Season 12 superfinal, games 47-54

After 54 games Stockfish leads 11-6 with 37 draws. The draw rate continues to be low, now at 68.5%.

Evals stayed close to 0 in game 24.1. The engines castled in opposite directions, neither attacked with pawns. Instead the engines shuffled and exchanged pieces, the game reached a RB vs RN position and ended in a repetition. Game 24.2 was similar. Queens were exchanged early and by move 34 only RBN vs RNN remained. Komodo created a queen side passer and pushed it to the 7th rank. Stockfish exchanged everything to stop the passer, drawing the game in a knight vs pawns tablebase position.

The engines exchanged almost all the pieces by move 27 of game 25.1, only RB vs RB remained. The game reached a tablebase draw on move 42. Komodo gave a pawn early in game 25.2 and Stockfish thought it had an advantage for a while. Komodo attacked the black king through the open queen side, evals came down to 0. After a series of exchanges Stockfish gave a rook for a bishop. The game reached a RN v BN position with Stockfish 3 pawns up, evals were at 0 and the game ended by the draw rule.

The engines castled in opposite directions in game 26.1, Both engines pushed pawns towards their opponent's king. Stockfish's eval started around 1 and drifted down slowly. Komodo lost a rook for a knight and pawn, after exchanging queens only RRB vs RBB remained.



The evals started to increase as Stockfish outplayed Komodo in the ending. First it doubled rooks on the h file and won a pawn. Komodo's king found safety behind a white pawn. Then Stockfish exchanged pawns and opened more files for its rooks.




Komodo played defensively and refused to exchange rooks. Stockfish was able to drive the rook away and started to push the c pawn. The game was adjudicated when the pawn reached the 6th rank. Komodo could not avoid exchanging rooks and was about to lose a bishop as well.

In game 26.2 Komodo pushed pawns on the king side while Stockfish did not push on the queen side. Instead it gave a rook for a knight and pawn and opened a file on the queen side, weakening the white king's position.




Evals slowly increased as more and more pieces were exchanged. On move 41 only RN vs BN remained and evals were close to 2.




Komodo had a pawn majority on the queen side and its pawns slowly moved up the board. Stockfish tried to counter with its king side passer but it was stopped on the 2nd rank. The game ended with Komodo capturing the bishop with a nice combination. This opening had a biased result, though the two games were not similar.

Stockfish started game 27.1 with an eval around 1. It kept its king in the center uncastled and pushed pawns on the king side. Komodo exchanged a few pieces and managed to defend its king side pawns despite the open files and white's rooks. Komodo gave a pawn and exchanged a pair of rooks, evals fell down to 0. The game was adjudicated a draw with several pieces still on the board. In game 27.2 the evals stayed low, Komodo castled and did not open the king side. After many exchanges, in which Komodo gave a rook for a bishop, the game reached a QN vs QR position. The black king was under threat, Stockfish captured the knight and Komodo ended the game with perpetual check.

Season 12 superfinal, games 39-46

After 46 games Stockfish leads 10-5 with 31 draws. The draw rate continues to be low, now at 67%. Stockfish is firmly in the lead as the match approaches the halfway point.

Stockfish's eval increased over 1 after the start of game 20.1. On move 24 Stockfish went a pawn up and created a queen side passer. Komodo locked the king side and got the pawn back, evals started to come down. The white passer was stopped on the 6th rank and then the engines shuffled for a while. The engines started to exchange pawns and pieces, Komodo created a pair of central passers. Evals were 0 for a long time, the game ended with a perpetual check. In the reverse game 20.2 there were many early exchanges and the game reached a RB vs RB position on move 29. Evals came down as the engines exchanged pawns, the game was adjudicated when the evals were low enough to activate the draw rule.

