Season 15 division 2 statistics

A summary statistics table of previous stages and seasons. 

Draw rate, wins


Final draw rate was 53.6%.


Game termination

The three most common game termination causes were:
38.4% - TCEC win rule
19.6% - SyzygyTB
19.6% - TCEC draw rule

Ginkgo crashed twice in the division, but it was reduced to 16 threads and survived.

Moves per game


Median= 61
Average= 72.7

There were 16 games longer than 100 moves, the longest was 244 moves (Game 85, AllieStein - Fritz, draw).

Time per game (hours)


Median= 1:05
Average= 1:04

Openings

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


The engines did not have a lot of freedom to choose the opening variant. 78.6% of the game pairs repeated the same ECO code twice, and 76.8% repeated the same opening variant. Only in 1.8% of the game pairs (one game pair) the ECO first letter was not repeated twice.

Reverse pairs, wins


Reverse pairs, same moves 

Pairs of reverse games diverged very quickly, 37.5% diverged immediately out of book, 73.2% of the pairs diverged at most after 1 move. There were many long sequences of repeated moves, the longest was 25 (!) plys (Ginkgo - rofChade, games 58 and 86, QGD Chigorin defense, Ginkgo won as white).

Season 15 division 2 results

Final crosstable

Xiphos and AllieStein promote to division 1. These two engines dominated division 2 and there was no doubt they would be the top 2. Xiphos won the division despite losing twice to AllieStein. AllieStein continues to improve and move up, still it lost two games to rofChade, blundering in a rook endgame in one of them.

Nirvana and Fritz are relegated to division 3. Nirvana was very weak in this division. The bottom half of the table was more competitive and the second engine to relegate wasn't determined until the last game. Ginkgo crashed twice and was in danger of being disqualified, but it managed to survive after it was reduced to 16 threads.

The engines to play in division 1 are: Andscacs, KomodoMCTS, Fizbo, Chiron, Laser, Jonny, Xiphos and AllieStein. The time control will be 60+5, twice as long as the lower divisions.

Action in the last RR

After RR3: Xiphos +11, AllieStein +8, Booot +2, ChessBrain and Ginkgo +0, Fritz -3, rofChade -5, Nirvana -13. The top is almost surely decided, Nirvana will be last. In the bottom rofChade still has a chance to overtake Fritz, and if Ginkgo crashes it will be disqualified. 

In rounds 22-23 Xiphos beat ChessBrain and increased the gap from third place to 5 points. Fritz survived against AllieStein in a long endgame, AllieStein's eval was over 14 but it couldn't convert. Ginkgo beat Fritz, Fritz is only 0.5 points away from the relegation zone.

5 rounds to go: Xiphos +12, AllieStein +8, Booot +2, Ginkgo +1, ChessBrain -1, Fritz -4, rofChade -5, Nirvana -13.

In rounds 24-25 AllieStein beat Ginkgo, Ginkgo is still one crash from disqualification. Xiphos beat Fritz and rofChade beat Booot, rofChade and Fritz change places and rofChade is in 6th place. Then Fritz beat Booot, Fritz and rofChade are tied.

3 rounds to go: Xiphos +13, AllieStein +9, Booot and Ginkgo +0, ChessBrain -1, Fritz and rofChade -4, Nirvana -13. Xiphos and AllieStein promote. The last game of the division will be rofChade - Fritz, it may determine which engine will finish in 7th place.

In round 26 rofChade and Fritz drew their games. AllieStein beat Xiphos and Ginkgo beat Booot without crashing. In round 27 rofChade drew AllieStein while Fritz lost to ChessBrain, so rofChade only needed a draw in the last game to avoid relegation while Fritz needed a win. Ginkgo beat Nirvana without crashing, Booot beat Xiphos. In the last round Ginkgo survived its last game without crashing, and rofChade held a draw against Fritz, so that Fritz is relegated together with Nirvana.

Season 15 division 3 statistics

A summary statistics table of previous stages and seasons. 

Draw rate, wins


Final draw rate was 54.5%.


Game termination

The three most common game termination causes were:
40.2% - TCEC win rule
22.3% - SyzygyTB
18.8% - 3-Fold repetition

There were no crashes in the division.

Moves per game


Median= 66
Average= 74.8

There were 19 games longer than 100 moves, the longest was 198 moves (Game 90, Nemorino - Texel, Texel won).

