Season 18 superfinal games 21-30

After 30 games Stockfish leads 8-5 with 17 draws. Both engines won a game pair, first game pair win for Leela. Stockfish continues to lead by 3 points.

Stockfish gave a pawn early in game 21, then the engines reduced to a RB vs RB position on move 21. Leela created a queen side passer and Stockfish blocked it with its pieces. The white king could not move forward and the engines shuffled for a long time. Stockfish ended the game with a perpetual check. In game 22 all the pieces were on the board on move 25 except for one white pawn and a pair of bishops. The white queen went forward and captured two pawns, Leela opened the king side with a rook sacrifice and attacked the white king. A long series of exchanges led to a N vs BN position with white two pawns up and low evals. The game continued for 60 more moves with pawn moves and exchanges before it was adjudicated.

Games 23-24 started with a 20-ply book in the Queen's indian, Petrosian variation, with black a pawn up.

In game 23 evals increased slowly after the start, and most minor pieces were exchanged. Leela opened a file on the queen side and doubled its rooks there. Evals peaked at around 1.5, then the pawns stabilized and the engines started to shuffle. After 35 moves a pair of rooks was exchanged and then queens as well. The engines continued to shuffle in a RN vs RB position, evals slowly came down. The rooks were eventually exchanged as well, the game was adjudicated on move 175.

There were less exchanges after the start of game 24, Stockfish's eval stayed around 1. The black pieces were pushed back and Stockfish had more space, but it didn't seem to have a clear plan. After move 30 Leela played a few moves that were inaccurate according to Stockfish, on move 35 Stockfish's eval jumped over 3.

The white queen had a diagonal line of sight to the black king, Stockfish used the check to move the queen forward on the queen side, and a rook followed in the center. The attack on the black king included a temporary rook sacrifice, the attack was over in a QN vs RRB position on move 48. The material was close to equal yet Stockfish's eval was close to 10. Leela's eval was much lower, still hopeful.

The queen danced across the board, giving checks and preventing any counter Leela might try. After 20 moves Stockfish captured a rook and the game was adjudicated. Stockfish wins the game pair, its lead increases to +4.

Stockfish was a pawn up early in game 25, evals were low around 0.5. Stockfish moved its king side rook and decided not to castle its king at all. Evals increased slowly as Leela increased the pressure on the king side. It pushed a pawn to h6 and threatened mate, Stockfish saved its king but its king side rook was trapped in the corner.

Leela traded its bishop for the rook, and the black queen got trapped in the corner. Leela's queen came forward and captured two pawns before Stockfish could untangle itself.

The game continued for a few more moves, but the material advantage gave Leela the win.

As before Leela was a pawn up in game 26. This time Leela chose to castle long despite the partial pawn cover on the queen side, and its king side rook was free to move. Stockfish's eval was around 1 as it slowly moved pieces to the queen side. Leela doubled rooks on the g file and created pressure on the g2 pawn and the white king. Stockfish had to defend carefully, its position seemed very hard to hold, yet its eval went over 2.

Leela saw its attack wasn't going anywhere but it was difficult to retreat, it tried to exchange queens but Stockfish refused. Stockfish opened another file on the queen side and its two rooks trapped the black king, preparing for the final attack.

Stockfish traded a rook for the defending knight, the white queen left the g2 pawn and went on the atttack. Leela was forced to give its queen for a rook to avoid immediate mate, but it couldn't save the game. The white knight joined the attack and Stockfish won more material to win. Two white wins in this game pair, no change in the lead.

Games 27-28 started with black a pawn up. Stockfish's eval started at (negative) 1 in game 27, then came down after the engines exchanged a few minor pieces. After exchanging queens Leela had a bishop pair to compensate for the pawn. Evals were close to 0 after move 20, the engines continued to exchange pieces and Leela regained the pawn. In a king and pawns ending both engines promoted a pawn to a queen, the game ended in perpetual check. In game 28 Stockfish's eval again reached (negative) 1, then came down when the engines started a series of exchanges. By move 33 the game reached a knight ending, with adjudication 10 moves later.

Leela pushed pawns on the queen side in game 29. The black pieces were pushed back, Leela opened a file and one of the pawns reached a6. Evals slowly increased as the engines exchanged pieces and pawns, Stockfish couldn't get rid of the white a pawn. On move 31 the game reached a RB vs RN position.

Stockfish could have taken the a pawn but chose not to do it since that would bring the white rook to the 7th rank. Instead Stockfish exchanged the minor pieces and reduced to a rook ending, its eval jumping as it realized it was losing. The kings moved forward and the pawns stabilized, Stockfish found itself in near zugzwang. Leela pushed a pawn to g6, creating a second entry point to the 7th rank for its rook. Leela's eval went over 2.

The white king moved to the queen side and threatened to penetrate. Stockfish had to move its rook to stop the king, allowing Leela to capture the e pawn. Leela took over the 7th rank, another black pawn was captured and the game was adjudicated.

In game 30 Stockfish went a pawn up tried to attack on the king side but was not strong enough. The engines exchanged pieces, Leela regained the pawn and evals came down to 0. The exchanges continued until the game reached a drawn rook ending. Leela wins a game pair for the first time in the match, the lead is back to +3.

