Crosstable after RR2
Komodo is in the lead, one point ahead of Houdini and Stockfish. So far there have been only draws within the top 3 engines and their score is determined by their performance against the lower ranked opponents. Komodo is doing well, an improvement compared to last season. Houdini started with 6 draws, but came back in the reverse games. All three are looking good and anything can happen.
Chiron crashed 3 times in its first 5 games and was disqualified with automatic relegation. Jonny is the at the bottom of the table and is most likely to be the second relegated engine, it has no wins so far and Ginkgo is one point away in 6th.
A few interesting games
Stockfish - Jonny: Jonny held on until move 22 when it was tempted to take a pawn. Stockfish responded with a deadly king side attack while most of the black pieces were on the queen side. The game was over on move 31, just before Jonny started to lose material.
Jonny - Chiron: The game reached a R vs NN imbalance on move 15, Jonny two pawns up with an eval over 1. The engines started to shuffle in a closed position, Jonny avoided the 50-move draw and then evals became negative after move 100. The position opened a little and Chiron took over, the white king was in danger and Chiron used this to win a rook for a bishop and create a queen side passer. This led to a winning QR vs QRB position. Second loss for Jonny.
Fire - Ginkgo: Fire pushed its king side pawns leaving its king unprotected and the queen side undeveloped. Evals gradually increased, Ginkgo's king was trapped in the corner with a white pawn on f6.The position opened up and Fire won a pawn, and then a mate threat caused Ginkgo's defense to collapse, a win for Fire.
Komodo - Jonny: Komodo's eval started to increase on move 21 and was over 1 after a few moves. It opened the king side and Jonny had to be careful to defend its king. After a series of exchanges Komodo reduced to an opposite color bishops ending two pawns up, good for a win. Three losses in a row for Jonny, looks like it is going to be the punching bag of the division.
Houdini - Chiron: Chiron crashed in a losing position, this is Chiron's second crash in 4 games and it will probably be disqualified soon.
Komodo - Andscacs: Andscacs grabbed a pawn with its queen, evals were close to 0 until move 25. The black king was exposed and moved without castling, Komodo's eval jumped over 2 a few moves later. It attacked through the open diagonals, the black king was in danger even after Komodo sacrificed a knight. Andscacs tried to counter, giving back the material and pushing pawns against the white king, but it was hopeless. The pressure was too great, the game was adjudicated in a QRR vs QRN position with a white pawn on the 7th rank.
Andscacs - Jonny: Andscacs had an advantage from the opening. On move 30 Andscacs had a rook for a knight and a queen side passer with eval over 1.5. On move 43 the game reached a RR vs RN position, with Jonny up a pawn and evals above 2. It took more than 30 moves for Andscacs to find the path to a win, could Jonny have held? Now 4 losses to Jonny in 5 games.
Ginkgo - Komodo: Evals were close to 0 after 40 moves, then Komodo started to take over. Evals jumped over 1 after Komodo created a passer on the king side. Ginkgo gave a pawn and tried to counter with an attack on the black king. Komodo's eval continued to increase, it reduced to a winning BN vs RN position with passers on both sides. Komodo's 3rd win in the first RR, it is leading the table.
Chiron - Stockfish: Chiron crashed for the third time, and it is disqualified in the first RR.
Stockfish - Fire: See featured game below. First loss for Fire, a qualilty win for
Stockfish in 2nd place with a +2 score.
Andscacs - Ginkgo: Evals were close to 0 until move 32. Ginkgo had a pawn majority on the queen side, it started to push pawns and its eval started to climb. When it got a passer to the 3rd rank Ginkgo's eval was over 2. Andscacs gave a pawn and got nothing in return, when Ginkgo pushed a second passer to the 3rd rank the game was over. A small surprise, a win in black for Ginkgo, recall it promoted from division 1.
Ginkgo - Jonny: Evals remained close to 0 until move 42. After all the king side pawns were gone Ginkgo's king walked to the queen side for a little more safety. Jonny's king had no pawn support and the evals increased gradually. In a series of exchanges Ginkgo gave two rooks for a queen and the game reached a QN vs RRB position, evals around 1.5. For a while Jonny managed to hold but Ginkgo's queen was able to capture the two remaining black pawns and Jonny had no counter. Ginkgo's pawns started to advance and the game was soon over. Back to back wins for Ginkgo, it now has a 0 score.
Andscacs - Houdini: Material was equal, Andscacs controlled the center and had a weak bishop, Houdini had a pawn majority on the queen side and had a strong central knight. Evals remained close to 0 until move 48, then Houdini created a passer on the queen side and sent its queen forward. A mate threat caused Andscacs' defense to collapse, the game was adjudicated with Houdini up a rook for a bishop with connected passers. This is the first Houdini win of the division.
Houdini - Ginkgo: Houdini had a small eval advantage from the start, and although material remained equal the eval gradually increased. The black king was in the corner and the black pieces could hardly move, while Houdini slowly gathered its pieces for the attack. The king side opened and Houdini added two pawns to the attack, when the game was adjudicated Houdini saw mate in its PV, material still equal. Houdini's 2nd win, the usual top-3 is forming again.
Ginkgo - Stockfish: Stockfish placed a pawn-bishop formation deep in white territory on move 20, and evals started to increase for black. More black pieces joined the attack on the king side, Ginkgo exchanged pieces and gave a rook for the annoying bishop. The evals jumped higher, the white king was exposed and Stockfish's rooks had open files. Stockfish sacrificed a rook, exchanged queens and used the advanced pawn to get the rook back. The game reached a BN vs RN position, Ginkgo couldn't protect its pawns and Stockfish was soon winning. Stockfish joins Komodo in the lead with a +3 score.
Komodo - Ginkgo: The evals were close to 0 for 50 moves, then the small eval advantage Komodo had started to increase. It had a central passer, and after a few exchanges Komodo created a second passer, connected with the first. Ginkgo was a pawn up and it reduced to a double rook ending, the white central passers were too strong when they started to march. Komodo completes a double win against Ginkgo, regains the lead with +4.
Houdini - Jonny: Houdini had an eval advantage from the start, it gave a pawn to create an advanced passer on move 21. Houdini gave a second pawn and then exchanged its passer for a knight, evals over 1. The game reached a QBN vs QB position, Jonny had a passer on the 2nd rank but Houdini controlled the queening square. The white knight was very active, threatening the king and targeting Jonny's pawns. Houdini captured a pawn and exchanged queens, leading to a won BN vs B position. Houdini with a +3 score, joint second place with Stockfish.
Komodo - Fire: Komodo had an eval advantage from the start that increased gradually, it had a strong central knight and a space advantage. The engines shuffled behind their pawn lines, then after a series of exchanges Fire got rid of the white knight, it had a passer on the queen side while Komodo had a central passer. Evals started to jump as Komodo exchanged a pair of rooks, king safety became important as the black king was exposed while the white king had pawn support. Fire exchanged down to a rook ending, Komodo's rook controlled the 7th rank and could capture pawns easily. Komodo was 3 pawns up after a few moves and the game was adjudicated. Komodo extending its lead with a +5 score, Fire's second loss.
Featured game: Stockfish - Fire
Premier division, round 6
Link to game on TCEC
Stockfish had an eval advantage from the start. The engines exchanged the DSBs early and Fire's king position had dark squares weakness, it tried to compensate with its queen and knights.
Stockfish pushed pawns on the king side, with evals increasing. After exchanging the remaining bishops Stockfish's eval was over 2, the pawn line on the king side looked dangerous as the pawns were ready to charge.
Stockfish opened the g file and its eval jumped over 3. The white knights took over the center and Stockfish increased the pressure on the black king. Fire gave a rook to remove one of the white knights, after exchanging a pair of rooks the game reached a QRN vs QNN position.
Fire's king was still under threat, and Stockfish pushed its central pawns as well. The game was adjudicated before
Fire lost more material.
Tuesday, May 29, 2018
Saturday, May 26, 2018
Season 12 live blogging, Saturday May 26th
Houdini and Stockfish are playing, the first game they meet in this division, in the season even. They will play 8 times in the division, and 100 games in the superfinal if they both qualify.
This is the last game of the first RR. So far Chiron crashed 3 times and has been disqualified, all its past games ignored. In the new tournament settings all its future games have been canceled, so the season will be shorter than planned by 25%.
Komodo is leading with a +3 score, Stockfish with +1 is the only other engine with an above average score. Houdini has drawn all its games so far, but most of these have been with black (its white game against Chiron ended in a crash and was removed, it is now playing as white against Stockfish). Fire has one win and one loss, Ginkgo has two each. Andscacs has two losses and a win, while Jonny has 4 losses and will be relegated unless another engine starts to crash.
The games are 3 hours long or more, I will not be able to blog continuously through the whole game. We'll see how it goes.
move 14 - Stockfish played a rare move on move 9, right after book. Then on move 11 its king moved forward, no longer allowed to castle. The black rooks are connected and both are on the king side to counter a potential pawn attack by Houdini. Perhaps the black king will find safety on the queen side later.
--- had to go for an hour ---
move 34 - Stockfish managed to hold so far, evals remain close to 0. After exchanging all bishops Houdini attacked with its queen on the queen side, forcing the black king back to the center. Houdini doubled rooks on the open e file and Stockfish exchanged RR for Q. Houdini then won a rook for a knight and the game reached a Q vs RN position. The evals are low and Stockfish will hold, any result other than a draw will be a surprise.
move 39 - Evals are not moving, despite Houdini thinking for 10 minutes on move 36. This may take a while.
move 48 - Stockfish's eval has dropped to 0. The black rook now controlling the 4th rank, not allowing the white king to move forward.
move 55 - Houdini is still hopeful with eval over 0.2. I doubt it will blunder in its optimism, it just takes longer to realize it cannot win.
Houdini lowered its eval on move 64 and the draw rule was activated on move 70. Houdini completes the first round with 6 draws and in joint 3rd with Fire and Ginkgo. Komodo in the lead with a +3 score. Will this season be Komodo's comeback?
This is the last game of the first RR. So far Chiron crashed 3 times and has been disqualified, all its past games ignored. In the new tournament settings all its future games have been canceled, so the season will be shorter than planned by 25%.
Komodo is leading with a +3 score, Stockfish with +1 is the only other engine with an above average score. Houdini has drawn all its games so far, but most of these have been with black (its white game against Chiron ended in a crash and was removed, it is now playing as white against Stockfish). Fire has one win and one loss, Ginkgo has two each. Andscacs has two losses and a win, while Jonny has 4 losses and will be relegated unless another engine starts to crash.