Time per game (hours)


Median= 1:06
Average= 1:05

Openings

There were 2-move book openings in this stage chosen by Cato. The first letter of the ECO codes was distributed as follows:


The engines had a lot of freedom to choose the opening variant. Yet 50% of the game pairs repeated the same ECO code twice, and 50% repeated the same opening variant. In 17.9% 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, 42.9% diverged immediately out of book, 80.4% of the pairs diverged at most after 1 move. The longest repeated sequence of moves was only 5 plys (rofChade - Arasan, games 72 and 100, English symmetrical variation, rofChade won as black).

Season 15 division 3 results

Final crosstable


rofChade and AllieStein promote to division 2. AllieStein continues to move up the divisions in season 15. It played with a known bug this division, in which it did not look at lines where the opponent castles. This bug is now fixed for division 2. rofChade started season 14 in division 4, and this season it won division 3.

Pedone and Nemorino are relegated to division 4. This is the second season in a row that Nemorino starts in division 4, then promotes to division 3 and then returns to division 4.

Division 3 was very tight, there were no engines that were clearly better or worse than the others. All engines finished with at least 3 wins and 2 losses and it took a while for the rankings to stablilize. The fight against relegation continued until the very last game. 

The engines that will play in division 2: Fritz, Ginkgo, Xiphos, Booot, Nirvana, ChessBrain, rofChade, AllieStein.

Action in the last RR

After RR3: rofChade +4, AllieStein +3, Arasan +2, Gull 0, Nemorino, Vajolet and Pedone -2, Texel -3. rofChade, AllieStein and Arasan have a good chance of promotion, Nemorino, Vajolet, Pedone and Texel are fighting against relegation.

In rounds 22-23 Pedone lost twice to rofChade and AllieStein, while Texel won twice against Arasan and Nemorino. Arasan also beat Vajolet.  

5 rounds to go: rofChade +5, AllieStein +4, Arasan +2, Gull 0, Texel -1, Nemorino and Vajolet -3, Pedone -4. Arasan is losing touch with the leaders, Texel improved its position. 

In rounds 24-25 Nemorino lost twice against Vajolet and Pedone, its rivals in the bottom of the table. Texel beat Gull and lost to AllieStein, it remains in the middle of the table together with Gull. rofChade beat Arasan in a game where Arasan sacrificed pieces to get a pawn on the 7th rank, but the gamble failed. Arasan is 2 points behind the leaders.

3 rounds to go:  rofChade +6, AllieStein +5, Arasan +1, Gull and Texel -1, Vajolet -2, Pedone -3, Nemorino -5. Arasan is probably not going to catch the leaders, rofChade and AllieStein will probably promote. The bottom is still tight, Pedone has a good chance of escaping relegation since its remaining games are against lower ranked opponents. Nemorino is almost surely going to be last. The last game of the division is Pedone-Vajolet in a King's Gambit, it might determine which engine relegates in the end.

There were 4 draws in round 26. In round 27 Texel beat Pedone, and as a result Pedone could only escape relegation if it beat Vajolet in the last game. rofChade beat Gull in black, Gull and Vajolet were tied with score -2 one point away from Pedone. However, Gull was safe due to better head to head results against Pedone and Vajolet. Arasan beat Nemorino, too late to catch up with the leaders. In the last round Arasan beat Gull, setting up  a possible three-way tie with Vajolet and Pedone. rofChade beat Texel, with no impact on the table. The last game turned out to be uninteresting, Vajolet had a small eval advantage, and Pedone didn't come close to winning. All the tiebreak speculation became irrelevent and Pedone joined Nemorino at the bottom of the table.

Season 15 division 3 live blog, Tuesday Mar 19th

Division 3 is being played, currently in the third RR out of 4. This is a very close division, only 1.5 points seperate 1st and 7th places after 18 rounds. All the engines have won and lost at least twice already and anything can happen.