Season 18 superfinal games 11-20

After 20 games Stockfish leads 6-3 with 11 draws. There were two openings with two white wins each, Stockfish's lead is unchanged. Leela's performance is still a hot topic in the chat. My understanding is that a combination of Leela's software update and the way it uses the CPU on the GPU machine is making it 20% slower in the superfinal. It is not clear how much weaker this makes Leela, if at all. I think it is mostly a small sample issue. In any case I prefer to focus on the games.

The engines exchanged most minor pieces early in game 11. Stockfish pushed a pawn to f3 while Leela took over the center, the position opened and Stockfish went a pawn up. The game reached a QR vs QR position with low evals, the engines mainly shuffled or gave checks until the game was adjudicated. In game 12 again most minors were exchanged early, though the position was different. The game reached a QRR vs QRR position on move 28 with most pawns still on the board. The engines shuffled and evals came down, then the position opened and reached a queen ending. Stockfish ended the game with perpetual check.

In game 13 Stockfish chose to have a BNN vs Q imbalance early. Stockfish's eval was negative, it thought it had an advantage. The black king walked to the corner, Leela could take a rook for a bishop for many moves but it wouldn't give up its bishop. By the time the rook moved to safety evals became positive. All the black pawns were on the board and the black heavy pieces were not effective behind the pawn line. The engines opened the pawn wall on the king side, evals increased over 2 as Leela placed two minor pieces that blocked the pawns in the center and further limited the black piece movement.

Stockfish gave a rook for a knight and opened the position. Both engines had advanced passers on the king side. Stockfish's queen was strong enough to keep several white pieces busy blocking the black passer. Eventually Stockfish lost the passer while reducing to a RBB vs RB position, evals were very high.

Leela had control of the dark squares, its DS bishop made sure the black pawns could not advance. The white king came forward, probably planning to attack the c6 pawn. The game continued for 20 moves before it was adjudicated.

Leela played more peacefully in game 14 and evals were close to 0. Leela had a bishop pair advantage and it went a pawn up on move 20, Stockfish controlled the center and its two knights were deep in black territory, one on the 7th rank. On move 27 Stockfish's eval suddenly jumped over 3.5, it seemed to gift a knight and sacrifice a rook but then it captured the queen and a bishop after a discovered check. The game reached a QR vs RRB position on move 33.

Stockfish was sure it was winning, Leela was not convinced. Stockfish protected the back rank to avoid mate, and it slowly pushed the b pawn forward. Leela countered with the d pawn, Stockfish captured the pawn and the engines exchanged a pair of rooks. 

Stockfish had to juggle between giving checks and avoiding a back rank mate. It took about 20 moves to force the rook to leave the 8th rank and then Stockfish was able to promote its pawn and the game was adjudicated. There were two white wins in this opening, but the games were very different and the opening doesn't seem to be biased.

Games 15-16 started with the Budapest defense, white a pawn up. In game 15 the engines opened a file on the king side, exchanged a pair of rooks and the white queen came forward. A series of early exchanges led to a BNN vs BNN position with white a pawn up. Evals were over 1 for a while, Stockfish reduced to a same color bishop ending and moved its remaining pawns out of the bishop's reach. Leela's king could not move forward to attack the pawns and the engines shuffled for a long time. When Leela used its pawns to open a path for its king there were not enough pawns to win. In game 16 Leela avoided opening the king side and it regained the pawn early. Stockfish moved its queen forward, it captured 3 pawns on the queen side and then was traded for two black rooks. Evals came down, Leela opened the king side and threatened the white king. Both engines thought there was nothing to play for and the game was adjudicated on move 40, still with many pieces on the board. Both engines managed to hold in a difficult opening.

Games 17-18 started with a 31-ply book sequence, a popular line in the Sicilian Najdorf variation. White had a pawn on f6 and the g file was open, and the black king was stuck in the center. In game 17 Leela moved a rook to g7, daring Stockfish to exchange it for a bishop so its pawn would advance to g7. Stockfish refused and tried to attack on the queen side, Leela traded BN for R and stopped the attack. Leela created an advanced passer in the center, the black king's situation looked scary.

Stockfish had no choice but to capture the passer, opening the d file. The black king tried to find cover on the queen side, protected only by pieces. Leela used a queen exchange to get its f pawn to the 7th rank in a RR vs RBN position.

Stockfish captured the e pawn at the cost of losing its remaining pawns. The rook pair was very strong on an open board, eventually Leela captured the knight and the game was adjudicated.

In game 18 Stockfish's eval shot up and was already over 7 after playing 4 moves. It sacrificed a knight in the center and pushed its pawn to f7, forcing the black king to move. Leela captured the pawn, it was up a piece but its king was almost defenseless in the center.

Stockfish sacrificed its bishop and cleared the way for its rook to join the attack. The white queen and rook attacked together, chasing the black king and regaining material. When the attack was over only N vs BN remained, Stockfish 3 pawns up.

The white queen side pawns were unstoppable. Stockfish pushed them forward, Leela gave the bishop to stop a promotion and the knight ending was a win for Stockfish. There were two white wins in this opening, the evals suggest it was too biased.

