Stockfish had an eval advantage close to 1 from the opening in game 27. There were many early exchanges and by move 21 only RBN vs RBB remained. Stockfish had a slightly better pawn structure, Houdini seemed solid enough and had a bishop pair.
The engines shuffled for a while with evals about the same. Houdini locked the queen side pawns, Stockfish thought this was a mistake and its eval jumped close to 2. Houdini's pieces were blocked and Stockfish's king slowly walked all the way to g4 on the king side.
The additional piece on the king side was enough to cause Houdini's defense to crumble. Stockfish could play h5, or it could free a piece for another task as the king takes over, while the black pieces could only stand and watch. After carefully arranging its pieces Stockfish forced a bishop exchange and created a passer on the king side. The game was adjudicated before Houdini lost a piece.
In game 28 more pieces and pawns remained on the board. Houdini had a small eval advantage when the engines began to shuffle after move 20. About 45 moves later, with a few pawn moves here and there, the engines exchanged the remaining knights and continued to shuffle. Eventually Stockfish opened the position, leading to many exchanges and a tablebase draw on move 155. This was the fourth opening in a row with a 1.5-0.5 score for Stockfish.
Game 29 started with a long book sequence and an eval advantage for Stockfish. People on the chat didn't like Stockfish's early c6 move, though who can argue with such a strong engine. After several exchanges the position on the board was simplified with low evals. On move 32 the white c pawn became an advanced passer.
The c pawn advanced to c7 where Houdini blocked it. Stockfish's eval climbed over 1 yet Houdini thought it could hold the position. The engines exchanged queens and then Houdini decided to give a rook for a bishop in order to get rid of the advanced passer. The result was a RR vs RB ending with very few pawns, both evals were over 1, it was not clear what Stockfish could do to improve.
Stockfish's eval started to increase, slowly at first and then it jumped over 10, Boom! Suddenly the black king was in danger of a backrank mate. The 3 pawns in the center also restricted its movement. Houdini's eval stayed around 1 for 6 moves before it saw the danger, unbelievable.
The game was adjudicated quickly, the PV showed how Houdini was going to lose a piece. We haven't seen an engine blind to the danger for so long in a late stage game in TCEC for a long time. I was reminded of a similar situation in season 5, but I wouldn't have believed this could happen in this superfinal.
In game 30 Stockfish stormed the king side and created a hole in the pawn wall. Stockfish's eval was over 1, Houdini still felt safe though its king was exposed and its pawns were blocking the way for its own pieces.
Houdini was a pawn up and it made a weak attempt to counter on the queen side. Stockfish had many pieces on the king side, Houdini was forced to give back the pawn and allow its pieces to help. The white king was under a lot of pressure, unable to move and facing two black pawns supported from behind.
Houdini exchanged all the pieces it could to save its king. In the process Houdini lost a knight, and the game reached a RRN vs RR position with black a pawn up, a win for Stockfish. This is the fifth straight opening that Stockfish wins, and this time it won with white and with black. The last time this happened in a TCEC superfinal was in season 5, Komodo had a double win in games 25-26 against Stockfish.
No comments:
Post a Comment