Evals stayed close to 0 throughout game 21.1, despite the opening which was a King's Gambit variant. Stockfish sacrificed a knight for pawns early in the game, this removed the threat on its king. Komodo was behind in development, the engines exchanged most pieces and got a R vs BN imbalance with Stockfish 2 pawns up. On move 26 only RB vs BBN remained, Stockfish thought it had a small advantage. Komodo push a passer on the king side, after Stockfish captured it the engines shuffled until draw adjudication. The reverse game 21.2 started with a knight and bishop sacrifice by Komodo. Stockfish gave the bishop back and exchanged queens, Komodo put a rook and a passer on the 7th rank. Evals were 0 as all this was happening. On move 31 only RN vs RBN remained, and Stockfish pushed a second passer to the 7th rank. Komodo stopped a passer and traded pieces to get a knight ending and a repetition draw.


Stockfish's eval constantly increased after the start of game 22.1. All the pieces were on the board in a closed position, the first capture occured on move 35 when Stockfish went up a pawn. After a few exchanges the position became static and the engines started shuffling, Stockfish eval close to 2 and Komodo's eval close to 1. It was a bit surprising that Komodo chose to move a pawn on move 68.



Stockfish's eval jumped a level, there was another piece exchange that changed the pawn structure a little. Then the engines resumed their shuffling. The white king walked to the queen side to get away from a potential attack on the h file. Komodo was playing on increment for a long time, yet the evals remained constant. The 50-move counter reached 15 when Stockfish captured a second pawn with an eval jump, it thought Komodo finally made a mistake.




Komodo's eval started to increase very rapidly, it saw it was in trouble. Stockfish had 3 pawn breaks available on the a, e and h files and Komodo scrambled to rearrange its pieces in defense. Eventually Stockfish gave a rook for a knight, taking out an important defender and creating a queen side passer.



The white queen side pawns began to march and Komodo's defense collapsed when two connected pawns reached the 6th rank.

In the reverse game 22.2 the engines opened a file on the queen side early. Komodo's eval increased from the start and then came back down under 0.5. The engines locked the center, exchanged a few pieces and started to shuffle. Komodo avoided the 50-move draw but the game was adjudicated by the draw rule a few moves later.

The 23rd opening was only one ply long, 1.g4!!, the Grob opening. In game 23.1 evals started around negative 1 and drifted slowly downwards. Stockfish kept its king in the center uncastled. It pushed pawns on the king side, Komodo won a pawn on the queen side, both sides did not look safe for the white king. Komodo gave two pawns and created a central passer, both evals jumped over 1 while Stockfish chose to castle long.




The engines had a long PV agreement, evals slowly increased as they played the line, Komodo's passer reached the 3rd rank. The black major pieces moved to the queen side, threatening the white king. Stockfish gave a bishop to remove the passer and evals jumped over 2. After exchanging queens only RR vs RRN remained.


Komodo captured a pawn and exchanged a pair of rooks, the R vs RN ending was a win.

In the reverse game 23.2 Stockfish's eval started at 1.5 and increased steadily. Both kings stayed in the center uncastled. Stockfish pushed pawns on the king side and Komodo sacrificed a knight for two pawns to open up the king side and expose the black king. Stockfish's knights were very active, both with outposts on the 4th rank.



The engines played out a long line that appeared in both their PVs. After exchanging rooks Stockfish captured a pawn and created passers on the queen side. The engines exchanged queens and traded pawns until only RN vs RBN remained, Stockfish with its last pawn and its eval approaching 4.




Stockfish captured one of the white pawns, and the game was adjudicated without playing out the win. Stockfish's PV suggested that Komodo was going to lose the knight, Komodo's PV headed for a won knight ending. Both engines won this opening with black, first biased opening of the match.

Season 12 superfinal, games 33-38

After 38 games Stockfish leads 8-4 with 26 draws. Komodo is still behind but it is fighting back. The match is not really one-sided like in season 11,  this makes it much more interesting. There are many decisive games, the draw rate is at 68%.

Evals stayed close to 0 throughout game 17.1, despite Komodo having a pawn advantage from the start. On move 20 the game reached a BN for R imbalance and evals were 0. Komodo had an advanced passer on the queen side, Stockfish broke through the king side with a knight sacrifice and ended the game with a repetition check.