The first capture in game 19 was on move 20, then a series of exchanges opened the position. The engines shuffled for a while and evals were low, Stockfish's eval jumped over 1 when Leela didn't retake a pawn and went on the attack instead. The engines exchanged queens and Stockfish captured a second pawn, its king was threatened by the white rooks and bishops together with a pawn on h6. After a series of exchanges the game reached a RB vs RB position, Leela went a pawn up with a passer. Stockfish's pieces prevented the passer advancing and the engines shuffled, evals came down and eventually Leela just gave the passer up. After a second long shuffle the engines reduced to a drawn 7-man position and the game was adjudicated. In game 20 Stockfish castled long and concentrated on the king side. After a few exchanges evals came down, Stockfish opened a file on the king side and Leela exchanged a pair of rooks. Stockfish doubled rook and queen and threatened the black king, Leela traded RB for Q and reached a QN vs RBN position. The queen side was locked, the engines exchanged knights and the white queen could not win the game on its own. The game was adjudicated a few moves later.

Season 18 superfinal games 1-10

After 10 games Stockfish leads 4-1 with 5 draws. This result is quite unexpected, considering that Leela beat Stockfish last season, not too long ago. The chat is buzzing, with fans of the two engines arguing and trying to figure out what is going on. Various explanations have been suggested, including the engine software updates, Leela's configuration and the upgraded hardware. It could be a random result of the small sample size, likewise for the result of the previous superfinal, but who wants to believe in statistics? Another contributing factor is the openings. As Jeroen explained, the openings for this superfinal are more biased as he preferred more decisive games, at the cost of having more openings with wins for both engines. Perhaps Stockfish is better in such openings? More decisive games means more work for me, but I'm not complaining too hard.

Games 1-2 started with a Sicilian 2-ply opening. Game 1 continued in the Kan Polugaevsky variation. There were very few exchanges and evals stayed low. Stockfish started shuffling before move 20, after a few exchanges on move 29 both engines continued to shuffle and the game was adjudicated. In game 2 the engines played a B50 Sicilian with d6 and d3 played. There were many early exchanges and only QRB vs QRN remained on move 24. Evals were close to 0 and the game was adjudicated before move 40. 

Leela had an eval advantage from the start of game 3, Stockfish seemed to be holding in the first 20 moves. The center was blocked, Leela pushed pawns across the board and kept its king uncastled. The first exchange was on move 20, Stockfish prepared a king side attack and Leela concentrated on the queen side. Stockfish's eval went over 1, on move 27 it moved its queen away from defending the b7 pawn, but it didn't expect Leela to take it. 

Leela captured the pawn and Stockfish's eval immediately jumped as it saw the blunder (so why did it make the move in the first place??). Leela traded Q for RR, then a quick series of exchanges led to a RB vs BN ending, white a pawn up.

Stockfish had a passer that had to be watched but the material was enough for Leela to win.

In game 4 Stockfish pushed pawns on the king side and castled long. It prepared a king side attack, doubling rooks on the g file, and its eval went over 2. Leela refused a bishop sacrifice but it couldn't prevent the king side being opened. Leela moved its pieces to surround the king, that was left without pawns.

Stockfish shifted its rooks to the open e file, and created a passer on the queen side to increase the pressure. It gave a rook for a knight and prepared for the final attack. 

The white rook and queen moved forward, Leela defended its king and exchanged pieces until only QN vs QB were left. Stockfish was only a pawn up, but after Leela ran out of checks Stockfish forced a queen exchange and the b passer was unstoppable. There were two white wins in this opening, score is 1-1.

Stockfish moved its king early in game 5, out of known theory. Evals increased over 1, the engines opened a file on the queen side, Stockfish stabilized its pawns and started to shuffle. Leela couldn't find an entry point, it offered a knight but Stockfish wouldn't open the king side. Evals slowly came down, after exchanging a few pieces Leela started to shuffle as well, the game ended with a 50-move draw.

Game 6 continued along a more popular line, Stockfish's eval was around 1.5. Leela gave a pawn early to activate its pieces and create a king side threat. Stockfish concentrated on the queen side, it opened a file and placed a knight on c5. Leela created a queen side passer and gave a rook for a knight and pawn to remove the white knight. Stockfish's eval was over 2.

Stockfish moved its heavy pieces to the queen side, it blocked the passer with a rook while taking over the open file. Leela played passively and evals continued to increase .On move 41 Stockfish exposed the black king with a rook sacrifice. 

The white queen came forward, it chased the black king while capturing pawns. Stockfish was 3 pawns up when it exchanged queens with only RB vs RNN remaining. Leela couldn't stop the white king side pawns marching and the game was adjudicated. Stockfish wins first game pair and goes into the lead. 

There was a series of early exchanges in game 7, Leela was up a rook for a knight while Stockfish was two pawns up with connected passers on the queen side. Stockfish's eval came down to 0, Leela's king blocked the passers while its pieces couldn't find a way through the black defense. The position opened with more exchanges, Stockfish gave a piece and its queen side passers and Leela was left without pawns in a RB vs N position. Stockfish had two connected pawns, close enough to promotion so they were worth a rook. The game ended in a tablebase draw. 

In game 8 there were more pieces on the board. Stockfish thought Leela's 15th move was a mistake and its eval jumped over 1.5. It immediately gave two minors for 3 pawns and its eval jumped higher while Leela was completely calm. After exchanging queens the center opened, Stockfish captured two more pawns and created a queen side passer. The white rooks controlled the center and the black king was exposed, yet Leela's eval remained close to 0.

Leela realized something was wrong only on move 27. It gave back some of the material but couldn't save its king. Stockfish attacked with its rooks and managed to capture a black rook.