We join the Arasan - AllieStein game at the start of round 19. These two engines are the current joint leaders with a +2 score.

move 12: The opening is a Petrov, on move 7 AllieStein chose a sideline that was played by Kasparov and Karpov in the 1980s. Its eval jumped over 1.5, perhaps related to the known "castling bug" as Arasan castled right before that. After a few exchanges AllieStein's eval came back down. On move 12 queens were exchanged, this could be a short draw.

move 19: more exchanges and only RRN vs RRB remain on move 17. Evals are close to 0, black's pawn structure seems slightly worse but I doubt it matters. This will be a draw unless one of the engines blunders badly.

move 28: Arasan offered a pawn to exchange the knight and bishop, AllieStein refused. The white knight then blocked both black rooks on the back rank. The center pawns were isolated and were removed. Evals climbed over 0.3 for a while, though I don't see any real advantage for white.

move 41: AllieStein pushed pawns on the king side, evals came down to 0. Arasan was expecting a rook trade, AllieStein had another plan and its eval became negative. Arasan pushed a pawn on the queen side, AllieStein didn't take to keep its king safe from checks. The engines traded pawns on both sides, AllieStein went a pawn up but its king was exposed.

move 51: Arasan wants to finish the game with a repetition, AllieStein keeps playing. The engines exchanged down to a rook ending and reached a tablebase draw.

Nothing really happened in this game, both Arasan and AllieStein continue to lead the table but now the other engines have a chance to get closer to the top with a win.

Season 15 division 4b and playoffs results

Final crosstable 4b group


pirarucu and AllieStein will go the division 4 playoffs, Wasp and Marvin will participate in the cup event later this season.

pirarucu started strong with 8 wins in the first 9 games, it held on to first place despite a weaker second half. AllieStein and Wasp were better in the second half of the division, Wasp crashed in a critical game against AllieStein and this meant AllieStein had the tiebreak advantage. In the end there was no need for a tiebreak.

There were two crashes in this division, in both cases the crashing engine was not new to TCEC (Wasp and chess22k). Of the 5 new engines Jumbo and Bagatur were weak, AllieStein, Marvin and Monollith were better. There were two NN based engines in the division, ScorpioNN and AllieStein. The network crashed in one game (chess22k - AllieStein) and due to a new TCEC rule the game was not replayed but resumed, with added time to compensate for the lost information.

Final crosstable playoffs


AllieStein and Nemorino progress to division 3.

AllieStein dominated the playoffs with 8/12 wins and one loss. It beat Nemorino in all their 4 games and had a positive score against the other engines as well. This performance was surprising, in division 4b it struggled to qualify for the playoffs. This was explained in part by the worse performance of pirarucu. After finishing first in division 4b pirarucu crashed twice in the playoffs and was reduced to 16 threads. Without the crashes pirarucu could have been the second qualifier.

AllieStein and Nemorino join Gull, Texel, Arasan, Vajolet, Pedone and rofChade in division 3. Can AllieStein continue to climb, a second NN engine in the higher TCEC divisions? I don't think so, some of the other engines have new versions and Gull should also be tough to beat despite not being developed anymore. We shall soon see...

Season 15 division 4b statistics

A summary statistics table of previous stages and seasons. 

Draw rate, wins


Final draw rate was 28.9%. This is higher than in the 4a group but still very low for TCEC. 


Game termination

The three most common game termination causes were:
45.6% - TCEC win rule
28.9% - SyzygyTB
11.1% - TCEC draw rule

There were 2 crashes in the stage, as well as one network failure that stopped a game - which was completed later.

Moves per game


Median= 61.3
Average= 67.4

There were 11 games longer than 100 moves, the longest was 188 moves (Monolith - ScorpioNN, game 60, ScorpioNN won).

Time per game (hours)


Median= 1:01
Average= 1:02

Openings

There were 2-move book openings in this stage chosen by Cato. The first letter of the ECO codes was distributed as follows:


The engines had a lot of freedom to choose the opening variant. Only 53.3% of the game pairs repeated the same ECO code twice, and 46.7% repeated the same opening variant. In 8.9% 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, 44.4% diverged immediately out of book, 84.4% of the pairs diverged at most after 1 move. The longest repeated sequence of moves was only 5 plys (chess22k - ScorpioNN, games 39 and 84, King's pawn Colorado counter, two draws ; AllieStein - Gavioya, games 41 and 86, Old Benoni defence, AllieStein won as black)

Season 15 division 4a results

Final crosstable



The number of participating engines continues to grow, this season there are 20 candidate engines in division 4, in season 14 there were 12. As a result division 4 is divided into two, 4a and 4b, each group plays a single DRR. The top two engines from each group will play a 2 DRR playoff to decide the two engines that will promote to division 3. The top 4 engines from each group will participate in the cup event later in the season.