In the reverse game 17.2 Stockfish was a pawn up and had a negative eval of about 0.5. Komodo locked the center and Stockfish concentrated forces on the king side. Stockfish's eval came down after Komodo exchanged pawns on the king side. Komodo then gave another pawn and created a central passer, its eval starting to increase.




The white passer on e6 could not be touched and was a constant threat. Stockfish's eval increased as Komodo took the initiative. Komodo captured the two weak pawns on the queen side and equalized material. Stockfish had no counter, it had to block the passer and had little room to move. In desperation it gave a knight for the passer. 




Komodo traded down to a QNN vs QN position.The extra piece was enough to give Komodo the win, though the game was adjudicated before the win was clear. In Komodo's PV it captured the black queen a few moves later. Komodo is fighting to stay in the match.

Stockfish's eval was about 1.5 from the start of game 18.1, a Benoni opening. Komodo's position was very cramped, Stockfish pushed pawns on the king side and opened the h file. Komodo's king was in danger while all the black pieces were on the queen side, Stockfish's eval was over 2.



Komodo's counter was a passer on the queen side. Stockfish attacked on the h file, Komodo gave two pawns in the hope that the white pawns will shield its king. The evals continued to increase all the time.




Stockfish exchanged pieces and won two more pawns. After exchanging queens the game was adjudicated, the material advantage was enough for a win.

The reverse game 18.2 developed very differently. Stockfish opened a file in the center and another on the queen side, allowing its pieces to move more freely. The engines exchanged pieces until reaching a drawn rook ending. 

Game 19.1 started with a knight sacrifice by Stockfish that opened the king side and exposed the black king. Komodo immediately gave the material back and exchanged queens. The engines continued to exchange pieces, the game reached a RB vs RN position on move 30. Evals came down but Stockfish thought it still had a small advantage. As a result the engines continued to shuffle for more than 40 moves to agree on a draw.

In the reverse game 19.2 the king side stayed intact, Komodo was a pawn down but had an eval advantage as Stockfish's king moved and could not castle. Komodo found a great square for its knight on e6, controlling a lot of black territory. The black king side rook was miserable in the corner.




Evals started to increase, Stockfish traded knights to get rid of the black knight on e6, Komodo simply moved its other knight to the same square. Komodo then opened the center and got the pawn back, its pieces were much better placed and the king side pawns were ready to march.




Stockfish was essentially playing a rook down, evals jumped quickly as Komodo attacked. After a series of exchanges only Q vs RB remained, Komodo with a central passer.




Stockfish couldn't stop the white passer without losing a piece, a win for Komodo. Komodo played this opening much better than Stockfish.

Season 12 superfinal, games 27-32

After 32 games Stockfish leads 7-2 with 23 draws. The number of decisive games is quite high, Jeroen's openings are working better than expected. Most openings give a substantial advantage to one side, and yet none of the openings so far resulted in two wins, one per engine.

Stockfish's eval was around 1 from the start of game 14.1. Stockfish pushed pawns on the king side, leaving its king exposed. Komodo went up a pawn yet Stockfish continued to be ahead on eval. Stockfish created a passer in the center, and Komodo was reluctant to push its king side pawns to keep its king safe.




The position became static and the engines started to shuffle. The game seemed to be heading for a draw, evals slowly drifting down. Then with 19 moves left on the 50 move counter Stockfish's eval jumped as it pushed the d pawn forward. Komodo realized something was wrong two moves too late. Stockfish gave a second pawn and opened the queen side.




Komodo's defense collapsed quickly. Stockfish lost the d pawn but its a pawn became a passer. Komodo exchanged pieces, giving a rook for a knight, then it gave a knight to stop the white a pawn.




This meant Stockfish was a rook up in a RRB vs RB position. It took a while but there was no doubt Stockfish was winning.

In the reverse game 14.2 Komodo kept the king side closed. Stockfish's eval stayed close to 0, Komodo's eval was about 0.5 for a while. Komodo created a passer in the center but it wasn't strong enough. Stockfish exchanged pieces, clearing the queen side, evals were at 0 despite Komodo's bishop pair and queen side passer. The game was adjudicated by the draw rule.