Leela's last hope was its passer, Stockfish traded a rook for a bishop to get rid of it. The rook ending with 3 pawns up was as easy win. Such engine blindness in a superfinal is a surprise to me, I don't think there was such an extreme case in the last several seasons. Stockfish wins the game pair and increases the lead to +2.

Games 9-10 started with a King's Gambit accepted, with black up a knight for a pawn. In game 9 there were a few early minor piece exchanges. Stockfish was late in developing its king side and its king was uncastled in the center. Leela moved its heavy pieces to the half open f file, Stockfish prepared an escape route for its king to the queen side. Stockfish's (negative) eval increased steadily and was over 2.5 on move 26.

Leela moved its rooks forward, Stockfish captured pawns while Leela shifted its pieces to threaten the king on the queen side. In a series of exchanges Leela regained the piece and reduced to a queen ending with black a pawn up. Despite the small material advantage Stockfish's eval was close to 20 by then, Leela was more hopeful.

The ending was far from trivial and the game continued for more than 25 moves. After a few pawn exchanges Stockfish pushed its queen side pawns forward. The black king had more cover and the white king was exposed to checks, Stockfish used this to get a pawn to the 2nd rank supported by its king. When the second pawn moved forward the game was adjudicated.

In game 10 both engines castled long early and evals came down. The black queen moved forward, Leela traded it for a rook and bishop to get a Q vs RBN imbalance on move 17. The white queen was very active, Leela's pieces defended each other and did not try to attack. Stockfish opened the queen side and threatened the black king, the engines mostly shuffled and the game was adjudicated when the pawns stopped moving. Stockfish wins another game pair and the lead is +3.

Season 18 premier division statistics

A summary statistics table of previous stages and seasons.  

Draw rate, wins

Final draw rate was 79.2%. 

Game termination

The three most common game termination causes were:

51.8% - TCEC draw rule
19.6% - TCEC win rule
17.3% - SyzygyTB

There were no crashes in the stage.

Moves per game

Median= 57.5
Average= 68.7

There were 23 games longer than 100 moves, the longest was 304 moves (rofChade - Leela, game 141, draw). 

Time per game (hours)

Median= 2:57
Average= 2:54

Openings

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

The engines had almost no freedom to choose the opening variant, all except two of the game pairs repeated the same ECO code and the same opening variant twice. 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, 32.1% diverged immediately out of book, 67.9% of the pairs diverged at most after 1 move. The longest repeated sequence of moves was 16 plys (Stoofvlees - Leela games 75 and 103, Alekhine's defence, modern, Larsen variation, two draws).

Season 18 premier division, final results

Final standings

Stockfish and Leela will play in the season 18 superfinal. Stockfish took the lead in RR3 and never looked back. Leela kept winning in a steady pace to end up in 2nd place. It finally won a game against a top 4 engine, though it was a result of a Stoofvlees blunder. AllieStein stayed in the race until the last two rounds. It had a potential tiebreak advantage over Leela but it couldn't find wins when it needed them, instead it lost to Komodo and couldn't recover. Stoofvlees made a late comeback in the last RR, after losing twice in RR5. Its +4 score after RR2 seems to be a statistical outlier. 

Fire and rofChade are relegated back to league 1. There were 4 leagues this season, and 7 of the 8 engines that promoted at the end of a league were then relegated back in the next league. Only Booot managed to survive its promotion to league 1.

Interesting games

Game 114, Stockfish - Fire: Fire didn't castle and moved its king early, Stockfish's eval increased over 1 as it gathered forces for a king side attack. Fire protected its king and Stockfish shifted to the center and created a passer. It gave a rook for a knight and used a mate threat to reduce to a QRN vs QRR position, with two pawns up. Fire had to block the passer with a rook, this was enough for a win though it took Stocfish a while. 

Game 115, AllieStein - Ethereal: AllieStein pushed a pawn to h6 and attacked the black king, while its king remained uncastled. Ethereal exchanged queens and captured two pawns, yet evals increased quickly. The black king was trapped in the corner, Ethereal relieved the pressure by exchanging pieces until only R vs B remained. The white pawns were out of the bishop's reach, AllieStein captured pawns until getting a tablebase win.

Game 117, rofChade - Stoofvlees: Stoofvlees made a few moves in the opening that were not optimal, rofChade went a pawn up and evals were around 1. Stoofvlees pushed pawns on the king side, it then gave a second pawn and attempted to attack the white king. rofChade gave the pawns back and was confident in its defense despite the two black rooks on the 2nd rank. Then Stoofvlees blundered badly, it went ahead with the attack and missed the threat against its king. It saw the mate coming 4 moves later, too late. rofChade gets its first win in the division, a real blow for Stoofvlees' superfinal chances.

Game 121, Stockfish - rofChade: Stockfish had an eval of 1 from the start. rofChade had problems developing the queen side while Stockfish attacked on the king side, evals increased quickly. Stockfish gave a rook for a bishop and pawn, creating a pawn majority on the queen side and a threat to the black king on the long diagonal. After exchanging queens Stockfish created passers on both sides and rofChade's defense collapsed. Stockfish started to win material and the game was adjudicated.

Game 123, Komodo - Fire: Komodo pushed pawns on the king side and didn't castle its king. The g file opened and one white pawn reached h6, Komodo's eval was over 1. Fire had no real counter, evals slowly increased as Komodo moved both its rooks and its queen to the g file. Fire had to give material to save its king and the game was adjudicated.