The games are 3 hours long or more, I will not be able to blog continuously through the whole game. We'll see how it goes.
move 14 - Stockfish played a rare move on move 9, right after book. Then on move 11 its king moved forward, no longer allowed to castle. The black rooks are connected and both are on the king side to counter a potential pawn attack by Houdini. Perhaps the black king will find safety on the queen side later.
--- had to go for an hour ---
move 34 - Stockfish managed to hold so far, evals remain close to 0. After exchanging all bishops Houdini attacked with its queen on the queen side, forcing the black king back to the center. Houdini doubled rooks on the open e file and Stockfish exchanged RR for Q. Houdini then won a rook for a knight and the game reached a Q vs RN position. The evals are low and Stockfish will hold, any result other than a draw will be a surprise.
move 39 - Evals are not moving, despite Houdini thinking for 10 minutes on move 36. This may take a while.
move 48 - Stockfish's eval has dropped to 0. The black rook now controlling the 4th rank, not allowing the white king to move forward.
move 55 - Houdini is still hopeful with eval over 0.2. I doubt it will blunder in its optimism, it just takes longer to realize it cannot win.
Houdini lowered its eval on move 64 and the draw rule was activated on move 70. Houdini completes the first round with 6 draws and in joint 3rd with Fire and Ginkgo. Komodo in the lead with a +3 score. Will this season be Komodo's comeback?
Monday, May 21, 2018
Season 12 division 1 final results
Final Crosstable
The final draw rate is 58%.
Ginkgo and Jonny are promoted to the premier division, Gull and Texel are relegated to division 2. The promotion race was very close for most of the division. Fizbo and Booot were dropped only in the last two double rounds, Fizbo with a double loss to Ginkgo and Booot with a double loss to Fizbo.
It is a little sad to see Gull relegated. Only a few seasons ago it was one of the top engines, but it is not being developed and new engines are constanty improving. Check out season 8 stage 3 coverage (part 1, part 2, part 3, part 4, part 5, part 6), this was the stage where Stockfish had a bug causing it to lose on time. Gull led the stage and was close to reaching the superfinal.
There was one crash in the division, Fritz crashed against Jonny. In the end this crash did not have an effect of the final results, Jonny would have promoted without that point, and Fritz was not in danger of relegation as a result of the lost point.
The premier division will start soon. This is the main event of the season, Jonny and Ginkgo will be joined by Stockfish, Houdini, Komodo, Fire, Chiron and Andscacs. The engines will play 8RRs with a time control of 90+10. Will Stockfish dominate like it did in season 11? Can Komodo make a come-back? Can Ginkgo and Jonny survive this division? We will soon find out.
A few interesting games
Laser - Fizbo: The engines gradually exchanged most pieces, reaching a QB vs QB position on move 39. Laser was a pawn up but evals were close to 0. Fizbo was expecting a perpetual check draw, but it blundered quite unexpectedly and allowed Laser to take over and force a queen exchange. The black king was trapped and the white passers could not be stopped by the black bishop alone. A weird blunder by Fizbo, score down to +1.
Texel - Booot: Booot had a space advantage and an open file for attacking the white king. Texel's pieces were on the queen side and were not able to defend the king, its passers were not a strong enough threat. Booot won a rook for a knight and the game was adjudicated before it won more material. Booot now has +3.
Jonny - Fritz: The game reached a QR vs QR position on move 25, material was equal but Jonny had more space and the initiative. Fritz tried to defend its pawns, Jonny created a connected pair of passers on the queen side and reduced to a won rook ending. In the reverse game Fritz was a pawn up with evals slowly increasing when it crashed. This was the first crash of the division, Fritz down to -3 still above the relegation zone, Jonny joins Ginkgo in the lead with +4.
Fizbo - Texel: Fizbo's eval jumped above 1 when it managed to make a hole in the pawn wall and to push a passer to h6. Texel was not worried at first, after queens were exchanged Fizbo chased the black king out of the corner and the passer became very dangerous. Texel had no counter and the game was adjuducated before it started to lose material. The reverse game reached a QRN vs QRR position on move 26, Texel two pawns up and evals close to 0. Fizbo's eval jumped when Texel started a series of exchanges that resulted in an imbalance of a R vs RR with 4 pawns up for white. Texel realized its mistake too late. With this double win Fizbo completes 4 wins against Texel in the division, it stays close to the lead with a +3 score.
Laser - Gull: Laser was a pawn up in a closed position, Gull had a bishop pair but Laser's knights were more effective. After the center opened a little the evals started to increase, all rooks were exchanged and Laser was much more active. After exchanging queens Laser captured another pawn the game reached a NN vs BN position and the marching white passers gave Laser the win. Gull at -5 is deep in the relegation zone, Laser with +1 is starting to look up.
Fritz - Booot: Fritz launched an early king side attack, Booot got its king out of trouble but was two pawns down on move 20. Evals continued to increase as the engines exchanged pieces, the game was adjudicated after Fritz captured a third pawn. Fritz at -2 keeping a distance from relegation, Booot with +2.
Ginkgo - Jonny: Jonny's king was stuck in the center, evals increased quickly as Ginkgo arranged its pieces behind the pawn line. After opening the center and exchanging most pieces Ginkgo was a pawn up in a QR vs QR position with eval over 2. The game was over when Ginkgo's king side pawns marched forward while keeping the king shielded from attacks. In the reverse game Ginkgo's king walked to the king side and Jonny castled long and pushed its king side pawns forward. Ginkgo could hardly move in a QRR vs QRR position, Jonny captured two pawns and exchanged queens leading to a won double rook ending. Two white wins in this opening, Ginkgo and Jonny stay in the lead.
With 5 double rounds to go the scores are: Ginkgo +4, Jonny +4, Fizbo +3, Booot +2, Laser +1, Fritz -2, Gull -5, Texel -7.
Fritz - Texel: Fritz had a bishop pair advantage early in the game. Evals constantly increased, after exchanging queens Fritz captured two pawns and created a pair of connected passers on the queen side, evals more than 2. Texel had to give a rook to stop the passers, and though Fritz was left without pawns it was winning in a RRB vs RN position. Fritz at -1 is 2 points away from the relegation zone.
Booot - Ginkgo: Evals were close to 0 in a closed position, when Ginkgo opened the center and the engines exchanged pieces the evals increased over 1. The game reached a RB vs RB position, the bishops were bosh DSBs, all the black pawns were on dark squares while almost all the white pawns were on light squares. Booot created two passers and pushed them to the 7the rank. Ginkgo had to lose a piece to stop them, a win for Booot. Booot catches up with Ginkgo and Fizbo with +3, Jonny in the lead with +4.
Fizbo - Fritz: Fritz was a pawn up from the opening. Fizbo gave a rook and a knight and could have drawn with perpetual check, but it chose to play on, trade a rook for the black queen and reach a Q vs RRN position with two pawns up. Fritz quickly took the initiative, it had to stop the white passers while avoiding checks from the queen. It took a while but Fritz managed to stop the passers and keep one pawn for the win.
Jonny - Gull: Jonny had an eval advantage of over 1 from the start, there were only a few exchanges in the firsts 40 moves. The engines started to exchange pieces and evals gradually increased, after exchanging queens Jonny saw the win. With only N vs B left the bishop was trapped and the engines raced with their passers. Both engines queened but the QN vs Q ending was a win for Jonny.
Ginkgo and Booot managed only draws against Texel and Laser.
With 3 double rounds to go the scores are: Jonny +5, Ginkgo +3, Booot +3, Fizbo +2, Laser +1, Fritz 0, Gull -6, Texel -8.
Jonny - Fizbo: two draws. Jonny pushed hard as black but a pawn advantage was not enough.
Ginkgo - Fritz: The game reached a BN vs R imbalance on move 35, evals started to increase despite Fritz' connected pair of passers on the queen side. Ginkgo moved its king forward, exchanged down to a BN vs R ending, and created a king side passer. The black rook couldn't get back in time, a win for Ginkgo.
Texel - Laser: Texel created a pair of connected passers and Laser couldn't stop them without losing material. In the reverse game Laser was a pawn up in a QB vs QN position, it took a while before Laser found how to convert the advantage to a win. Two white wins in this opening.
Only Ginkgo improves to +4 in the double round, two double rounds left, the promotion race is still wide open.
Fizbo - Ginkgo: Ginkgo won in black, see featured game below. In the reverse game Ginkgo was up a pawn by move 18. After capturing a second pawn Ginkgo reduced to a RB vs RB position. Fizbo got one pawn back but after exchanging rooks Ginkgo was winning. Two wins for Ginkgo and it leads the table with a +6 score. Fizbo with a 0 score is out of the race almost surely.
Jonny - Booot: With only RB vs RN left Booot gave the knight for two pawns to stop an advanced passer. Jonny had only one pawn left, it managed to keep it and slowly capture the black pawns. In the end Booot could either lose the last pawn or get mated. In the reverse game Booot pushed pawns on the queen side and exposed the black king. Jonny gave a rook for a knight to escape the strong attack on its king, Booot converted the QR vs QB ending. Another opening with two white wins.
After this double round Fizbo is out of the race, the top scores are Ginkgo +6, Jonny +5, Booot +3 with two games left.
Fizbo - Booot: The game reached a QRN vs QRB position on move 31. Fizbo's eval slowly increased, it took more than 40 moves for Fizbo to reduce to a rook ending two pawns up and a win. In the reverse game Booot had an eval advantage until move 39, it had a material advantage of a rook for a bishop while Fizbo had a bishop pair and queen side passers. Booot gave the material back and reduced to a QR vs QR position, Fizbo was a pawn down but had a passer on the 2nd rank, its eval jumped over 1. Booot was forced to lose the rook for the passer, and two pawns were not enough compensation for a rook. This double win for Fizbo means that Booot is out of the race as well, Ginkgo and Jonny promote to the premier division.
Texel - Jonny: Jonny gave a knight for a strong queen side attack on the white king. At first evals were in favor of Texel, then Texel traded RR for Q and evals became negative. Texel lost pawns and gave a bishop to stop a passer, it won some material back through a passer of its own and the game reached a Q vs RB ending with 3 black pawns. Jonny slowly moved the pawns forward, with its pieces creating a shield against attacks from the white queen. Texel could not stop the pawns and the game was adjudicated.