Stockfish's eval climbed over 1 after the start of game 15.1. The engines castled in opposite directions, Stockfish pushed pawns on the king side and Komodo controlled an open file on the queen side. The engines started to shuffle essentially from move 27, evals staying constant. On move 70 Komodo moved a pawn unexpectedly. It then won a rook for a bishop, Stockfish's eval did not move but Komodo's eval came down.




Komodo was playing on increment, its king found safety behind the white pawns and Stockfish realized it couldn't attack on the king side. Then suddenly evals turned negative. Komodo doubled rooks on the g file and pushed its passer, even though this weakened its king's defense. Through a rook exchange the passer reached the f3 square in a QBN vs QRB position, evals over 1 for Komodo.



Both engines were blitzing moves, for a while Komodo looked for a way to improve and then evals continued to increase. Komodo gave a pawn to open the queen side, then used this to force a queen exchange. Evals were close to 5 at this point.



The white pawns started to fall as Stockfish had to block the passer and could not protect them. The black king came forward and three king side passers were good for a win. An amazing turnaround for Komodo.

There were many early exchanges in the reverse game 15.2. On move 25 only QRN vs QRB remained. Komodo was a pawn up but evals came down to 0, the game ended in a draw when the pawns stopped moving. Komodo wins its second minimatch, after winning the first one and losing 6 since.

Stockfish's eval was again over 1 quickly in game 16.1. The engines castled in opposite directions, this time Komodo castled long. Komodo made a hole on the king side and Stockfish pushed pawns on the queen side. After a few exchanges Stockfish had the bishop pair advantage.

For a while Stockfish's plan was not clear, the pieces moved around and evals did not change much. Then Stockfish traded a bishop for a knight, losing its bishop pair but gaining a pawn majority on the queen side.

Komodo struggled to stop the white queen side pawns and to keep its king protected. It tried to counter on the king side, giving a knight for 2 pawns and threatening the white king. Stockfish exchanged queens, with an eval over 4.

Stockfish was a piece up in a RRB vs RR position.The game was soon adjdicated though we didn't get to see how Stockfish converts the win.

In the reverse game 16.2 Stockfish exchanged queens early. The engines continued to exchange pieces until on move 31 only RB vs RN remained. Evals fell to 0 and the game was adjudicated as soon as the pawns stopped moving. Another minimatch win for Stockfish.

Season 12 superfinal, games 17-26

After 26 games Stockfish leads 5-1 with 20 draws. After losing the first game Stockfish is in charge with 5 wins in the first quarter of the match. This superfinal may be as one-sided as the previous one.

In game 9.1 Stockfish started with an eval over 1 and it constantly increased. After a few exchanges the black f file was open and stockfish concentrated its pieces as it prepared for a king side attack. By move 22 both evals were over 2.




Komodo placed a rook on f5 to block the white pawn and prevent a further weakening of the king side. Stockfish could take the rook with its bishop but first it moved its king to the corner and captured a pawn to clear the queen side. Evals continued to increase, the end was in sight. 




Stockfish opened the g file and then allowed Komodo to exchange rooks, without losing material. Instead Komodo won a pawn and equalized, though this opened the f file as well.



The white pieces were too strong and the black king was vulnerable.The game was adjudicated before Stockfish started to win material, the PV showed it was about to win a piece.

In the reverse game 9.2 Komodo's eval started low and increased to about 0.7. There was no king side attack, instead the engines exchanged pieces until only RN vs RN remained on move 30. Komodo was a pawn up with central passer, Stockfish gave a pawn to create a passer on the queen side. Both engines captured the other's passer to avoid queening, then Stockfish got a pawn back. Komodo's advantage was not enough in a rook ending, Stockfish held the draw. Another minimatch for Stockfish, a missed chance for Komodo.

Stockfish had an eval of under 1 in game 10.1, there were almost no exchanges until move 24. After a series of exchanges Stockfish was a pawn up, with a pawn majority on the king side. Evals came down as Komodo captured a pawn on the queen side and equalized. The engines gradually exchanged all the pieces and the game ended in a tablebase draw. In game 10.2  the evals stayed close to 0. Stockfish thought it had a small advantage for a while, but after exchanging pieces and opening the queen side both evals came down to 0. The game ended in an early draw with many pieces still on the board.

