Wednesday, November 7, 2018

Season 13 superfinal, games 79-86

After 86 games Stockfish leads 12-5 with 69 draws.
The draw rate continues to decrease, now at 80.2%. The sequence of decisive minimatches ended with two openings the engines drew, but then there was the first opening with a biased 1-1 result. Stockfish's lead is 7 points with 14 games left.

Stockfish had an eval over 1 at the start of game 79. The engines castled in opposite directions and evals started to increase. Stockfish had a space advantage and Komodo's pieces were pushed back and defensive. Stockfish pushed pawns on the queen side and threatened the black king.

A series of exchanges eliminated most pieces. Komodo gave a bishop for a pawn and only QRB vs QR remained on move 30. Komodo's king was hiding in the corner, most of the black pawns were invisible to the bishop. Evals were high but Stockfish still had to find a way to win.

Komodo's only hope was to use the exposed white king to threaten mate or a perpetual check draw. Stockfish had to be careful,  it managed to get its king to a square where it couldn't be attacked and brought the rook forward.

Komodo exchanged queens, it was 3 pawns up but the RB vs R ending was a win for Stockfish.

Komodo had an eval of about 0.5 after the start of game 80. There were a few exchanges, the last one on move 17 when Komodo won a pawn. Stockfish doubled rooks in an open file on the king side, Komodo blocked it with a knight. The engines started to shuffle, with a few pawn moves that locked the queen side. Evals stayed around 1, and only came down when the 50 move draw was close. This was the 5th decisive miniatch in a row, Stockfish extends its lead to 7 points.

Stockfish had an eval advantage less than 1 from the start of game 81. There were no exchanges until move 21, the center was blocked and the engines played behind their pawn lines. The engines opened a file on the king side, Komodo doubled its rook there and its eval turned negative for a while. Stockfish created a passer on the queen side and evals returned to 0. After a few exchanges the engines started to shuffle until adjudication. In game 82 Stockfish opened the center and gave a rook for a knight and two pawns. Evals stayed low, the engines exchanged down to a QRR vs QRN position. The rooks did not have open files and the engines again shuffled until adjudication.

Stockfish started game 83 with an eval around 1. Komodo gave a rook for a knight and pawn, then a series of exchanges left only RRB vs RBB on move 17. The engines started to shuffle, with a few pawns moves that reset the 50 move counter. Stockfish's eval came down on move 115, it took 20 more moves to adjudicate the game. In game 84 Komodo's eval started around 0.5 and gradually increased. Komodo castled long and Stockfish kept its king uncastled in the center. Stockfish pushed pawns on the queen side and weakened the white king's defenses. In a series of exchanges Komodo gave a rook for a bishop, only QRB vs QRR remained. The exposed black king compensated for the lost material, evals were over 1. The black king marched to the queen side, leaving Komodo to capture pawns. Stockfish gave the material back and reduced to a queen ending one pawn down. Evals came down and the game was adjudicated a draw.

Stockfish started game 85 with an eval over 1 that gradually increased. The engines opened the center and castled in opposite directions. Stockfish had the bishop pair advantage in an open position, its pieces seemed to be preparing an attack on the king side. Komodo had open files for its rooks on the queen side.

Komodo's queen side attack was weak and nothing came out of it. Stockfish won a pawn while losing one of its bishops, evals continued to increase steadily. The engines exchanged down to a double rook ending, Stockfish captured another pawn and was two pawns up.

Stockfish had passers on both sides. The 3 connected passers on the queen side were also guarding the king so Stockfish didn't push them forward immediately. First it focused on the king side, exchanging pawns and gaining a third pawn in the process.

The material advantage was more than enough to secure a win for white.

In game 86 evals were around 0.5 from the start. There were no exchanges until move 16, the center was blocked and both engines kept the kings uncastled. Komodo pushed a pawn on the king side and made a hole in the pawn line. Stockfish had already moved its queen side rook, it decided to move its king and not castle. Komodo gave a rook for a knight and created a central passer, evals jumped over 1.

Komodo found safety for its king by castling long. On move 29 the engines had a long 15 move PV agreement which they played out. Stockfish gave the rook back for a bishop, the engines exchanged queens, Komodo won a pawn and created a second passer on the queen side. After a few more exchanges only RNN vs RBN remained, Komodo with two pawns and Stockfish down to one.

This is the sort of position that only chess engines can understand, humans can stare at the evals and try to figure out what is going on, preferably using a strong chess engine... For a long while the evals fluctuated, Komodo tried to strengthen its queen side, support its pawns and allow them to move forward. Stockfish blocked the white pawns while protecting its remaining pawn. Eventually Komodo captured the black pawn and exchanged a knight for a bishop. Evals started to increase again.

The RN vs RN ending was still very tricky. It took Komodo almost 40 more moves before the game was adjudicated. The 50 move counter was down to 10 as the pawns still hadn't advanced, the PV showed that a win was still very far off yet both engines agreed it was inevitable.
This was the first opening of the superfinal with a biased 1-1 result. The book was only 4-plys long, the two games had very different characteristics. I would argue that the engines had a lot of influence on the result, more than the opening itself.

No comments:

Post a Comment