Featured game: Fizbo - Ginkgo
Division 1, double round 13
Link to game on TCEC
Fizbo had an eval advantage from the start. There were almost no exchanges, the position was closed and Fizbo had more space and a knight outpost near the black king and queen. Ginkgo's eval remained low, its position looked solid.
Ginkgo exchanged the forward white knight, which was then replaced by pawns. Nevertheless the black king was relatively safe while the white king was quite exposed. Evals became negative as Ginkgo removed all the pawns surrounding Fizbo's king and opened the position.
Fizbo had to defend against Ginkgo's many threats, in particular to deny the f3 square from the black knight where it could fork several pieces. Ginkgo opened the a8-h1 diagonal, adding mate threats with its queen.
Fizbo's defense collapsed and it began to lose material, the game was soon over.
The final draw rate is 58%.
Ginkgo and Jonny are promoted to the premier division, Gull and Texel are relegated to division 2. The promotion race was very close for most of the division. Fizbo and Booot were dropped only in the last two double rounds, Fizbo with a double loss to Ginkgo and Booot with a double loss to Fizbo.
It is a little sad to see Gull relegated. Only a few seasons ago it was one of the top engines, but it is not being developed and new engines are constanty improving. Check out season 8 stage 3 coverage (part 1, part 2, part 3, part 4, part 5, part 6), this was the stage where Stockfish had a bug causing it to lose on time. Gull led the stage and was close to reaching the superfinal.
There was one crash in the division, Fritz crashed against Jonny. In the end this crash did not have an effect of the final results, Jonny would have promoted without that point, and Fritz was not in danger of relegation as a result of the lost point.
The premier division will start soon. This is the main event of the season, Jonny and Ginkgo will be joined by Stockfish, Houdini, Komodo, Fire, Chiron and Andscacs. The engines will play 8RRs with a time control of 90+10. Will Stockfish dominate like it did in season 11? Can Komodo make a come-back? Can Ginkgo and Jonny survive this division? We will soon find out.
A few interesting games
Laser - Fizbo: The engines gradually exchanged most pieces, reaching a QB vs QB position on move 39. Laser was a pawn up but evals were close to 0. Fizbo was expecting a perpetual check draw, but it blundered quite unexpectedly and allowed Laser to take over and force a queen exchange. The black king was trapped and the white passers could not be stopped by the black bishop alone. A weird blunder by Fizbo, score down to +1.
Texel - Booot: Booot had a space advantage and an open file for attacking the white king. Texel's pieces were on the queen side and were not able to defend the king, its passers were not a strong enough threat. Booot won a rook for a knight and the game was adjudicated before it won more material. Booot now has +3.
Jonny - Fritz: The game reached a QR vs QR position on move 25, material was equal but Jonny had more space and the initiative. Fritz tried to defend its pawns, Jonny created a connected pair of passers on the queen side and reduced to a won rook ending. In the reverse game Fritz was a pawn up with evals slowly increasing when it crashed. This was the first crash of the division, Fritz down to -3 still above the relegation zone, Jonny joins Ginkgo in the lead with +4.
Fizbo - Texel: Fizbo's eval jumped above 1 when it managed to make a hole in the pawn wall and to push a passer to h6. Texel was not worried at first, after queens were exchanged Fizbo chased the black king out of the corner and the passer became very dangerous. Texel had no counter and the game was adjuducated before it started to lose material. The reverse game reached a QRN vs QRR position on move 26, Texel two pawns up and evals close to 0. Fizbo's eval jumped when Texel started a series of exchanges that resulted in an imbalance of a R vs RR with 4 pawns up for white. Texel realized its mistake too late. With this double win Fizbo completes 4 wins against Texel in the division, it stays close to the lead with a +3 score.
Laser - Gull: Laser was a pawn up in a closed position, Gull had a bishop pair but Laser's knights were more effective. After the center opened a little the evals started to increase, all rooks were exchanged and Laser was much more active. After exchanging queens Laser captured another pawn the game reached a NN vs BN position and the marching white passers gave Laser the win. Gull at -5 is deep in the relegation zone, Laser with +1 is starting to look up.
Fritz - Booot: Fritz launched an early king side attack, Booot got its king out of trouble but was two pawns down on move 20. Evals continued to increase as the engines exchanged pieces, the game was adjudicated after Fritz captured a third pawn. Fritz at -2 keeping a distance from relegation, Booot with +2.
Ginkgo - Jonny: Jonny's king was stuck in the center, evals increased quickly as Ginkgo arranged its pieces behind the pawn line. After opening the center and exchanging most pieces Ginkgo was a pawn up in a QR vs QR position with eval over 2. The game was over when Ginkgo's king side pawns marched forward while keeping the king shielded from attacks. In the reverse game Ginkgo's king walked to the king side and Jonny castled long and pushed its king side pawns forward. Ginkgo could hardly move in a QRR vs QRR position, Jonny captured two pawns and exchanged queens leading to a won double rook ending. Two white wins in this opening, Ginkgo and Jonny stay in the lead.
With 5 double rounds to go the scores are: Ginkgo +4, Jonny +4, Fizbo +3, Booot +2, Laser +1, Fritz -2, Gull -5, Texel -7.
Fritz - Texel: Fritz had a bishop pair advantage early in the game. Evals constantly increased, after exchanging queens Fritz captured two pawns and created a pair of connected passers on the queen side, evals more than 2. Texel had to give a rook to stop the passers, and though Fritz was left without pawns it was winning in a RRB vs RN position. Fritz at -1 is 2 points away from the relegation zone.
Booot - Ginkgo: Evals were close to 0 in a closed position, when Ginkgo opened the center and the engines exchanged pieces the evals increased over 1. The game reached a RB vs RB position, the bishops were bosh DSBs, all the black pawns were on dark squares while almost all the white pawns were on light squares. Booot created two passers and pushed them to the 7the rank. Ginkgo had to lose a piece to stop them, a win for Booot. Booot catches up with Ginkgo and Fizbo with +3, Jonny in the lead with +4.
Fizbo - Fritz: Fritz was a pawn up from the opening. Fizbo gave a rook and a knight and could have drawn with perpetual check, but it chose to play on, trade a rook for the black queen and reach a Q vs RRN position with two pawns up. Fritz quickly took the initiative, it had to stop the white passers while avoiding checks from the queen. It took a while but Fritz managed to stop the passers and keep one pawn for the win.
Jonny - Gull: Jonny had an eval advantage of over 1 from the start, there were only a few exchanges in the firsts 40 moves. The engines started to exchange pieces and evals gradually increased, after exchanging queens Jonny saw the win. With only N vs B left the bishop was trapped and the engines raced with their passers. Both engines queened but the QN vs Q ending was a win for Jonny.
Ginkgo and Booot managed only draws against Texel and Laser.
With 3 double rounds to go the scores are: Jonny +5, Ginkgo +3, Booot +3, Fizbo +2, Laser +1, Fritz 0, Gull -6, Texel -8.
Jonny - Fizbo: two draws. Jonny pushed hard as black but a pawn advantage was not enough.
Ginkgo - Fritz: The game reached a BN vs R imbalance on move 35, evals started to increase despite Fritz' connected pair of passers on the queen side. Ginkgo moved its king forward, exchanged down to a BN vs R ending, and created a king side passer. The black rook couldn't get back in time, a win for Ginkgo.
Texel - Laser: Texel created a pair of connected passers and Laser couldn't stop them without losing material. In the reverse game Laser was a pawn up in a QB vs QN position, it took a while before Laser found how to convert the advantage to a win. Two white wins in this opening.
Only Ginkgo improves to +4 in the double round, two double rounds left, the promotion race is still wide open.
Fizbo - Ginkgo: Ginkgo won in black, see featured game below. In the reverse game Ginkgo was up a pawn by move 18. After capturing a second pawn Ginkgo reduced to a RB vs RB position. Fizbo got one pawn back but after exchanging rooks Ginkgo was winning. Two wins for Ginkgo and it leads the table with a +6 score. Fizbo with a 0 score is out of the race almost surely.
Jonny - Booot: With only RB vs RN left Booot gave the knight for two pawns to stop an advanced passer. Jonny had only one pawn left, it managed to keep it and slowly capture the black pawns. In the end Booot could either lose the last pawn or get mated. In the reverse game Booot pushed pawns on the queen side and exposed the black king. Jonny gave a rook for a knight to escape the strong attack on its king, Booot converted the QR vs QB ending. Another opening with two white wins.
After this double round Fizbo is out of the race, the top scores are Ginkgo +6, Jonny +5, Booot +3 with two games left.
Fizbo - Booot: The game reached a QRN vs QRB position on move 31. Fizbo's eval slowly increased, it took more than 40 moves for Fizbo to reduce to a rook ending two pawns up and a win. In the reverse game Booot had an eval advantage until move 39, it had a material advantage of a rook for a bishop while Fizbo had a bishop pair and queen side passers. Booot gave the material back and reduced to a QR vs QR position, Fizbo was a pawn down but had a passer on the 2nd rank, its eval jumped over 1. Booot was forced to lose the rook for the passer, and two pawns were not enough compensation for a rook. This double win for Fizbo means that Booot is out of the race as well, Ginkgo and Jonny promote to the premier division.
Texel - Jonny: Jonny gave a knight for a strong queen side attack on the white king. At first evals were in favor of Texel, then Texel traded RR for Q and evals became negative. Texel lost pawns and gave a bishop to stop a passer, it won some material back through a passer of its own and the game reached a Q vs RB ending with 3 black pawns. Jonny slowly moved the pawns forward, with its pieces creating a shield against attacks from the white queen. Texel could not stop the pawns and the game was adjudicated.
Featured game: Fizbo - Ginkgo
Division 1, double round 13
Link to game on TCEC
Fizbo had an eval advantage from the start. There were almost no exchanges, the position was closed and Fizbo had more space and a knight outpost near the black king and queen. Ginkgo's eval remained low, its position looked solid.
Ginkgo exchanged the forward white knight, which was then replaced by pawns. Nevertheless the black king was relatively safe while the white king was quite exposed. Evals became negative as Ginkgo removed all the pawns surrounding Fizbo's king and opened the position.
Fizbo had to defend against Ginkgo's many threats, in particular to deny the f3 square from the black knight where it could fork several pieces. Ginkgo opened the a8-h1 diagonal, adding mate threats with its queen.
Fizbo's defense collapsed and it began to lose material, the game was soon over.
Season 12 division 1 statistics
Draw rate, wins
Final draw rate was 58.0%.