In game 11.1 Komodo's eval was 0 from move 13 and Stockfish's eval came down to 0 a few moves later. The position was open with both kings uncastled in the center. The engines quickly exchanged most pieces, Komodo gave a rook for a knight and the game reached a RR vs RN position on move 27. The white pawns were vulnerable, Komodo managed to capture 2. The game ended with the white king trapped, just before a repetition draw. In game 11.2 Stockfish captured a pawn and its eval was negative from the start. The king side was open and both engines castled long. Stockfish had a connected pair of passers on the king side and Komodo had an advanced passer in the center. Stockfish gave a bishop to stop the white passer, and managed to keep one passer on the king side. The game reached a QN vs Q position, and ended in a draw  just before Komodo lost its knight for the black passer.

The pawns in the center and on the queen side were locked in game 12.1. Komodo pushed pawns on the king side, Stockfish's eval was around 1, it had more space and the black king was exposed. Evals dropped to 0 as Stockfish gave a rook for a knight and pawn, opening the king side. The engines exchanged most pieces and the game ended in a repetition draw. In game 12.2 the engines opened the queen side and exchanged a few pieces. Then after mostly shuffling for 25 moves the engines exchanged down to a QBN vs QBB position, evals came down to 0 and the game ended in another draw.

Stockfish had an eval advantage from the start of game 13.1, its eval rose to about 1.3 and Komodo's eval to about 0.7. Komodo kept its king in the center uncastled as it pushed pawns on the king side. After a series of exchanges only RBB vs RNN remained, evals staying constant for a while.




Both engines had passer, Stockfish's bishop pair was an advantage in an open position. The white pawns slowly advanced on the king side, and Komodo chose to allow the f pawn through without exchanging. This meant both engines had connected pairs of passers, the evals suggested that Stockfish's pair was stronger.




Evals jumped higher as Stockfish managed to place its rook behind the black pawns. Komodo's king walked to the queen side and captured a pawn, Stockfish moved its king forward and Komodo gave a knight for one of the passers. The game was adjudicated before Komodo lost more material for the second white passer.

Komodo played the opening of the reverse game 13.2 in a completely different way. Its eval was close to 0 while Stockfish's evals turned negative after a while. There was a pawn exchange on move 18 and the engines started to shuffle, with no exchanges until move 60. After opening the queen side the engines exchanged most pieces until only RBN vs RBN remained. The evals came down to 0 and the game ended in a draw. Stockfish handled this opening much better.

Season 12 superfinal, games 9-16

After 16 games Stockfish is leading 3-1 with 12 draws.

Stockfish had an eval advantage from the start of game 5.1, Komodo was less impressed as usual. The engines castled in opposite directions, Stockfish pushed pawns on the king side facing the black king. After opening the king side Stockfish had two pawns that looked dangerous, Komodo blocked them and used them as shields. Komodo had a passer in the center and its eval was still low.



Stockfish cleared the black queen side pawns and created a passer, Komodo exchanged pieces and removed the immediate threat to its king. The game reached a QRB vs QRN position, Stockfish's eval was over 2 but it was not clear how it would improve.




For a while Stockfish moved its pieces with no apparent purpose. This changed when it managed to capture the d pawn. Now it had a connected pair of passers on the queen side, still on the 2nd rank but very threatening in the long run, especially after the queens were exchanged.




Komodo did not have enough strength to stop the advancing passers, the game was soon adjudicated. 

There were many early exchanges in the reverse game 5.2. Evals were close to 0 after only RB vs RB were left on move 33, the game ended as soon as the pawns stopped moving and the draw rule was activated. Stockfish in the lead after 5 openings.

Stockfish started game 6.1 with an eval around 1. The engines castled in opposite directions, Stockfish pushed pawns on the king side. Evals started to increase as the white king moved to the corner away from Komodo's pieces. Stockfish opened the king side and its eval was over 2.




Komodo exchanged pieces to stop the king side attack. After it gave a rook for a knight and pawn the game reached a QRR vs QRN position, Stockfish's eval was over 3.