Game 129, AllieStein - rofChade: rofChade was up a rook for a bishop exchange early in the game, though AllieStein had a small eval advantage. AllieStein had a strong bishop pair that limited the black pieces. AllieStein castled long, when it was ready it opened the king side with its pawns. For a while the engines seemed to be shuffling, though AllieStein's eval continued to increase. Eventually rofChade gave back the rook to get rid of one of the white bishops. AllieStein pushed a passer to the 7th rank and rofChade gave a bishop to stop it. AllieStein had enough pawns to win in a RB vs R ending.

Game 131, Stoofvlees - Leela: Stoofvlees had an eval advantage in a closed position, Leela defended well. The position opened and the engines reduced to a RN vs RB position with low evals. Suddenly Leela's eval jumped over (negative) 3, Stoofvlees blundered and Leela captured the knight a few moves later. Stoofvlees was 3 pawns up but its pawns were isolated and it couldn't defend them all. Leela kept one pawn safe, it slowly captured black pawns until reaching a tablebase win.

Game 134, Leela - Ethereal: The engines castled in opposite directions, two pawns were exchanged on the queen side and Leela pushed a pawn to h6 on the king side. Evals increased slowly while the engines exchanged pieces and pawns. Leela pushed pawns on the king side, after a series of exchanges only QRN vs QRB remained and Leela had an advanced passer in the center. The engines exchanged queens and the white king moved forward to support the passer. Ethereal gave a rook for a knight to get rid of the passer and the R vs B ending was a win for Leela.

Game 136, AllieStein - Komodo: There were many exchanges after the start and the game reached a RRB vs RRB position on move 22, with AllieStein a pawn up. AllieStein had a 3 to 1 pawn majority on the queen side, after exchanging a pair of rooks its eval jumped quickly to double digits. AllieStein knew it was winning but getting there took 40 moves. First it exchanged pawns on the queen side and created connected passers. Komodo concentrated on blocking the passers, AllieStein captured two pawns on the king side and then reduced to a winning rook ending.

Scores after 35 rounds: Stockfish +9, Leela +7, AllieStein +6, Stoofvlees +2, Komodo -1, Ethereal -4, rofChade -9, Fire -10. Stockfish and Leela continue to lead with two wins each. AllieStein had its best RR so far with 3 wins, it is still in the race. Stoofvlees is practically out after losing twice, including a huge upset by rofChade. No news at the bottom other than now Fire is last. The only question is which engine will beat Fire and rofChade, and how this will affect the race at the top of the table. So far the reverse games have been less exciting, we'll see how the last RR goes.

Game 141, rofChade - Leela: The engines shuffled from move 28, with pawn moves that reset the 50-move counter. On move 158 the engines exchanged a pair of rooks, on move 221 the queens were exchanged and only RB vs RB remained. Leela may have missed a win, rofChade's eval was over (negaitve) 2 for a while. However evals came down as the game continued, adjudication finally came on move 304. 

Game 145, Stoofvlees - rofChade: The engines opened the king side, both kings were exposed and the black king was especially vulnerable on the long diagonal. Stoofvlees gave a rook for a bishop and created an advanced passer in the center. Evals increased quickly while the engines exchanged pieces until only BB vs RN remained. Stoofvlees went 3 pawns up, it promoted a pawn before the game was over.

Scores after 38 rounds: Stockfish +9, Leela +7, AllieStein +6, Stoofvlees +3, Komodo -1, Ethereal -4, rofChade Fire -10. Only one decisive game in 3 rounds. AllieStein couldn't get anyting playing white against Fire. If it wins a game it will have a tiebreak advantage over Leela, so the superfinal race is still open.

Game 154, Stoofvlees - Fire: The black queen moved forward, Stoofvlees used its absence to place a pawn on f6 before exchanging queens. Evals increased as the engines exchanged pieces, they were over 3 when only BN vs BN remained. Material was equal, Stoofvlees moved its king forward and Fire couldn't stop it and protect all its pawns. Stoofvlees captured the white king side pawns and created passers, enough for a win.

Game 159, Leela - Stoofvlees: The game reached a BB vs R ending on move 31, evals over 1.5. Stoofvlees had a passer on the 2nd rank, it took Leela a few moves to get rid of it. Analysis engines were sure Leela was winning, and that move 43 was a mistake. Leela did not find a way to win, the engines shuffled with a few pawn moves in between. The game was adjudicated on move 143.

AllieStein only managed a draw playing black against rofChade, while Leela missed a win against Stoofvlees. No change at the top, only two more rounds to play.

Game 164, Komodo - AllieStein: See featured game below.

At precisely the wrong moment AllieStein lost a game. One round to go there is still a theoretical possibility that AllieStein will finish ahead of Leela, AllieStein has to win in black against Stoofvlees and Leela has to lose in white to Fire.

AllieStein only managed a draw against Stoofvlees, and Leela qualified for the superfinal before its last game. Leela was close to a win against Fire but the game ended in a draw. In any case AllieStein wouldn't have qualified even if it had won its last game. 

Featured game: Komodo - AllieStein
Premier division, game 164
Link to game on TCEC

Komodo had an eval advantage of around 1 from the opening. AllieStein tried to attack on the king side, its queen moved forward and captured a pawn, while its king remained uncastled in the center. Komodo opened a file in the center, its pieces were ready to attack the black king. AllieStein's queen came back quickly, evals were over 2 already on move 19.