Game termination
The three most common game termination causes were:
39.3% - TCEC win rule
33.9% - TCEC draw rule
12.5% - 3-Fold repetition
There was one game that ended in a crash.
Moves per game
Median= 59.5
Average= 70.1
There were 14 games with more than 100 moves.
Time per game (hours)
Median= 2:06
Average= 2:05
Openings
There was no change in the opening variants information, still not as detailed as in previous seasons.
There were 5-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 - 12 times
QGD, King's Indian, Sicillian defence - 8 times each
(the naming problem is evident here, the Sicillian openings are not all given the same name)
The engines had some freedom to choose the opening variant. In 87.5% of the game pairs the same ECO code was repeated, all of them repeated the same opening variant, another demonstration of the lack of detail in opening variants. 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, 89.3% of the pairs diverged at most after 1 move. Remarkably there were no pairs that repeated more than 2 moves.
Final draw rate was 58.0%.
Game termination
The three most common game termination causes were:
39.3% - TCEC win rule
33.9% - TCEC draw rule
12.5% - 3-Fold repetition
There was one game that ended in a crash.
Moves per game
Median= 59.5
Average= 70.1
There were 14 games with more than 100 moves.
Time per game (hours)
Median= 2:06
Average= 2:05
Openings
There was no change in the opening variants information, still not as detailed as in previous seasons.
There were 5-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 - 12 times
QGD, King's Indian, Sicillian defence - 8 times each
(the naming problem is evident here, the Sicillian openings are not all given the same name)
The engines had some freedom to choose the opening variant. In 87.5% of the game pairs the same ECO code was repeated, all of them repeated the same opening variant, another demonstration of the lack of detail in opening variants. 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, 89.3% of the pairs diverged at most after 1 move. Remarkably there were no pairs that repeated more than 2 moves.
Friday, May 18, 2018
Season 12 live blogging, Friday May 18th
There are 3 double rounds left in division 1 and the race for promotion is very close. Today we follow the game between the leader Jonny and Fizbo which is joint second with Ginkgo and Booot. This is the reverse game, the previous game ended in a draw.
move 8 - Jonny diverges by retaking with the queen instead of the pawn, this is the less typical move here. Jonny with an eval advantage, but Jonny is usually an optimist.
move 13 - Fizbo is a pawn up but its pawn structure is worse. Fizbo needs to castle its king. Evals dropped after Jonny's last move, Jonny is seeing a drawn rook ending in its PV.
move 15 - Fizbo castled and evals are close to 0. Jonny is taking a long time to make moves, almost twice as much as Fizbo, but it doesn't seem to have a plan.
move 18 - a pair of rooks exchanged, if the engines start trading pieces this may be a short draw.
move 23 - Fizbo's eval turned negative and came back to 0, it is seeing a repetition draw in the PV but I doubt Jonny will choose this line. Fizbo is still a pawn up.
move 28 - a series of exchanges and only RB vs RN are left. Jonny's eval is still positive and increasing, can it outplay Fizbo in the ending?
--- away from keyboard for 20 minutes ---
--- back again ---
move 50 - for a while Jonny thought it was better, it moved its king forward and captured a pawn. However, evals came back down, Fizbo exchanged the minor pieces and the rook ending looks like a draw.
move 55 - Jonny lowered its eval and the draw rule kicked in.
This result is not good for Fizbo, now Ginkgo and Booot have a chance to increase the gap to Fizbo and to get closer to Jonny and a promotion spot.
move 8 - Jonny diverges by retaking with the queen instead of the pawn, this is the less typical move here. Jonny with an eval advantage, but Jonny is usually an optimist.
move 13 - Fizbo is a pawn up but its pawn structure is worse. Fizbo needs to castle its king. Evals dropped after Jonny's last move, Jonny is seeing a drawn rook ending in its PV.
move 15 - Fizbo castled and evals are close to 0. Jonny is taking a long time to make moves, almost twice as much as Fizbo, but it doesn't seem to have a plan.
move 18 - a pair of rooks exchanged, if the engines start trading pieces this may be a short draw.
move 23 - Fizbo's eval turned negative and came back to 0, it is seeing a repetition draw in the PV but I doubt Jonny will choose this line. Fizbo is still a pawn up.
move 28 - a series of exchanges and only RB vs RN are left. Jonny's eval is still positive and increasing, can it outplay Fizbo in the ending?
--- away from keyboard for 20 minutes ---
--- back again ---
move 50 - for a while Jonny thought it was better, it moved its king forward and captured a pawn. However, evals came back down, Fizbo exchanged the minor pieces and the rook ending looks like a draw.
move 55 - Jonny lowered its eval and the draw rule kicked in.
This result is not good for Fizbo, now Ginkgo and Booot have a chance to increase the gap to Fizbo and to get closer to Jonny and a promotion spot.
Wednesday, May 16, 2018
Season 12 division 1 after RR2
Crosstable after RR2
Division 1 completed 2 of 4 RRs, what a contrast this division is to division 2. There were no crashes, the draw rate is higher at 58.9%, the games are longer. It feels like the engines in this division take chess more seriously. Every engine has won and lost at least one game, there are no games where the result is known in advance.
There are 4 engines with a good chance of promoting to the premier division. Ginkgo is currently in the lead, 1 point ahead of Booot, Jonny and Fizbo. Fizbo and Ginkgo were relegated from the premier division in season 11, and it is not easy to get back.
Gull and Texel are in the relegation zone, 1.5 points behind the nearest rival. Texel may be on its way back to division 2, where it played earlier in the season. It is sad to see Gull in danger of relegation, it was a top engine a few seasons ago but has not been updated for a while now.
A few interesting games
Ginkgo - Gull: The game reached a QRN vs QRB position on move 31, Ginkgo saw a win and reduced to a N vs B ending, it took Gull 9 moves to see it was going to lose. This was the first decisive game of the division. The reverse was a draw.
Jonny - Fritz: Jonny gave a rook for a knight and pawn, exposing the black king. Fritz counter-attacked the white king, and the game reduced to a RB vs BN position and a win for Fritz. This was the first win in black in the division.
Fizbo - Texel: Texel was two pawns up and it tried to attack on the king side. Fizbo's king hid behind a black pawn and was completely safe. Fizbo counter-attacked and Texel's position collapsed quickly. In the reverse Fizbo had a bishop pair advantage in an open position. It gave a rook for a knight and created a connected pair of passers that secured the win. A double win for Fizbo, leading the table.
Gull - Laser: Gull opened the king side after 34 moves and Laser gave a knight to stop the white rooks from invading. This led to a QBB vs QB position with Gull two pawns behind, and an unstoppable passer on the queen side was enough for a Gull win. Gull can still be strong, though it hasn't changed in years and has no tablebase support.
Booot - Fritz: Booot opened the king side with a pawn storm, the result was a pawn up for Fritz with both kings uncastled in an open position. Booot's pieces soon had the black king in a mating net, Booot pushed passers on the queen side and Fritz' defense collapsed. In the reverse game Booot exchanged many pieces and equalized quickly.
Gull - Fizbo: Gull was two pawns up and had an eval of over 2, but it failed to notice the danger to its king. Fizbo found safety for its king behind a white pawn and then created a mating net with pawns in the center and rooks in an open file on the king side. Gull realized too late that it had no defense without losing material.
Fritz - Texel: Fritz gave a rook for a bishop and two pawns, it thought it was slightly better with a queen side passer and a bishop pair. Texel used its pieces to attack the white king and force a queen exchange, evals turned negative. The game reached a BN vs RB position and seemed to be heading for a draw, when Fritz blundered, exchanged pawns and reduced to a black won 6-man position. Apparently Frits only used a 5-man tablebase, and it took 25 moves for it to see the loss.
Gull - Jonny: After a very long shuffle in a QBN vs QBN position Jonny found a way to improve after move 100, with time running out. After exchanging queens it won two pawns while Gull placed a passer on the 7th rank. Jonny blocked the white passer and pushed its own pawns, Gull was forced to lose its pieces to stop them. In the reverse game Jonny trapped a black bishop and was a pawn up by move 20, with eval over 2. After creating a central passer and exchanging most pieces Jonny won the game easily. With this double win Jonny has a positive +1 score, while Gull is sinking with -3.
Texel - Ginkgo: Texel had an eval advantage and was two pawns up on move 25, then in a very complicated position Ginkgo found a surprise move, and 10 moves later at the end of a series of exchanges Ginkgo was up a rook for 3 pawns. The game was soon over. In the reverse game Ginkgo kept its initial advantage, it gradually exchanged pieces until reaching a rook ending. The material was balanced but the white king and pawns were much better, Ginkgo soon had an advancing passer that Texel couldn't stop. A second double loss for Texel, it joins Gull at the bottom with a -3 score. Ginkgo joins Fizbo at the top with a +3 score.
Booot - Laser: After many early exchanges the game reached a BB vs R imbalance on move 25. Laser had two extra pawns while Booot had a central passer, its bishop pair very strong in an open position. Evals increased gradually, after exchanging queens Laser gave a rook for a bishop to remove the black passer. The remaining RB vs R position was a win for Booot. Laser with a second loss is close to the relegation zone, Booot with a second win is close to the top.
Gull - Booot: Gull had a strong pair of knights deep in Booot's territory, it used an open file on the king side to threaten the black king. Booot gave a bishop and reduced to a NN vs N position, but Gull kept enough pawns alive to secure the win. This was the first loss for Booot in the division. In the reverse game Booot had a space advantage and evals increased gradually. Gull's king was exposed and its pieces on the wrong side of the board. Gull gave a rook for a knight, but the pressure on the king gave Booot the win. First opening with two white wins.
Laser - Texel: Laser was a pawn up in a RRB vs RRN position. Apparently Texel blundered and gave a second pawn, it realized the mistake a move later. After a series of exchanges evals were over 2 in a RR vs RN position, Laser's queen side passer secured the win. This was Laser's first win in the division.
Fizbo - Ginkgo: The engines castled in opposite directions, Fizbo pushed pawns on the king side and opened a file with a pawn reaching the 7th rank. Ginkgo remained calm and exchanged pieces, its eval around 0.5. Fizbo's eval was over 2 but seemed to be stable, but it jumped when Ginkgo gave a rook for a bishop. Ginkgo regained the material a few moves later, but then it also realized the resulting rook ending was lost. In the reverse game Ginkgo won, see featured game below. Another biased opening, first losses for both Ginkgo and Fizbo, both still in the lead.