Stockfish avoided exchanging queens, it used the exposed black king to make threats and checks. In this way it managed to capture a pawn and only then to exchange queens. Evals continued to increase all the time.




It was clear that Stockfish was winning, it was just a matter of technique. The white pawns started to advance on the queen side, supported by the king. When Stockfish captured a pawn and created a passer the game was over.

In the reverse game 6.2 evals stayed low. Both engines castled short and there were no attacks. After a series of exchanges most pieces were gone by move 39, and a few moves later the game reached a queen ending. Evals were 0 and when pawns stopped moving the draw rule was activated. A second straight minimatch win for Stockfish.

Stockfish started game 7.1 with a rook for a knight advantage, eval around 1.4. Komodo's king was uncastled in the center, it found safety on the queen side. After move 25 evals started to fall, the position opened and the engines started to exchange pieces. On move 36 only RRB vs RBB remained, Komodo's bishop pair and advanced passer were strong enough to counter Stockfish's rook pair and hold the draw.
In game 7.2 evals turned negative despite the material advantage for white, Komodo's queen side was undeveloped and Stockfish attacked on the king side. Komodo exchanged queens and bishops to stop the immediate threat, Stockfish pushed pawns on the king side and created an advanced passer. Komodo gave a rook for a knight to remove the passer, the game reached a RBN vs RBN position on move 40 and the evals started to drift down to 0 and a draw.

In game 8.1 most pieces were on the board in an open position, Stockfish's eval fluctuated under 1 and Komodo's eval was around 0.5. Most of the action was on the queen side. After exchanging queens on move 43 the evals started to come down, and they dropped to 0 when the game reached a RN vs RB position on move 57. The game ended in a tablebase draw. In game 8.2 there were many early exchanges. Stockfish's eval was close to 0 throughout the game, Komodo's was close to 0.5 for a while and then came down. After exchanging queens the game reached a RN vs RB position, the game ended in a repetition.

Season 12 superfinal, games 1-8

This season's superfinal is between Stockfish and Komodo.

After 8 games the engines are tied 1-1 with 6 draws.

The first opening of the season was a King's Gambit. In game 1.1 evals started close to 0 with a small advantage to Komodo. Komodo gave the pawn back, and although Stockfish's king was uncastled and exposed it seemed relatively safe. Evals started to increase as Komodo pushed pawns on both sides, the white king walked towards the corner and trapped its own rook.



Stockfish moved the g pawn to free its rook, exposing its king further. Evals were over 2 after Komodo gave a rook for a bishop, it got the material back a few moves later and the game reached a QRB vs QRB position.




Material was equal but the white king was too vulnerable. Komodo gave 2 pawns and opened a file for its rook, a mate threat forced Stockfish to lose a rook for a bishop.



The game was soon adjudicated, Komodo was about to exchange queens for the win.

In the reverse game Stockfish kept the negative eval around 0.8 from the start. It was a pawn up but did not attack the exposed white king. Komodo exchanged pieces and on move 30 the game reached a RRB vs RRB position. Evals came down and Komodo held the draw. Komodo wins the first minimatch.

Stockfish started game 2.1 with evals around 1. Komodo's eval dropped close to 0 by move 20 while Stockfish remained hopeful. It gave a rook for a bishop, and its strong bishop pair attacked the exposed black king. Komodo traded a knight for one of the bishops and evals dropped to 0. The game ended in a drawn B vs R tablebase position (6-man, first time it is used in TCEC).
In the reverse game the center was locked, there were long pawn lines across the board and very few exchanges. Stockfish had more space on the king side and its eval turned negative. For many moves Stockfish planned to open the king side and attack, its eval constant at 1 while Komodo's eval was 0. The attack never came and the game ended in a 50 move draw.

Stockfish started game 3.1 with an eval close to 1 while Komodo's eval stayed close to 0. The black king was uncastled and Stockfish dominated the dark squares on the king side. However, Komodo kept its pawn line intact and strong and Stockfish did not find a way in.