Komodo targeted the pinned black knight, AllieStein couldn't castle because the king was a crucial defender. The black pieces surrounded their king, with very little room to move. AllieStein's eval was over 10 on move 25.

Komodo attacked through the center, it gave a rook for a bishop and exposed the black king. Komodo was behind in material but the black king was in front of its defenders, facing 4 white pieces and with nowhere to hide.

The game was adjudicated, both PVs showed how AllieStein loses material within a few moves. There were two white wins in this opening, though the games were very different. I don't think anyone expected Komodo to beat AllieStein this fast, if at all.

Season 18 premier division, after RR4

Standings after RR4

Stockfish leads the table, Leela is a point behind while Stoofvlees and AllieStein are still in the race. Stockfish had an outstanding RR3 with a +5 score, including wins against Stoofvlees and AllieStein. This was enough for the lead though it drew all the games in RR4. Leela continues to accumulate wins at a slower pace. Stoofvlees led after RR2 but had only one win since then to compensate for its loss and it dropped to 3rd.

At the bottom Fire and rofChade are on their way back to league 1, it is mostly a question which will finish last. Ethereal and Komodo are stable in the middle of the table, Ethereal is still without a win in the division.

There were 10 decisive games in RR3, of which 5 were Stockfish wins. There were a few more games with winning chances, at least one missed win by AllieStein. RR4 was less interesting with only 3 decisive games, all Fire losses.

Interesting games

Game 57, Leela - rofChade: rofChade was a pawn up with a queen side passer on the 2nd rank that Leela blocked with a rook. Leela exposed the black king and regained the pawn, evals stayed low and the engines shuffled for a while. Leela thought rofChade's move 50 was a blunder, within a few moves Leela captured the black passer and the black defense collapsed. The engines exchanged pieces and reduced to a RBN vs RBN position, Leela was 3 pawns up when the game was adjudicated.

Game 58, Stockfish - Fire: Stockfish had an eval advantage close to 1 after the start, Fire seemed to be holding as the engines slowly exchanged pieces. On move 35 the game reached a QRN vs QRB position and Stockfish's eval jumped over 2. The queens were exchanged and the engines started to capture pawns, Stockfish's eval climbing fast. On move 55 Stockfish went a pawn up, there were only a few pawns left and Stockfish seemed to have an unstoppable passer. Fire gave its last pawn in desperation and the game was adjudicated.

Game 59, AllieStein - Ethereal: AllieStein had a promising QRB vs QRB position with a passer on the 7th rank. Ethereal found a way to exchange pieces and get rid of the passer, the game reached a B vs pawns ending. Ethereal was 2 pawns up, all its pawns out of the bishop's reach. AllieStein wouldn't lower its eval, the game ended in a 50-move draw.

Game 61, rofChade - Stoofvlees: Stoofvlees sacrificed a bishop for pawns, but its attack on the white king wasn't strong enough. The engines reduced to a NN vs R position and the game ended in a tablebase draw.

Game 63, Fire - AllieStein: AllieStein missed a win in this game. Fire was two pawns up in a B vs BN position, with advanced connected passers on the 6th and 7th ranks. AllieStein probably made a wrong choice on move 47, its knight had to protect a pawn and its king blocked the passers, while the black bishop's targets were protected by the white king. The engines shuffled, AllieStein prolonged the game with pawn moves until adjudication on move 171.

Game 65, Stockfish - rofChade: Stockfish had an eval advantage from the start. It pushed the king side pawns forward and didn't castle its king. rofChade moved its queen forward on the queen side and captured a pawn, Stockfish immediately attacked while the black queen was away. It gave a rook for a knight, opened the king side with its pawns and took advantage of the fact that the black queen side pieces were undeveloped. Stockfish used the king side attack to regain the material, and created connected passers. Together with a mate threat this was enough for a win. 

Game 67, Komodo - Fire: Fire was up a pawn early but Komodo pushed a pawn to h6 and trapped a black bishop in the corner. The engines exchanged pieces gradually until only RBB vs RBB were left on move 34. Komodo was effectively a bishop up, the game was adjudicated after a few moves. This is Komodo's first win in the division. 

Game 72, Stockfish - AllieStein: See featured game below. First decisive game in the top 4, first loss for AllieStein.

Game 73, AllieStein - rofChade: AllieStein had a strong central pawn and a space advantage. The pawn became a passer, and eventually rofChade gave a knight to get rid of it. The game reached a RBN vs RN position on move 54, evals were high but AllieStein played 30 more moves of slow progress before the game was adjudicated.

Game 74, Komodo - Stockfish: Stockfish had an eval advantage under 1 (negative) from the start, Komodo had two isolated pawns on the queen side. The engines mainly shuffled and evals stayed constant until move 40. A few piece exchanges later Komodo managed to connect its pawns while Stockfish had a bishop pair and a queen side passer. Stockfish's eval jumped to around 3 on move 60, it used its passer to reduce to a R vs BB ending. The bishop pair was stronger than the rook, the game was adjudicated before Stockfish started to capture pawns.

Game 78, Leela - Ethereal: All the pieces were on the board until move 29, evals increased slowly. Leela had two strong knights in the center, Ethereal gave a rook to take one out. The engines exchanged pieces until only RR vs RBN remained, white two pawns up. Ethereal created a passer but it wasn't strong enough, Leela moved the rooks to the back rank and managed to capture the black knight. A few moves later Leela went a pawn up and the game was adjudicated. 