Booot - Jonny: The game started with balanced evals, on move 21 Jonny took the initiative and attacked the exposed white king. The attack lasted more than 15 moves until queens were exchanged, Jonny's eval remained above 2. Booot reduced to a rook endgame a pawn down, Jonny's pawns were better and it was able to convert the win. Jonny gets closer to the leaders with a score of +2.
Laser - Fritz: Laser had an eval advantage from the start that was close to 2 on move 20 and kept increasing. On move 30 Laser was 2 pawns up, then it exchanged pieces until reaching a rook ending and an easy win.
Booot - Fizbo: Fizbo was happy with its opening for the first 20 moves, then evals started to turn in Booot's favor. On move 29 only QRB vs QRB remained, Booot pushed a central passer that got to the 7th rank with eval over 1. Fizbo tried to counter by attacking the white king, eventually Booot combined the queening and checks on the black king to win a piece and the game. Booot staying close to the top with +2.
Laser - Ginkgo: The engines had a few exchanges, then they stabilized the pawns and started to shuffle. On move 78 Ginkgo gave a rook for a knight and pawns, broke through the center and destroyed the white king's defense with its advancing pawns. Laser had to lose material to avoid mate, but it was lost in any case. Ginkgo takes the lead with a +4 score.
Fritz - Gull: Fritz was a pawn up in a RB vs RN position. For a while Gull held the position, but after reducing to a rook ending the black pawns started to fall and Fritz won. Gull joins Texel at the bottom with a -4 score.
Featured game: Ginkgo - Fizbo
Division 1, double round 6
Link to game on TCEC
Ginkgo was ahead on eval, it pushed pawns on the king side and threatened to expose the black king. Its own king remained uncastled in the center and Fizbo gave a pawn to open the center, and then seemed to ignore its rook being attacked.
For 5 moves the rook remained under attack and Ginkgo did not take it. Each move Fizbo had a different threat that Ginkgo had to defend against first. Finally the rook moved away, but now Ginkgo captured a hanging knight and was a full piece ahead on move 22.
Fizbo's eval was below 1, it compensated by attacking the exposed uncastled white king. The white bishop was also blocked in the corner and weak. However, evals continued to increase as Fizbo's attack was not effective and Ginkgo had a king side threat on the black king.
After both kings found relative safety the piece advantage became evident. Ginkgo forced its way on the king side, it untangled its knight and started to make mating threats. Fizbo couldn't protect its king side pawns and the added threat of a white passer finished the game.
Division 1 completed 2 of 4 RRs, what a contrast this division is to division 2. There were no crashes, the draw rate is higher at 58.9%, the games are longer. It feels like the engines in this division take chess more seriously. Every engine has won and lost at least one game, there are no games where the result is known in advance.
There are 4 engines with a good chance of promoting to the premier division. Ginkgo is currently in the lead, 1 point ahead of Booot, Jonny and Fizbo. Fizbo and Ginkgo were relegated from the premier division in season 11, and it is not easy to get back.
Gull and Texel are in the relegation zone, 1.5 points behind the nearest rival. Texel may be on its way back to division 2, where it played earlier in the season. It is sad to see Gull in danger of relegation, it was a top engine a few seasons ago but has not been updated for a while now.
A few interesting games
Ginkgo - Gull: The game reached a QRN vs QRB position on move 31, Ginkgo saw a win and reduced to a N vs B ending, it took Gull 9 moves to see it was going to lose. This was the first decisive game of the division. The reverse was a draw.
Jonny - Fritz: Jonny gave a rook for a knight and pawn, exposing the black king. Fritz counter-attacked the white king, and the game reduced to a RB vs BN position and a win for Fritz. This was the first win in black in the division.
Fizbo - Texel: Texel was two pawns up and it tried to attack on the king side. Fizbo's king hid behind a black pawn and was completely safe. Fizbo counter-attacked and Texel's position collapsed quickly. In the reverse Fizbo had a bishop pair advantage in an open position. It gave a rook for a knight and created a connected pair of passers that secured the win. A double win for Fizbo, leading the table.
Gull - Laser: Gull opened the king side after 34 moves and Laser gave a knight to stop the white rooks from invading. This led to a QBB vs QB position with Gull two pawns behind, and an unstoppable passer on the queen side was enough for a Gull win. Gull can still be strong, though it hasn't changed in years and has no tablebase support.
Booot - Fritz: Booot opened the king side with a pawn storm, the result was a pawn up for Fritz with both kings uncastled in an open position. Booot's pieces soon had the black king in a mating net, Booot pushed passers on the queen side and Fritz' defense collapsed. In the reverse game Booot exchanged many pieces and equalized quickly.
Gull - Fizbo: Gull was two pawns up and had an eval of over 2, but it failed to notice the danger to its king. Fizbo found safety for its king behind a white pawn and then created a mating net with pawns in the center and rooks in an open file on the king side. Gull realized too late that it had no defense without losing material.
Fritz - Texel: Fritz gave a rook for a bishop and two pawns, it thought it was slightly better with a queen side passer and a bishop pair. Texel used its pieces to attack the white king and force a queen exchange, evals turned negative. The game reached a BN vs RB position and seemed to be heading for a draw, when Fritz blundered, exchanged pawns and reduced to a black won 6-man position. Apparently Frits only used a 5-man tablebase, and it took 25 moves for it to see the loss.
Gull - Jonny: After a very long shuffle in a QBN vs QBN position Jonny found a way to improve after move 100, with time running out. After exchanging queens it won two pawns while Gull placed a passer on the 7th rank. Jonny blocked the white passer and pushed its own pawns, Gull was forced to lose its pieces to stop them. In the reverse game Jonny trapped a black bishop and was a pawn up by move 20, with eval over 2. After creating a central passer and exchanging most pieces Jonny won the game easily. With this double win Jonny has a positive +1 score, while Gull is sinking with -3.
Texel - Ginkgo: Texel had an eval advantage and was two pawns up on move 25, then in a very complicated position Ginkgo found a surprise move, and 10 moves later at the end of a series of exchanges Ginkgo was up a rook for 3 pawns. The game was soon over. In the reverse game Ginkgo kept its initial advantage, it gradually exchanged pieces until reaching a rook ending. The material was balanced but the white king and pawns were much better, Ginkgo soon had an advancing passer that Texel couldn't stop. A second double loss for Texel, it joins Gull at the bottom with a -3 score. Ginkgo joins Fizbo at the top with a +3 score.
Booot - Laser: After many early exchanges the game reached a BB vs R imbalance on move 25. Laser had two extra pawns while Booot had a central passer, its bishop pair very strong in an open position. Evals increased gradually, after exchanging queens Laser gave a rook for a bishop to remove the black passer. The remaining RB vs R position was a win for Booot. Laser with a second loss is close to the relegation zone, Booot with a second win is close to the top.
Gull - Booot: Gull had a strong pair of knights deep in Booot's territory, it used an open file on the king side to threaten the black king. Booot gave a bishop and reduced to a NN vs N position, but Gull kept enough pawns alive to secure the win. This was the first loss for Booot in the division. In the reverse game Booot had a space advantage and evals increased gradually. Gull's king was exposed and its pieces on the wrong side of the board. Gull gave a rook for a knight, but the pressure on the king gave Booot the win. First opening with two white wins.
Laser - Texel: Laser was a pawn up in a RRB vs RRN position. Apparently Texel blundered and gave a second pawn, it realized the mistake a move later. After a series of exchanges evals were over 2 in a RR vs RN position, Laser's queen side passer secured the win. This was Laser's first win in the division.
Fizbo - Ginkgo: The engines castled in opposite directions, Fizbo pushed pawns on the king side and opened a file with a pawn reaching the 7th rank. Ginkgo remained calm and exchanged pieces, its eval around 0.5. Fizbo's eval was over 2 but seemed to be stable, but it jumped when Ginkgo gave a rook for a bishop. Ginkgo regained the material a few moves later, but then it also realized the resulting rook ending was lost. In the reverse game Ginkgo won, see featured game below. Another biased opening, first losses for both Ginkgo and Fizbo, both still in the lead.
Booot - Jonny: The game started with balanced evals, on move 21 Jonny took the initiative and attacked the exposed white king. The attack lasted more than 15 moves until queens were exchanged, Jonny's eval remained above 2. Booot reduced to a rook endgame a pawn down, Jonny's pawns were better and it was able to convert the win. Jonny gets closer to the leaders with a score of +2.
Laser - Fritz: Laser had an eval advantage from the start that was close to 2 on move 20 and kept increasing. On move 30 Laser was 2 pawns up, then it exchanged pieces until reaching a rook ending and an easy win.
Booot - Fizbo: Fizbo was happy with its opening for the first 20 moves, then evals started to turn in Booot's favor. On move 29 only QRB vs QRB remained, Booot pushed a central passer that got to the 7th rank with eval over 1. Fizbo tried to counter by attacking the white king, eventually Booot combined the queening and checks on the black king to win a piece and the game. Booot staying close to the top with +2.
Laser - Ginkgo: The engines had a few exchanges, then they stabilized the pawns and started to shuffle. On move 78 Ginkgo gave a rook for a knight and pawns, broke through the center and destroyed the white king's defense with its advancing pawns. Laser had to lose material to avoid mate, but it was lost in any case. Ginkgo takes the lead with a +4 score.
Fritz - Gull: Fritz was a pawn up in a RB vs RN position. For a while Gull held the position, but after reducing to a rook ending the black pawns started to fall and Fritz won. Gull joins Texel at the bottom with a -4 score.
Featured game: Ginkgo - Fizbo
Division 1, double round 6
Link to game on TCEC
Ginkgo was ahead on eval, it pushed pawns on the king side and threatened to expose the black king. Its own king remained uncastled in the center and Fizbo gave a pawn to open the center, and then seemed to ignore its rook being attacked.
For 5 moves the rook remained under attack and Ginkgo did not take it. Each move Fizbo had a different threat that Ginkgo had to defend against first. Finally the rook moved away, but now Ginkgo captured a hanging knight and was a full piece ahead on move 22.
Fizbo's eval was below 1, it compensated by attacking the exposed uncastled white king. The white bishop was also blocked in the corner and weak. However, evals continued to increase as Fizbo's attack was not effective and Ginkgo had a king side threat on the black king.
After both kings found relative safety the piece advantage became evident. Ginkgo forced its way on the king side, it untangled its knight and started to make mating threats. Fizbo couldn't protect its king side pawns and the added threat of a white passer finished the game.