Evals started to jump when Stockfish played g4 and made a small hole in the pawn line on the king side. Komodo's pieces had no access to the king side, and Komodo tried to counter on the queen side. The evals of both engines implied the end was near.



Stockfish sacrificed a knight to open the king side completely. The game reached a QR vs QRN position, Stockfish with an advanced king side passer and the black king under threat. Adjudication came a few moves later. 

In the reverse game 3.2 Komodo had an eval advantage from the start. It did not sttack the king side, the pawn lines stablized and the engines started to shuffle on move 20. On move 52 Stockfish opened the king side, the game reached a QRR vs QRR position and the evals quickly dropped to 0 ending the game with a draw. Stockfish wins its first minimatch and equalizes the score.

Stockfish was up a pawn with eval over 1 at the start of game 4.1. Komodo exchanged pieces until only QR vs QR remained on move 29. Evals stayed constant for a long while, Stockfish had two passers that weren't strong enough. Eventually the game reached a drawn rook ending. In game 4.2 Stockfish's eval was 0 from move 20. The engines gradually exchanged pieces and Komodo's eval came down as well. The game reached a drawn rook ending, like the previous game.

Season 12 premier division final results

Crosstable after RR8



Komodo and Stockfish qualify for the superfinal. Komodo with 4.5/6 outscored Stockfish and won the division. This is despite having lost once to Stockfish, mainly due to a better result against Jonny. Houdini was close behind the two leaders but could not close the gap with a RR score of 4/6.

Jonny and Chiron are relegated to division 1. Jonny lost 3 games in the last RR, including a fourth loss to Ginkgo in the division.

Fire finished 4th after beating Andscacs in the reverse game.

The superfinal is due to start on Monday 18.6.18, it will consist of 100 games, 50 openings chosen by Jeroen Noomen. The time control will be 120+15, game duration will be close to 4 hours. We haven't had a Stockfish - Komodo superfinal since season 8. Stockfish is my favorite, though I hope it won't be as one-sided as the previous season's match against Houdini.

A few interesting games

Houdini - Komodo: Houdini was a pawn up from the start, with eval close to 1. Komodo reduced to a rook ending, and the pawn advantage was only good for a draw. Houdini could really use a win in this game, it is still 3rd 1 point behind Komodo.

Fire - Andscacs: Fire had an eval advantage from the start, Andscacs gave a rook for a knight early in the game. The engines traded most pieces and on move 32 only RR vs RB remained. Andscacs had 2 extra pawns, evals were a little over 1 and increased gradually. The endgame was slow and technical, it took Fire more than 40 moves to reach a winning 6-man position. Fire wins after losing the reverse game, it opens a 1 point gap in the race for 4th place.

Jonny - Komodo: Jonny blundered in the opening, Komodo gave a pawn but its eval jumped over 1. Jonny's king was uncastled in the center, exposed to attacks. By move 22 Komodo regained the pawn and its eval was over 2. Komodo went a pawn up and then slowly exchanged pieces. The game reached a QN vs QB position, Komodo two pawns up and winning. Komodo improves to +12

Stockfish - Andscacs: Stockfish had an eval advantage from the start, the engines exchanged pieces and the game reached a RRB vs RRB position on move 29. Stockfish blocked the queen side and managed to create a passer on the king side. The evals started to increase, it took 20 more moves to convert the win. Stockfish improves to +13.

Houdini - Ginkgo: Houdini had more space and evals gradually increased. On move 35 evals were over 2, Houdini opened the position and Ginkgo gave a knight for two pawns. It took a while for Houdini to convert but it was in control and Ginkgo had no defense. Houdini keeps its hope alive, with +10 it is still 1 point away from 2nd place, only 3 games left.

Andscacs - Komodo: Komodo's eval turned negative after it created a central passer, evals stayed close to 0 for 40 moves. Komodo's knight moved to the 3rd rank, and after opening the king side the evals jumped over 3. After a series of exchanges the game reached a RBN vs Q position. The black king used checks on the exposed white king to create double threats on Andscacs' pieces, and the passer was traded for the white bishop. The remaining black pawns were soon unstoppable. Komodo and Stockfish are tied with +13. Houdini has 3 more games, white against Fire and Jonny, black against Stockfish. It needs wins to qualify for the superfinal.

