Stockfish is the winner of the season 22 superfinal. The final score is 28-9 with 63 draws. In the last 10 games there was no change in the +19 margin. There were 4 game pairs with two draws, in the final game pair KomodoDragon managed to win as white for the 9th time in the match, and Stockfish won the reverse game as it had done in the previous 8 times this happened.
Stockfish was clearly better than KomodoDragon in the match, it took the lead right from the start and steadily increased the gap until the end. The remaining questions were whether Stockfish would break the highest gap record (it didn't, the record is +20 from season 12, also Stockfish - Komodo) and whether KomodoDragon would be able to win a game pair (failed with 9 wins, first TCEC that the winner does not lose a game pair). In the bookmaker competition of which lines led to a decisive game pair the final score was Jeroen Noomen (even pairs) 11 - 8 GM Matthew Sadler (odd pairs).
A total of 48 engines participated in season 22. The lower leagues were 1 DRR events with 12-14 engines and about one hour per game. Berserk started the season in the qualification league, and it was expected to advance since it had a good performance in the Swiss event of season 21. Indeed it had very good results in the qualification league and league 4 and it managed to advance twice. League 3 was a different story, first since the competition was harder, but also due to the promotion bottleneck.
Several engines that played in higher leagues during season 21 did not participate in season 22 - Xiphos and AllieStein were the most prominent of these. As a result the season 21 relegation was disrupted and there were several weaker engines that started in higher leagues in season 22. From league 3 and on there were only 2 promotion spots per league and this created a bottleneck, judging by their strength there were more engines that "deserved" to promote than there were spots available.
In league 3 Berserk was one of the 3 leaders but it only managed 3rd place behind Koivisto and Seer. Koivisto also performed well in the season 21 Swiss event and was expected to be a strong competitor, Seer was more of a surprise candidate. League 2 was a 2DRR event with 8 engines, Koivisto won the league convincingly. Seer couldn't repeat its success, it was in the leader group but only came 3rd behind Minic. League 1 was a 2DRR event with slightly longer games. Koivisto led from the start and it seemed to be heading to the premier division. However in the last RR it crashed once and lost two more games, and as a result it dropped to 3rd place behind rofChade and SlowChess. Another surprise was that Ethereal was only 5th in league 1, after appearing in the premier division every season since season 13 except once.
The premier division was a 4DRR event with 2-hour games. From the start it was clear that Stockfish, KomodoDragon and Leela were in a competition for the superfinal, while the other 5 engines were only trying to avoid relegation. In 60 game pairs between the top3 and bottom5 engines the top3 engines won 48 and lost none. Within the top3 all the games were drawn except one game Stockfish won against KomodoDragon. However, of the 3 leaders it was Leela that failed to win game pairs against the weaker engines the most. It lagged behind the other two leaders and the gap slowly increased, in the end Leela was 3rd and 3 points behind 2nd place. Stockfish won the premier division and KomodoDragon was 0.5 points behind. This was quite an upset, Leela hadn't missed a superfinal since season 14 except once, and the last time KomodoDragon was in the superfinal (then just Komodo) was season 13.
This season there was a match between 3rd and 4th places of the premier division (infrafinal = infi), to determine 3rd place (??). Leela won the 50-game match against rofChade easily. After the superfinal another match is played between the winner of the infrafinal and the loser of the superfinal (subfinal = subfi). This match between Leela and KomodoDragon is still ongoing and it is less one sided, it appears Leela is going to win. I don't know if these events will return in the future, it looks like a way of compensating Leela fans for its absence in the superfinal, to show the world that it is still better than KomodoDragon.
There are many more events and bonuses planned for season 22, including a cup, a Swiss tournament, an FRC (chess 960) event. There is a 1-day per move event that already started, Stockfish and Leela play a game pair with time control of 12 hours per move. So far 1 move has been completed, the event will be resumed later and played in the background - not sure what that means. This is an intriguing and totally unwatchable event. We can also look forward to various bonus events (like VSOB - viewer submitted openings) before season 23 starts.
A word about Cato (aka Nelson Hernandez). As I mentioned in the premier division report, Cato has retired from TCEC and will not provide opening books in the future as he had done since season 6 (!!). Cato is a TCEC icon, his database (Catobase) of chess games is legendary, he is one of the founding fathers of TCEC as we know it now. Cato's Fault is universal, and I can safely say that he is to blame that I am still watching and reporting about TCEC after so many seasons. His TCEC openings are a big part of why TCEC is a fun event to watch, and I always look forward to his chat appearances, either for chess or general knowledge or for a healthy smile.
The Covid pandemic is still with us, two and a half years and 6 TCEC seasons so far. The situation is getting better though, the Omicron variant spread very fast but caused a less severe disease. In many countries there are less limitations on public gatherings and masks are no longer mandatory. It is not over yet but there is hope that life can return to a more normal mode.