Game 79, Stockfish - Stoofvlees: Stoofvlees castled long, Stockfish's eval jumped over 1 after it opened the queen side. The white queen and knight chased the black king to the center, then Stockfish's eval jumped again when Stoofvlees gave a rook for a knight. The black king moved forward as Stockfish closed in. Stoofvlees had a 21 minute long think on move 30, the first time it realized it was in trouble. However, Stoofvlees' eval didn't increase fast enough, the white rooks got involved and the game ended with mate on the board. First loss for Stoofvlees in the division.

Game 81, Komodo - rofChade: Komodo had an eval advantage from the start, it pushed the black pieces back and evals increased to around 2. rofChade gave a rook for a knight and pawn, the game reached a RB vs BN ending that was a win for white, though the game continued for 20 more moves.

Scores after RR3: Stockfish +7, Leela +4, Stoofvlees +3, AllieStein +2, Komodo -1, Ethereal -2, Fire -5,  rofChade -8. Stockfish takes the lead after a fantastic RR with 5 wins, including wins against Stoofvlees and AllieStein. Leela took advantage of the losses of its opponents to move up to 2nd place, after a solid +2 in the RR. Komodo had a positve score after it beat Fire and rofChade, it moves away from the relegation line. Fire and rofChade continue to lose games, probably both will relegate. The reverse games are next.

 

Game 91, AllieStein - Fire: AllieStein tried to sacrifice a bishop for pawns and to expose the black king, Fire refused to take back and accepted the pawn loss. AllieStein's eval jumped over 2, and it continued to increase quickly as the engines exchanged pieces. The game reached a knight ending, Fire continued to play another 30 moves before its eval was high enough for the win rule.

Game 98, Stoofvlees - Fire: Stoofvlees pushed pawns on the king side and its eval increased over 1. Fire tried to use open files on both sides but had no real threats. Stoofvlees placed a protected knight on f6, its eval increasing steadily. By move 39 Stoofvlees was a pawn up in a BNN vs BNN position, 10 moves later it was two pawns up in a BN vs NN ending. The game was adjudicated after Stoofvlees went 4 pawns up.

Game 112, Leela - Fire: The engines castled in opposite directions, evals started to increase when Leela pushed pawns on the king side facing the black king. Fire hid its king behind a white pawn and tried to counter on the queen side, but its attack was too weak. Fire gave a rook for a bishop and the game reached a RR vs RB ending. Leela was not in a hurry, the black bishop did not have pawn targets. There were several 2-fold repetitions, eventually Leela found the winning path by capturing pawns and pushing a passer.

Featured game: Stockfish - AllieStein
Premier division, game 72
Link to game on TCEC

Stockfish started with an eval advantage under 1, evals stayed relatively low after the opening. The f file was half open early, nevertheless AllieStein decided to castle short. Stockfish doubled its rooks on the f file and on move 22 it sacrificed both its rooks for BN and exposed the black king. Evals increased to around 1.

Stockfish attacked on the king side, AllieStein exchanged pieces and gave some of the material back to save its king. The game reached a QB vs QR position with white 3 pawns up, evals did not change a lot and it was still possible that AllieStein could hold. 

Stockfish's eval fluctuated wildly for a few moves, the black king was in the corner with the rook pinned. After a few moves with low evals Stockfish found an advantage again, AllieStein captured 2 pawns and chased the white king around while Stockfish's eval increased over 4. When the checks ran out AllieStein was in zugzwang, 

AllieStein started pushing pawns and Stockfish captured them. AllieStein finally realized it was losing when it was down to its last pawn. Stockfish pushed a passer to the 7th rank and the game was adjudicated. 

Season 18 premier division, after RR2

Standings after RR2

Stoofvlees is leading with a +4 score, followed by AllieStein, Leela and Stockfish with +2 each. Stoofvlees is the surprise of the division so far, I don't think anyone expected it to perform so well. The leading group is unbeaten so far, there is a clear divide between engines with positive scores and those with negative scores.

Fire and rofChade, the promoting engines, are at the bottom. I think rofChade will relegate, Fire still has a chance to survive. Komodo is dangerously close to relegation, which I didn't expect.

The division started with 5 straight decisive games, but the draw rate has increased significantly since then to around 80%.

There are three NN engines in the leading group, it is quite possible that Stockfish will not reach this season's superfinal. The last time TCEC had a superfinal without Stockfish was in season 10 (Houdini - Komodo), this was the only season this happened since season 4.

Interesting games
Game 1, Leela - rofChade: Leela gave two pawns and developed a strong king side attack. rofChade exchanged pieces and reduced to a QRN vs QRN position, its king behind a white pawn. In desperation rofChade gave a rook and exposed the white king to checks. Leela moved its king to safety and reducd to a RN vs N ending, 4 pawns down but still winning.
Game 2, Stockfish - Fire: The game reached a same color bishop ending., Stockfish a pawn up with a passer. The engines shuffled for a few moves with evals constant. Then Stockfish saw a win, though Fire didn't think it blundered. It took Fire a few moves to see what happened, Stockfish found a way to push the black bishop away and moved its king forward. The game was adjudicated before we saw the win played. 
Game 3, AllieStein - Ethereal: AllieStein had an eval advantage from the start, the black king stayed in the center uncastled and evals increased steadily. The black king side rook was effectively trapped, Ethereal traded it for a knight to slow AllieStein on the king side. The game reached a RR vs RN position, AllieStein's eval was winning though the game continued for another 20 moves.
Game 4, Komodo - Stoofvlees: Stoofvlees went a pawn up and created a pawn majority on the king side. The queen side opened and Stoofvlees' (negative) eval started to increase. Komodo did not defend accurately, Stoofvlees reduced to a QBN vs QBN position and targeted the white backward e pawn. The white king and knight couldn't reach the pawn in time, Stoofvlees exchanged queens and captured it. The two pawns advantage was enough for a win.