Houdini - Fire: Houdini couldn't get anything going in the game, it ended in a draw with opposite color bishops.
 
Houdini - Jonny: Houdini created a queen side passer on move 15, evals stayed close to 0 but started to increase after a few pieces were exchanged and the passer became stronger. On move 35 Houdini added a second connected passer and evals were over 1. Jonny had to block the advancing pawns on the queen side, Houdini opened the king side and attacked the black king. Jonny's defense collapsed and the game was adjudicated before it lost material. Houdini is 1 point behind, its only hope is to beat Stockfish as black in the last game.

Jonny - Ginkgo: Jonny created a central passer on move 22, it reached the 6th rank but stopped there. Evals were close to 0 until Ginkgo gave a pawn and created a supported passer on the 3rd rank on move 27.  Jonny gave a rook for a knight and evals jumped over 1, then it traded Q for RB with evals over 2. Ginkgo captured the weak white passer, and to complete the win Ginkgo gave back a rook for a knight and vreated a second connected passer. The two black pawns were unstoppable.

Komodo - Fire: See featured game below. Komodo needed a draw to ensure it qualifies for the superfinal, and won the game.

Stockfish - Houdini: Houdini still had a chance to reach the superfinal. If it won it would be tied with Stockfish and the win would also give it the head to head tiebreak. However, the opening favored Stockfish and all Houdini could do was hold the draw. After 30 moves Stockfish's came down, and it took another 45 moves until the game was adjudicated a draw. 

Featured game: Komodo - Fire
Premier division, round 56
Link to game on TCEC

The evals were close to 0 for the first 25 moves. The center was blocked and most pieces were on the board. Evals started to increase and Komodo's eval was more than 1. The engines continued to shuffle behind the pawn lines. Fire had a bishop pair, but its bishops did not have room to move, especially the DSB. Komodo had more space and a knight which has an advantage in a closed position.




The engines exchanged a few pieces, for a while the evals remained constant but then started to increase again. The black DSB could not move because the white knight controlled its potential squares. Then the white queen penetrated the black pawn line.




Komodo captured the a pawn and Fire had to deal with the queen side passer. The black defense was collapsing under pressure and Komodo forced a queen exchange.





Fire could not stop the white king walking to the queen side, and then the a pawn resumed its march forward and the game was over.

Season 12 premier division statistics

Note: I disregarded the 5 Chiron games in the report.


Draw rate, wins

 

Final draw rate was 64.3%.


Game termination

The three most common game termination causes were:
35.7% - TCEC win rule
34.5% - TCEC draw rule
14.9% - 3-Fold repetition, SyzygyTB

There was no games that ended in a crash, other than Chiron's 3 crashes that led to its early disqualification.  

Moves per game

 

Median= 61.5
Average= 65.8

There were 11 games with more than 100 moves, including one with 184.5 moves (Komodo - Fire, round 42)

Time per game (hours)

 
Median= 3:02
Average= 2:54

Note the left tail, some games tended to be shorter than expected. This is a well known TCEC phenomenon where in late stages some engines lower their eval and agree on draws as early as possible. There were 35 games with duration less than 2:30 hours, 31 of which were draws (88.6%, much higher than the draw rate).

Openings

There was no change in the opening variants information, still not as detailed as in previous seasons.

There were 8-move book openings in this stage chosen by Cato. The first letter of the ECO codes was distributed as follows:



If we use the opening 'family name' (using format FAMILY_NAME: VARIANT....) the top 3 are:
Sicillian - 38 times
English - 20 times
King's Indian - 14 time

The engines had very little freedom to choose the opening variant. In 97.6% of the game pairs the same ECO code was repeated, 92.9% repeated the same opening variant. In all of the game pairs the ECO first letter was repeated twice.

Reverse pairs, wins

 

Reverse pairs, same moves 

 

Pairs of reverse games diverged very quickly, 77.4% of the pairs diverged at most after 1 move.