During this season Russia invaded Ukraine and war broke out, it is still going on 2 months later. The destruction and human suffering is enormous, many lives have been lost in battle, millions have fled their homes and are refugees in their country and outside in neighboring countries. I have no illusion that there can be a happy end to this conflict, but I do hope there is an end and soon.
That's it for me this season, hope to be here for the next one.
Go TCEC!!
================================
Games 91-92 started with a 12-ply book in the Caro Kann, Bronstein-Larsen variation, played in many high level human games 30-40 years ago. Game 91 left theory very quickly. There were no pawn exchanges after the start. The engines exchanged a pair of bishops, other than that they played quiet moves mostly behind the pawn lines. Evals came slowly down, on move 43 there was a short series of exchanges that left a RRB vs RRN position. Evals were low enough for the draw rule. In game 92 the engines followed human theory for a few moves, they followed a PV agreement and exchanged queens. Stockfish moved its king forward without castling and KomodoDragon castled long. There were many pawn moves, Stockfish created a passer in the center and a pawn exchange opened a central file. Evals came down, another pawn exchange opened a file on the queen side while Stockfish created a passer on the king side. Stockfish gave a pawn and moved a rook forward, this gave KomodoDragon a passer that moved to the 2nd rank. All the threats canceled out, a series of exchanges reduced to a RB vs R position with KomodoDragon 2 pawns up, the game was quickly adjudicated.
Games 93-94 started with an 8-ply book in the Alekhine's defense, modern Schmid variation, that was played in a Carlsen - Nakamura game. In game 93 after the pieces were developed there were a few exchanges, then KomodoDragon gave a rook for a bishop and weakened the black pawn structure. The engines cleared the queen side pawns, KomodoDragon went a pawn up and created a passer. Evals came down to 0, KomodoDragon pushed the passer forward and eventually regained material. The game reached a QB vs QB ending and was adjudicated. In game 94 the engines opened a file in the center, then played a PV agreement where a pair of rooks was exchanged and Stockfish went a pawn up. Stockfish gave two pawns and created a queen side passer, after exchanging queens the passer moved to the 7th rank. In a RBN vs RBB position evals came slowly down, KomodoDragon blocked the white passer and the engines exchanged pawns on the king side. The game was adjudicated after a minor piece exchange.
Games 95-96 started with an 11-ply book in the Vienna Mengarini variation, with an advantage for black. In both games black captured a pawn and forced the white king to move. In game 95 Stockfish traded both its bishops for knights, and its king didn't castle as well. The engines exchanged queens and (negative) evals came down. The exchanges continued, the game was adjudicated in a RB vs RN position. In game 96 KomodoDragon castled short and gave the pawn back. The engines exchanged pieces and only QRB vs QRB remained on move 26. The engines mostly shuffled and (negative) evals came down, eventually the game reached a same color bishop ending. KomodoDragon was a pawn up and its eval was too high for the draw rule, the game was adjudicated 40 moves later.
Games 97-98 started with a 24-ply book in the Sicilian Richter-Rauzer variation that appeared in a Mikhail Tal game from 1973. One pair of pawns and one pair of knights were exchanged, the engines castled in opposite directions. The games then repeated another 40 plys, at the end of which only rooks and bishops remained on the board, evals were around 0.5. In game 97 evals dropped to 0 after some shuffling, the engines continued to exchange pieces and pawns and the game was adjudicated in a king and pawns endgame. In game 98 the engines quickly reduced to a drawn rook ending and the game was adjudicated.
Games 99-100 started with a 16-ply book in the Sicilian Scheveningen, Tal variation, played in high level human games. In both games white castled long immediately after book, the opposite direction of the black castling. In game 99 KomodoDragon pushed pawns on the king side and Stockfish pushed pawns on the queen side. Evals started to increase after a pawn exchange on the king side, KomodoDragon attacked the black king. Stockfish avoided mate and exchanged pieces, the game reached a RBB vs RBB position with white a pawn up.
The engines continued to trade pawns and exchanged a pair of bishops. KomodoDragon had two connected passers on the queen side, much stronger than the passer Stockfish had on the king side. Stockfish lost the bishop for a passer and the game ended in a tablebase win.
In game 100 the pawns mostly stayed back, the engines opened the center and exchanged pieces. Queens were off early, by move 27 only RRB vs RRB remained. Evals were around 1.5, Stockfish had a passer that KomodoDragon covered with its rooks, other than that the advantage wasn't obvious. However Stockfish's eval slowly increased, and on move 31 both evals jumped.
The engines reduced to a rook ending and predicted mate in their PVs. KomodoDragon lost a rook for a passer, it was able to queen a pawn but too late to stop mate. There were two white wins in this game pair, the score is 28-9.
No comments:
Post a Comment