Game 5, rofChade - Stoofvlees: rofChade went a pawn up and opened a file on the queen side, Stoofvlees' eval became negative. Stoofvlees threatened the white king on a long diagonal, forcing the white pieces back to defend. The black pieces moved forward on the queen side, Stoofvlees regained the pawn and reduced to a RN vs RB position. Both engines had passers but Stoofvlees' pieces were stronger. It managed to capture all the white pawns, 4 pawns up when the game was adjudicated.

Game 9, Stockfish - rofChade: The game reached a RBN vs Q imbalance, black two pawns up. After some shuffling, and perhaps inaccuracies from rofChade, Stockfish found a combination to capture two pawns and exchange a pair of rooks. rofChade then traded a promoting passer with a bishop and only RN vs Q remained, white two pawns up. rofChade thought it could hold until it realized Stockfish had an unstoppable passer. 

Game 15, Leela - Komodo: The engines castled on opposite sides and the center was blocked. All pawns were on the board and the engines mostly played behind their pawn lines. On move 42 the first pawns were exchanged, Komodo opened a file on the queen side and then sacrificed a knight for pawns to attack the white king. Leela exchanged queens and found safety for its king, the game reached a RRB vs RR position with black 3 pawns up. Leela made slow progress, it made sure to keep one pawn safe and won the RB  vs R ending. 

Game 22, Leela - Ethereal: The game reached a BBN vs Q imbalance early. Leela thought it had an advantage, its eval got close to 1.5. However, after exchanging all rooks Ethereal defended well and evals slowly came down. The engines reduced to a 7-man drawn position and Leela kept playing until move 176. 

Game 25, Komodo - rofChade: Komodo had a strong king side attack. It forced the black king to run to the center, evals were around 2. Most pieces were exchanged and only Q vs RR remained, Komodo couldn't break rofChade's defense and the game ended in a draw after 149 moves. 

Scores after RR1: Stockfish Leela Stoofvlees +2, AllieStein +1, Fire Ethereal -1, Komodo -2,  rofChade -3. The table is split into positive and negative scoring engines. At the top Stoofvlees is in the lead together with Leela and Stockfish, after winning two games playing black. At the bottom rofChade is last, which was expected. Komodo is weaker than I assumed, it used to be be better that Fire and Ethereal. Still many games to be played, the reverse games may change everything.

Game 33, Stoofvlees - rofChade: Stoofvlees' eval increased quickly from the start. It pushed the black pieces back with its pawns, then gave a pawn and attacked on the king side. rofChade gave back the pawn and gave a bishop to force a queen exchange. The black king was out of immediate danger, but Stoofvlees created an advanced passer, both engines agreed that together with the extra piece Stoofvlees was winning.

Game 35, AllieStein - Fire: AllieStein sacrificed a knight for pawns and exposed the black king. Fire kept its king safe, while AllieStein went 3 pawns up with two advanced passers. A series of exchanges reduced to a RB vs RB postion, AllieStein had a protected passer on the 7th rank that kept the black pieces busy, it captured the remaining black pawns and the game was adjudicated. 

After round 9 Stoofvlees moves into the lead, and AllieStein joins the chasing group together with Leela and Stockfish.

Game 42, Stoofvlees - Fire: See featured game below. 

Game 44, AllieStein - Stockfish: The engines opened two files, their remaining pawns locked in three pawn islands. All pieces were on the board until move 44, AllieStein had more space and its eval was around 2. Stockish's eval went over 2 for a short while, then came back down after the queens were exchanged. Stockfish covered all entry points, AllieStein couldn't break the fortress. The game ended in a 50 move draw. 

After round 11 Stoofvlees increases its lead to a full point, with a good chance of being first at the end of the first double RR. 

Featured game: Stoofvlees - Fire
Premier division, game 42 
Link to game on TCEC

Fire went two pawns up from the start, also exposing the white king. Stoofvlees thought it had a small advantage, it had a bishop pair and the black king couldn't castle. Stoofvlees pushed the black pieces back, gave two more pawns and captured a knight.

Fire had all its pawns on the board until move 41. The engines mainly shuffled behind the pawn lines and exchanged pieces. Evals stayed low, the game reached a RB vs R ending on move 47. 

Fire was 3 pawns up and evals were drawish. The white king came forward while the bishop kept watch over the black queen side pawns. AllieStein's eval increased above 1, Fire was close to zugzwang. All the black pawns could not move safely, the black king was trapped in the corner, only the black rook moved and it had no safe square behind its pawns. I'm not sure what was Stoofvlees' threat, but somehow Fire felt it had to move its pawns on move 68. Fire lost control of the position very quickly and its eval jumped over 5.

Stoofvlees was careful to keep its a pawn safe, the bishop made sure the black passers couldn't advance. The white king and rook had more freedom to operate, it took a while but Stoofvlees captured the black passers and the game was adjudicated in a 7-man winning position.