The two qualifiers from division 4a are Nemorino and RubiChess. Nemorino was relegated from division 3 and was the clear winner of this group with 13/18 wins and one loss. RubiChess is new to TCEC, it qualified through a tiebreak with Winter due to 2 head-to-head wins. Winter and Rodent will play in the cup event. There were 4 more engines in this group that are new to TCEC: Cheese, Minic, Topple and Igel. These engines ranked last in this group. It should be noted that there were no crashes, unusual considering there were so many new engines. Well done to the authors and administrators.

The TCEC website continues to evolve and improve. The new archive site has the same look as the live games site, you can see the new design in the table above. There is a new administrator team this season, making sure the games run smoothly, fixing and updating the websites and keeping the public informed through the chat. Well done and keep up the good work!

Season 15 division 4a statistics

A summary statistics table of previous stages and seasons. 

Draw rate, wins

 
Final draw rate was 21.1%. This is very low for TCEC, the lowest draw rate in a stage at least since season 7.


Game termination

The three most common game termination causes were:
65.6% - TCEC win rule
18.9% - SyzygyTB
8.9% - TCEC draw rule

There were no crashes in the stage, quite unusual for a stage with several engines new to TCEC. Well done to the authors and the organizers!

Moves per game


Median= 59.8
Average= 64.9

There were 8 games longer than 100 moves, the longest was 135 moves (Topple - Minic, game 54, Topple won).

Time per game (hours)


Median= 1:03
Average= 1:03

Openings

There were 2-move book openings in this stage chosen by Cato. The first letter of the ECO codes was distributed as follows:


The engines had a lot of freedom to choose the opening variant. Yet 64.4% of the game pairs repeated the same ECO code twice, and 62.2% repeated the same opening variant. In 15.6% 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, 46.7% diverged immediately out of book, 77.8% of the pairs diverged at most after 1 move. The longest repeated sequence of moves was 11 plys (Rodent - Cheese, games 4 and 49, Sicilian Alapin variation, two wins for white)

S14 bonus match, Leela - Stockfish, bookless openings

Leela beat Stockfish in a 100 rapid game match with no book, the end result was 16-4 with 80 draws. Leela learned a lot about openings in its training, and it appears to be better than Stockfish from the initial position, at least in shorter time controls. The match also shows that when the engines are allowed to choose the openings we get far less variety than what we usually see in TCEC.

In the 50 games Leela was white it always started with e4, Stockfish played either c5 (Sicilian) or e6 (French). Most of the Sicilians were a Najdorf, English attack variation, 30 games (60%) started with the same 15-ply opening. The French Defense games were mostly Steinitz Boleslavsky variations, 16 games (32%) had the same 13-ply opening. After 20 plys there were only 12 different sequences with information content ( = distribution entropy) equivalent to a space of size 6.4.

Stockfish played e4 in 76% of its white games, it started 24% with d4. Leela answered e4 with e5, leading to many Ruy-Lopez games and a few Italian games. 25 games (50%) started with an 16-ply Berlin Defense line, 11 games (22%) were Giuoco Pianissimo and more variable. Leela answered d4 with Nf6, and this led to QGD games, 7 games (14%) started with a 16-ply sequence in the Vienna variation. After 16 plys there were 19 different sequences with information content ( = distribution entropy) equivalent to a space of size 7.9.

Leela is more stubborn than Stockfish in its opening moves. All the major opening classes in the first few moves - Sicilian, French, Ruy Lopez, Italian, QGD - were determined by Stockfish either in black or in white. Stockfish expanded the opening tree almost every move in at least one opening line, again in both white and black. Leela, on the other hand, almost always played the same move when presented with the same position. In the first 24 plys of 100 games I found only 3 cases where Leela expanded the opening tree, on move 7 as black in QGD (game 29), on move 10 as black in Ruy Lopez (games 45 vs 97), and on move 10 as white in French (game 84).

As white Leela won 6/30 in the Sicilian Najdorf, and lost 2/30. In the French Defense Leela won 5/17 games and lost none. As black in the Italian games Leela won 4/11 and lost 1/11, not a good choice for Stockfish. In the Ruy Lopez Leela won 1/27, the rest were draws. Stockfish won 1/8 in the QGD.