Sunday, May 13, 2018
Season 12 live blogging, Sunday May 13th
We look at the Ginkgo - Fritz game, game 36 in division 1. Ginkgo played in the premier division in season 11 and was relegated while Fritz promoted from division 2 this season. Ginkgo was in the lead after 4 double rounds, together with Fizbo, with a +3 score. Fritz had a score of -1 and is ond point above the relegation zone. In the previous game Fritz was white and had an eval advantage, Ginkgo exchanged most pieces and held the draw.
move 16 - the game quickly diverged from the previous game. Ginkgo is ahead on eval, so far keeping the pieces on the board. Fritz' king is uncastled in the center, Ginkgo has more space and has a strong knight in the center.
move 20 - Fritz exchanged knights in the center and castled. The center is mosly controlled by Ginkgo, the white pieces are well developed. The black queen is on the a file in a defensive role, Fritz would probably prefer to trade queens.
move 26 - the engines exchanged a pair of rooks, and most of the queen side pawns are gone. Ginkgo's eval stays below 1, can Fritz hold?
move 31 - Ginkgo is better and has the initiative, but Fritz seems to be solid enough. Both engines have a passer on the queen side that isn't moving yet. Ginkgo offered a knight, Fritz did not take (unlike its PV) to keep its king safe.
move 37 - Ginkgo offered the knight again and this time Fritz took it, Ginkgo regained the material in a series of exchanges that left QB vs QN. It seems the game is headed for a draw.
move 46 - Ginkgo exchanged queens, it is a pawn up in a B vs N ending and evals are close to 0. Waiting for the pawns to stop moving and the draw rule.
move 52 - the draw rule is activated, Fritz holds the draw.
I thought Ginkgo's starting advantage would develop into a win, but not this time.
move 16 - the game quickly diverged from the previous game. Ginkgo is ahead on eval, so far keeping the pieces on the board. Fritz' king is uncastled in the center, Ginkgo has more space and has a strong knight in the center.
move 20 - Fritz exchanged knights in the center and castled. The center is mosly controlled by Ginkgo, the white pieces are well developed. The black queen is on the a file in a defensive role, Fritz would probably prefer to trade queens.
move 26 - the engines exchanged a pair of rooks, and most of the queen side pawns are gone. Ginkgo's eval stays below 1, can Fritz hold?
move 31 - Ginkgo is better and has the initiative, but Fritz seems to be solid enough. Both engines have a passer on the queen side that isn't moving yet. Ginkgo offered a knight, Fritz did not take (unlike its PV) to keep its king safe.
move 37 - Ginkgo offered the knight again and this time Fritz took it, Ginkgo regained the material in a series of exchanges that left QB vs QN. It seems the game is headed for a draw.
move 46 - Ginkgo exchanged queens, it is a pawn up in a B vs N ending and evals are close to 0. Waiting for the pawns to stop moving and the draw rule.
move 52 - the draw rule is activated, Fritz holds the draw.
I thought Ginkgo's starting advantage would develop into a win, but not this time.
Thursday, May 10, 2018
Season 12 division 2 statistics
Draw rate, wins
Final draw rate was 47.3%. If we ignore the crashed games the draw rate was 52%.
Game termination
The three most common game termination causes were:
40.2% - TCEC win rule
26.8% - TCEC draw rule
14.3% - SyzygyTB
There were 10 games that ended in a crash.
Moves per game
Median= 59.3
Average= 64.9
There were two games that ended in crashes in under 20 moves, and there were 12 games with more than 100 moves.
Time per game (hours)
Median= 1:10
Average= 1:11
Openings
The openings information was fixed in division 2, and the openings section of the report returns. It should be noted that previously the opening variants were much more detailed than in the current version, I don't know if this is still under development.
There were 2-move book openings in this stage chosen by Cato. The first letter of the ECO codes was distributed as follows:
For each double round all eight games had the same book sequence (reverse games are played consecutively). There were 14 book sequences, each repeated 8 times. For some sequences all the openings had similar codes (e.g. Sicillian, Hungarian variation, or Scandinavian defense). In others the engines had more freedom to choose.
If we use the opening 'family name' (using format FAMILY_NAME: VARIANT....) the top 3 are:
QGD - 18 times
Sicillian - 16 times
Alekhine's defence, Caro-Kann, English, Scandinavian defence, Vienna game - 8 times each
The engines had a lot of freedom to choose the opening variant. Only 50% of the game pairs repeated the same ECO code twice, and 48.2% repeated the same opening variant, the high percentage is related to the lack of detail in opening variants. In 7.1% of the game pairs the ECO first letter was not repeated twice.
Reverse pairs, wins
Reverse pairs, same moves
Pairs of reverse games diverged very quickly, 69.6% of the pairs diverged at most after 1 move. One game pair repeated 5.5 moves after book (Arasan - Hannibal, games 53 and 54, Caro-Kann, Tartakower (Nimzovich) variation).
Final draw rate was 47.3%. If we ignore the crashed games the draw rate was 52%.
Game termination
The three most common game termination causes were:
40.2% - TCEC win rule
26.8% - TCEC draw rule
14.3% - SyzygyTB
There were 10 games that ended in a crash.
Moves per game
Median= 59.3
Average= 64.9
There were two games that ended in crashes in under 20 moves, and there were 12 games with more than 100 moves.
Time per game (hours)
Median= 1:10
Average= 1:11
Openings
The openings information was fixed in division 2, and the openings section of the report returns. It should be noted that previously the opening variants were much more detailed than in the current version, I don't know if this is still under development.
There were 2-move book openings in this stage chosen by Cato. The first letter of the ECO codes was distributed as follows:
For each double round all eight games had the same book sequence (reverse games are played consecutively). There were 14 book sequences, each repeated 8 times. For some sequences all the openings had similar codes (e.g. Sicillian, Hungarian variation, or Scandinavian defense). In others the engines had more freedom to choose.
If we use the opening 'family name' (using format FAMILY_NAME: VARIANT....) the top 3 are:
QGD - 18 times
Sicillian - 16 times
Alekhine's defence, Caro-Kann, English, Scandinavian defence, Vienna game - 8 times each
The engines had a lot of freedom to choose the opening variant. Only 50% of the game pairs repeated the same ECO code twice, and 48.2% repeated the same opening variant, the high percentage is related to the lack of detail in opening variants. In 7.1% of the game pairs the ECO first letter was not repeated twice.
Reverse pairs, wins
Reverse pairs, same moves
Pairs of reverse games diverged very quickly, 69.6% of the pairs diverged at most after 1 move. One game pair repeated 5.5 moves after book (Arasan - Hannibal, games 53 and 54, Caro-Kann, Tartakower (Nimzovich) variation).
Season 12 division 2 final results
Final Crosstable
HOWEVER, there were many crashes in the division and two engines - Arasan and Hannibal - were disqualified and automatically relegated to division 3. Their games were also ignored and as a result the final scores were: Fritz 13, Texel 11.5, Xiphos 11.5, ChessBrain 9.5, Nirvana 9, Vajolet2 5.5. Xiphos lost the tiebreak to Texel since it had a crash, but it also lost in the head-to-head tiebreak.
The crashes were the most significant feature of the division, unfortunately. There were 10 of these, Arasan had 4, Hannibal 3 and one each for Fritz, Xiphos and ChessBrain. There were 3 crashes in the first double round alone. The crashing engines of the first few rounds had their parameters changed to try to stop them from crashing again. This worked for about 30 games, and then Arasan crashed for the third time. If this wasn't enough in the 12th and 13th double rounds Hannibal crashed twice, causing it to be disqualified right at the end of the division.
The promotion race was very close, mainly between Xiphos, Fritz and Texel. It was hard to follow the action because of the crashes, the crosstable did not take Arasan's disqualification into account. Hannibal's disqualification changed everything in the last minute as well. Xiphos led the table at the halfway point and could have promoted for a third time this season - recall it started as a newcomer in division 4. In round 7 Fritz beat Xiphos and in round 10 Texel beat Xiphos. Then in round 11 Xiphos crashed, so it couldn't rely on tiebreaks to qualify. Entering the last round Fritz had a 0.5 point lead over Texel and Xiphos. Texel played against ChessBrain, and unexpectedly lost one game and got only one draw. Xiphos played Fritz in the last 2 games of the division, it needed at least a point to overtake Texel. However, Fritz beat Xiphos again, and in the last game Xiphos was ahead but could only get a draw.
DIvision 1 will start soon, Texel and Fritz will be joined by Fizbo, Ginkgo, Booot, Jonny, Gull and Laser. They will play 2 double RRs, same as division 2-4, but this time the time control will be 60+10 so games will be twice as long. I really hope we don't see as many crashes as in division 2, the engines should be stable at this level.
HOWEVER, there were many crashes in the division and two engines - Arasan and Hannibal - were disqualified and automatically relegated to division 3. Their games were also ignored and as a result the final scores were: Fritz 13, Texel 11.5, Xiphos 11.5, ChessBrain 9.5, Nirvana 9, Vajolet2 5.5. Xiphos lost the tiebreak to Texel since it had a crash, but it also lost in the head-to-head tiebreak.
The crashes were the most significant feature of the division, unfortunately. There were 10 of these, Arasan had 4, Hannibal 3 and one each for Fritz, Xiphos and ChessBrain. There were 3 crashes in the first double round alone. The crashing engines of the first few rounds had their parameters changed to try to stop them from crashing again. This worked for about 30 games, and then Arasan crashed for the third time. If this wasn't enough in the 12th and 13th double rounds Hannibal crashed twice, causing it to be disqualified right at the end of the division.
The promotion race was very close, mainly between Xiphos, Fritz and Texel. It was hard to follow the action because of the crashes, the crosstable did not take Arasan's disqualification into account. Hannibal's disqualification changed everything in the last minute as well. Xiphos led the table at the halfway point and could have promoted for a third time this season - recall it started as a newcomer in division 4. In round 7 Fritz beat Xiphos and in round 10 Texel beat Xiphos. Then in round 11 Xiphos crashed, so it couldn't rely on tiebreaks to qualify. Entering the last round Fritz had a 0.5 point lead over Texel and Xiphos. Texel played against ChessBrain, and unexpectedly lost one game and got only one draw. Xiphos played Fritz in the last 2 games of the division, it needed at least a point to overtake Texel. However, Fritz beat Xiphos again, and in the last game Xiphos was ahead but could only get a draw.
