: Spike-Gaviota, Spike ended the Stage on a 2-game winning streak. Black was looking a bit better until White got the queenside movement going with b4 (move 33). Black then just had too many weaknesses and succumbed. Junior-Houdini saw a strange h4 lunge by White, but then it seemed that Black was making little progress for a long time. Junior eventually chose to move b3 and double rooks on the a-file, and Houdini seized the chance to attack his King (White's f6 locked his possibilities, and only allowed Black to leak though). The clever King slide with Kg8, and then the final d5 push (after preparation) led Black to victory. The PGN is missing the engine output. Critter-Stockfish was largely anemic, only a minor White edge after the Queens went off on move 11.

Rybka-Hiarcs saw Black give up his queen for a rook, minor piece, and pawn from the sequence starting at move 14. White might prefer simply to play hxg3, but it is hard to say. Hiarcs thought White had more of an advantage than Rybka for the next 20 moves, but as the pieces started coming off, Rybka starting preferring White more strongly. Then Black gave a pawn but had some sort of dynamic fortress (both sides having cfh-pawns), finally choosing to simplify to a draw in KQP versus KRN. Neither side had 6-piece tablebases, and Rybka overvalued this much more than Hiarcs did.