DIvision 1 will start soon, Texel and Fritz will be joined by Fizbo, Ginkgo, Booot, Jonny, Gull and Laser. They will play 2 double RRs, same as division 2-4, but this time the time control will be 60+10 so games will be twice as long. I really hope we don't see as many crashes as in division 2, the engines should be stable at this level.
Sunday, May 6, 2018
Season 12 live blogging, Sunday May 6th
Xiphos led the table after 2 double rounds, a bit of a surprise considering it started in division 4 and promoted twice as second. ChessBrain which promoted as first from division 3 isn't doing so well and is closer to relegation in division 2. Nirvana another engine in the promotion race, currently in 5th place but only 1 point behind Xiphos in 2nd (Texel is the current leader with 2 extra games). Xiphos won both games these engines played in the first double round.
I am watching game 8.5, Nirvana in white.
move 10 - so far the material is balanced and the evals close to 0. Both kings are castled and safe, Xiphos has more space on the king side, Nirvana's pieces are concentrated on the queen side.
move 16 - Xiphos is pushing pawns on both the king and the queen side, its eval turning negative. Nirvana is still happy so far, but its position is starting to look a bit cramped.
move 20 - the black king is quite exposed but there are no white pieces that can threaten it. Xiphos is moving more pieces to the king side, Nirvana looks like it is ready to create a pawn wall and start shuffling.
move 25 - Xiphos exchanged a pair of knights on the king side, its eval positive again. Nirvana's eval is increasing, it is countering on the queen side. Xiphos closed the queen side with its pawns, the c6 pawn is weak, not sure if Nirvana can break through.
move 30 - knights are strong in closed positions, Xiphos exchanged the remaining knights in the hope this slows Nirvana down. Evals continued to increase, Nirvana's eval peaked over 1 and then went back down a bit. The f5 pawn is a second weakness, the black bishop is covering it too.
move 36 - can Nirvana break through? I think it will need a pawn move to do it, needs to prepare. Instead Xiphos opened up the center itself giving a pawn. Evals went down a bit, perhaps a counter attack on the king side?
move 42 - Xiphos moved a rook to the 2nd rank, but it is two pawns down. Nirvana tried to exchange the invading rook, Xiphos avoided the exchange and brought the second rook to help and exchanged it. Evals dropping to 0, I think Xiphos found a defense.
move 48 - Nirvana had a trapped rook, it gave a pawn to open a way out for it. Xiphos exchanged the rook, won a second pawn and the game reached a QB vs QB position. The evals are at 0, waiting for the draw rule.
move 55 - perpetual check draw, game over.
Next game Xiphos is white.
move 10 - the engines are playing a different QGD variant, this time it doesn't look like the position will remain closed for long. The engines exchanged pawns and cleared the center.
Evals are slightly positive, nothing out of the ordinary so far.
move 15 - Xiphos' eval is increasing, Nirvana is behind in development, its LSB is stuck on its initial square and the rooks are not connected. Nirvana is planning to move the bishop to b7, its queen is in an awkward square on a5.
move 21 - Xiphos exchanged pieces on the king side and weakened the black pawn structure. Evals jumping higher, Xiphos gave a pawn but has a strong rook targeting the black queen.
move 24 - a nice combination, Xiphos winning a rook for a bishop, still down a pawn. Only QRB vs QBB remaining, can Nirvana hold in the long run?
move 30 - Nirvana needs to counter the white queen. The black bishops are weak, especially the DSB, while the white knight is centralized and strong.
move 36 - evals are increasing, this is looking good for Xiphos. Xiphos sacrificed its knight for a pawn, Nirvana declined (probably some trap). Instead the engines exchanged queens, evals jumping again.
move 42 - Only RN vs BB remaining, Xiphos eval over 2.5. Xiphos captured another pawn while exchanging its knight for a bishop. The R vs B ending is probably winning, we shall soon see.
move 51 - Xiphos capturing a pawn, game will be over soon by the win rule.
move 55 - a win for Xiphos, it is back in the lead with a very good chance to promote again. Still another 6 double rounds to go, a lot can happen.
I am watching game 8.5, Nirvana in white.
move 10 - so far the material is balanced and the evals close to 0. Both kings are castled and safe, Xiphos has more space on the king side, Nirvana's pieces are concentrated on the queen side.
move 16 - Xiphos is pushing pawns on both the king and the queen side, its eval turning negative. Nirvana is still happy so far, but its position is starting to look a bit cramped.
move 20 - the black king is quite exposed but there are no white pieces that can threaten it. Xiphos is moving more pieces to the king side, Nirvana looks like it is ready to create a pawn wall and start shuffling.
move 25 - Xiphos exchanged a pair of knights on the king side, its eval positive again. Nirvana's eval is increasing, it is countering on the queen side. Xiphos closed the queen side with its pawns, the c6 pawn is weak, not sure if Nirvana can break through.
move 30 - knights are strong in closed positions, Xiphos exchanged the remaining knights in the hope this slows Nirvana down. Evals continued to increase, Nirvana's eval peaked over 1 and then went back down a bit. The f5 pawn is a second weakness, the black bishop is covering it too.
move 36 - can Nirvana break through? I think it will need a pawn move to do it, needs to prepare. Instead Xiphos opened up the center itself giving a pawn. Evals went down a bit, perhaps a counter attack on the king side?
move 42 - Xiphos moved a rook to the 2nd rank, but it is two pawns down. Nirvana tried to exchange the invading rook, Xiphos avoided the exchange and brought the second rook to help and exchanged it. Evals dropping to 0, I think Xiphos found a defense.
move 48 - Nirvana had a trapped rook, it gave a pawn to open a way out for it. Xiphos exchanged the rook, won a second pawn and the game reached a QB vs QB position. The evals are at 0, waiting for the draw rule.
move 55 - perpetual check draw, game over.
Next game Xiphos is white.
move 10 - the engines are playing a different QGD variant, this time it doesn't look like the position will remain closed for long. The engines exchanged pawns and cleared the center.
Evals are slightly positive, nothing out of the ordinary so far.
move 15 - Xiphos' eval is increasing, Nirvana is behind in development, its LSB is stuck on its initial square and the rooks are not connected. Nirvana is planning to move the bishop to b7, its queen is in an awkward square on a5.
move 21 - Xiphos exchanged pieces on the king side and weakened the black pawn structure. Evals jumping higher, Xiphos gave a pawn but has a strong rook targeting the black queen.
move 24 - a nice combination, Xiphos winning a rook for a bishop, still down a pawn. Only QRB vs QBB remaining, can Nirvana hold in the long run?
move 30 - Nirvana needs to counter the white queen. The black bishops are weak, especially the DSB, while the white knight is centralized and strong.
move 36 - evals are increasing, this is looking good for Xiphos. Xiphos sacrificed its knight for a pawn, Nirvana declined (probably some trap). Instead the engines exchanged queens, evals jumping again.
move 42 - Only RN vs BB remaining, Xiphos eval over 2.5. Xiphos captured another pawn while exchanging its knight for a bishop. The R vs B ending is probably winning, we shall soon see.
move 51 - Xiphos capturing a pawn, game will be over soon by the win rule.
move 55 - a win for Xiphos, it is back in the lead with a very good chance to promote again. Still another 6 double rounds to go, a lot can happen.
Saturday, May 5, 2018
Season 12 live blogging, Saturday May 5th
Xiphos and ChessBrain promoted from division 3, so far Xiphos is leading division 2 while ChessBrain is 1 point behind in 5th place. ChessBrain had an early crash, but changed its parameters and it is hoped it will be stable for the rest of division.
move 12 - the center is very crowded. There have been no exchanges so far, evals close to 0 and slightly negative in favor of ChessBrain. Both engines are not showing a PV, very hard for me to understand where the game is going.
move 18 - the DSBs were exchanged, first pieces off the board. ChessBrain pushed a pawn on the queen side, Xiphos is worried so far.
move 23 - Xiphos pushed d4 and e5 and opened the center, ChessBrain's eval is not stable but now it is close to 0 again.
move 29 - the center is wide open now, Xiphos concentrated its major pieces on the d file, so far ChessBrain avoided exchanging rooks. Evals continue to be close to 0.
move 35 - the position looks very balanced, pawns are equal on both sides, both kings are safe. If nothing unusual happens the game will end with the draw rule quickly.
move 43 - game could have been over, ChessBrain moved a pawn and reset the draw rule counter.
move 51 - ChessBrain gave a pawn on the queen side, its eval moving a little. What is its compensation, activity in the center? Evals back down to 0.
move 61 - ChessBrain captured the white knight, Xiphos counters with an attack on the black king, perpetual check draw.
All in all, not a very interesting game, at least it was played out to the end crash-free.
move 12 - the center is very crowded. There have been no exchanges so far, evals close to 0 and slightly negative in favor of ChessBrain. Both engines are not showing a PV, very hard for me to understand where the game is going.
move 18 - the DSBs were exchanged, first pieces off the board. ChessBrain pushed a pawn on the queen side, Xiphos is worried so far.
move 23 - Xiphos pushed d4 and e5 and opened the center, ChessBrain's eval is not stable but now it is close to 0 again.
move 29 - the center is wide open now, Xiphos concentrated its major pieces on the d file, so far ChessBrain avoided exchanging rooks. Evals continue to be close to 0.
move 35 - the position looks very balanced, pawns are equal on both sides, both kings are safe. If nothing unusual happens the game will end with the draw rule quickly.
move 43 - game could have been over, ChessBrain moved a pawn and reset the draw rule counter.
move 51 - ChessBrain gave a pawn on the queen side, its eval moving a little. What is its compensation, activity in the center? Evals back down to 0.
move 61 - ChessBrain captured the white knight, Xiphos counters with an attack on the black king, perpetual check draw.
All in all, not a very interesting game, at least it was played out to the end crash-free.
Season 12 division 2 crash watch
Division 2 is running, the first 4 double rounds are over. Unfortunately it appears that this is going to be one of the most unstable stages in recent TCEC history. There were three crashes in the first double round alone. So far there have been 5 crashes in the first 34 games, with 4 engines crashing: ChessBrain, Arasan (twice), Fritz, Hannibal. ChessBrain has an additional strike since it was updated between divisions.
With 88 more games to play and the 3-strike rule in effect there is a serious concern that more than 2 engines will be disqualified and relegated, causing an administrative headache about which engines will play in the division next season. All game results of disqualified engines are removed, this makes the current standings unreliable. The whole viewing experience is negatively affected, when there's always a chance an engine will crash mid-game.
The most probable cause of these crashes is the engine updates, with the developers unable to test their engines under TCEC conditions. TCEC tests the engines before each stage but this is not sufficient to catch all problems. The fact that the GUI is also under development makes the whole TCEC experience a bit shaky at the moment. Hopefully this is going to improve from now on.
With 88 more games to play and the 3-strike rule in effect there is a serious concern that more than 2 engines will be disqualified and relegated, causing an administrative headache about which engines will play in the division next season. All game results of disqualified engines are removed, this makes the current standings unreliable. The whole viewing experience is negatively affected, when there's always a chance an engine will crash mid-game.
The most probable cause of these crashes is the engine updates, with the developers unable to test their engines under TCEC conditions. TCEC tests the engines before each stage but this is not sufficient to catch all problems. The fact that the GUI is also under development makes the whole TCEC experience a bit shaky at the moment. Hopefully this is going to improve from now on.
Wednesday, May 2, 2018
Season 12 division 3 final results
Final crosstable
ChessBrain and Xiphos promote to division 2, Senpai and Wasp are relegated to division 4.
ChessBrain dominated the division, a big improvement compared to season 11 where it finished 7th in division 3. The fact that it is written in Visual Basic is also a little bizzare. Xiphos was also a surprise, it started in division 4 as a new engine and finished second behind Ethereal. In division 3 it also finished 2nd, but Ethereal only 4th two points behind.
Pedone crashed once in the fourth double round, and there was some concern about its stability after it had two crashes in season 11. However Pedone managed to finish all of its remaining games and to be a real candidate in the promotion race.
The race for 2nd place was exciting, there were 4 candidates with a good chance of promotion in the final rounds: Xiphos, Pedone, Ethereal and Bobcat. Xiphos was ahead, but it had to play ChessBrain in the last round and it lost its head to head games to both Pedone and Ethereal, so it couldn't afford a tiebreak. In the end it was Xiphos' better scores against the lower ranked engines that gave it the advantage, and the surprising win against ChessBrain sealed its promotion.
The TCEC website is being updated, with a new GUI being tested. I'm not sure I'm going to like the new version, it still doesn't have the full functionality of the current GUI but that is work in progress, Having the two sites running it is clear that the new GUI is much faster, the current GUI lags by 10s of seconds. We'll have to get used to the new I suppose, can't stop progress.
Division 2 will start soon, probably a day or two break judging by the previous divisions. ChessBrain and Xiphos will be joined by Hannibal, Nirvana, Texel, Arasan, Fritz and Vajolet. I expect another exciting race for promotion. Can Xiphos promote a third time, or will it be relegated back to division 3?
Final 3 rounds, fight for second place:
The standings at the start of round 12 were: Xiphos 14, Pedone, Ethereal, Bobcat 12.5.
Bobcat managed 0.5/2 points from its double game against ChessBrain. The first game reached a RN vs RN position on move 26, Bobcat held as black. In the reverse ChessBrain attacked with a pawn storm on the king side, Bobcat's position collapsed and the game was quickly over.
Pedone with a big 1.5/2 over Xiphos. The first game ended in a draw. The second game reached a RN vs RB position on move 31 with evals close to 0, then Pedone in black won a pawn and reduced to a rook ending, its eval jumping over 2. Xiphos sacrificed pawns to get a passer to the 7th rank, this cost Pedone its rook. However the 4 king side passers were stronger than the remaining white rook. With this win Pedone improves its chances, closing the gap to only half a point.
Ethereal with two draws against Senpai, only 1/2 points. A missed chance for Ethereal.
After round 12 the standings were: Xiphos 14.5, Pedone 14, Ethereal 13.5, Bobcat 13
Xiphos beat Bobcat with 1.5/2. In the first game Xiphos as black had a bishop pair and a better pawn structure, it outplayed Bobcat slowly. The black king found a safe square behind a white pawn, Xiphos captured a pawn and created an unstoppable passer. The second game reached a QB vs QB position on move 31, Bobcat had a central passer and an eval advantage but Xiphos held the draw. Bobcat is out of the race.
Pedone and Ethereal needed wins to get closer to Xiphos. The first game was a 20 move miniature that ended in a perpetual check. The second game reached a RN vs RN position on move 24, Pedone used its tablebase knowledge to reduce to a RN vs R draw. A disappointing result for both engines.
Going into the final round the standings are: Xiphos 16, Pedone 15, Ethereal 14.5
Xiphos had ChessBrain as its last opponent. Xiphos as black had a pawn advantage and a central passer, ChessBrain defended well and held the draw. In the reverse game ChessBrain attacked on the king side, but was not effective. Xiphos had a central passer that had to be watched, at the right moment Xiphos started a series of exchanges that led to a Q vs RB ending. Xiphos was two pawns up and winning. A surprise result, only the 2nd loss for ChessBrain in the division, Xiphos clinches the second promotion spot.
The remaining Ethereal and Pedone games were less interesting now that the race was over. Ethereal drew twice against Bobcat. Pedone lost as black against Nemorino, and then won as white.
ChessBrain and Xiphos promote to division 2, Senpai and Wasp are relegated to division 4.
ChessBrain dominated the division, a big improvement compared to season 11 where it finished 7th in division 3. The fact that it is written in Visual Basic is also a little bizzare. Xiphos was also a surprise, it started in division 4 as a new engine and finished second behind Ethereal. In division 3 it also finished 2nd, but Ethereal only 4th two points behind.
Pedone crashed once in the fourth double round, and there was some concern about its stability after it had two crashes in season 11. However Pedone managed to finish all of its remaining games and to be a real candidate in the promotion race.
The race for 2nd place was exciting, there were 4 candidates with a good chance of promotion in the final rounds: Xiphos, Pedone, Ethereal and Bobcat. Xiphos was ahead, but it had to play ChessBrain in the last round and it lost its head to head games to both Pedone and Ethereal, so it couldn't afford a tiebreak. In the end it was Xiphos' better scores against the lower ranked engines that gave it the advantage, and the surprising win against ChessBrain sealed its promotion.
The TCEC website is being updated, with a new GUI being tested. I'm not sure I'm going to like the new version, it still doesn't have the full functionality of the current GUI but that is work in progress, Having the two sites running it is clear that the new GUI is much faster, the current GUI lags by 10s of seconds. We'll have to get used to the new I suppose, can't stop progress.
Division 2 will start soon, probably a day or two break judging by the previous divisions. ChessBrain and Xiphos will be joined by Hannibal, Nirvana, Texel, Arasan, Fritz and Vajolet. I expect another exciting race for promotion. Can Xiphos promote a third time, or will it be relegated back to division 3?
Final 3 rounds, fight for second place:
The standings at the start of round 12 were: Xiphos 14, Pedone, Ethereal, Bobcat 12.5.
Bobcat managed 0.5/2 points from its double game against ChessBrain. The first game reached a RN vs RN position on move 26, Bobcat held as black. In the reverse ChessBrain attacked with a pawn storm on the king side, Bobcat's position collapsed and the game was quickly over.
Pedone with a big 1.5/2 over Xiphos. The first game ended in a draw. The second game reached a RN vs RB position on move 31 with evals close to 0, then Pedone in black won a pawn and reduced to a rook ending, its eval jumping over 2. Xiphos sacrificed pawns to get a passer to the 7th rank, this cost Pedone its rook. However the 4 king side passers were stronger than the remaining white rook. With this win Pedone improves its chances, closing the gap to only half a point.
Ethereal with two draws against Senpai, only 1/2 points. A missed chance for Ethereal.
After round 12 the standings were: Xiphos 14.5, Pedone 14, Ethereal 13.5, Bobcat 13
Xiphos beat Bobcat with 1.5/2. In the first game Xiphos as black had a bishop pair and a better pawn structure, it outplayed Bobcat slowly. The black king found a safe square behind a white pawn, Xiphos captured a pawn and created an unstoppable passer. The second game reached a QB vs QB position on move 31, Bobcat had a central passer and an eval advantage but Xiphos held the draw. Bobcat is out of the race.
Pedone and Ethereal needed wins to get closer to Xiphos. The first game was a 20 move miniature that ended in a perpetual check. The second game reached a RN vs RN position on move 24, Pedone used its tablebase knowledge to reduce to a RN vs R draw. A disappointing result for both engines.
Going into the final round the standings are: Xiphos 16, Pedone 15, Ethereal 14.5
Xiphos had ChessBrain as its last opponent. Xiphos as black had a pawn advantage and a central passer, ChessBrain defended well and held the draw. In the reverse game ChessBrain attacked on the king side, but was not effective. Xiphos had a central passer that had to be watched, at the right moment Xiphos started a series of exchanges that led to a Q vs RB ending. Xiphos was two pawns up and winning. A surprise result, only the 2nd loss for ChessBrain in the division, Xiphos clinches the second promotion spot.
The remaining Ethereal and Pedone games were less interesting now that the race was over. Ethereal drew twice against Bobcat. Pedone lost as black against Nemorino, and then won as white.
Season 12 division 3 statistics
Draw rate, wins
Final draw rate was 54.5%. The draw rate returned to normal after the very low value in division 4.
Game termination
The three most common game termination causes were:
42.9% - TCEC win rule
28.6% - TCEC draw rule
17.9% - SyzygyTB (Syzygy tables are used in the new TCEC, still 5-man adjudication only)
Moves per game
Median= 58
Average= 62.9
Time per game (hours)
Median= 1:12
Average= 1:13
Openings
The opening section is missing, due to changes in TCEC the division 3 openings were not reported correctly.
Reverse pairs, wins
Reverse pairs, same moves
Pairs of reverse games diverged very quickly, 85.7% of the pairs diverged at most after 1 move.
Final draw rate was 54.5%. The draw rate returned to normal after the very low value in division 4.
Game termination
The three most common game termination causes were:
42.9% - TCEC win rule
28.6% - TCEC draw rule
17.9% - SyzygyTB (Syzygy tables are used in the new TCEC, still 5-man adjudication only)
Moves per game
Median= 58
Average= 62.9
Time per game (hours)
Median= 1:12
Average= 1:13
Openings
The opening section is missing, due to changes in TCEC the division 3 openings were not reported correctly.
Reverse pairs, wins
Reverse pairs, same moves
Pairs of reverse games diverged very quickly, 85.7% of the pairs diverged at most after 1 move.
Subscribe to:
Posts (Atom)