Isn't Bd7 for black after BC4 stronger. Also what is this KF8 and then KG8. Seemed strange for the king to go back and forth.
Testing a line for soundness in Sicilian Dragon, Soltis Variation using engine matches
Sort:
Feb 1, 2022
0
#3
benonidoni wrote:
Isn't Bd7 for black after BC4 stronger. Also what is this KF8 and then KG8. Seemed strange for the king to go back and forth.
In the Game Review analysis tab, 9... Bd7 gets a cloud Stockfish evaluation of 0.51 while 9... Qa5 gets an evaluation of 0.69 so the two moves aren't far apart. I thought 9... Qa5 and 10... Bd7 could be transposed, but I could be wrong. on this. 9... Bd7 is a chess.com book move, and 9.... Qa5 isn't. The organizers at TCEC are involved in preparing the opening book for the TCEC superfinals. The TCEC website with an archive of games is at https://tcec-chess.com/ .
About the king moves, I have no idea what Stockfish on high-end hardware at long time control was calculating to produce Kf8 and Kg8.
Feb 2, 2022
0
#4
bernier96 wrote:
So, in the line above, Black's move 13... h5 looks suspect as regards quality. To gther more data, Istarted an engine match Stockfish vs Stockfish which always starts from a custom position:
CCRL (a website which publishes rating lists of chess engines) has chess opening statistics by ECO code at https://ccrl.chessdom.com/ccrl/4040/opening_report_by_eco.html for 40/40 games and https://ccrl.chessdom.com/ccrl/404/opening_report_by_eco.html for 40/4 games. The latter shows a White score of 64.3% for the B79 Sicilian ( Sicilian, dragon, Yugoslav attack, 12.h4). The engine match from the custom position in that post is now at +27 =90 -1 for a White score of 61%; it seems White can expect to win or draw.
I don't know if 9... Qa5 10... Bd7 makes a difference compared to 9... Bd7 10...Qa5, however the TCEC Season 17 opening book for the superfinal was prepared by Jeroen Noomen (2199 FIDE). In an article on the superfinal, GM Matthew Sadler mentions the Sicilian Dragon games, nos. 35 and 36, without commenting on the move order in game 36 ( https://tcec-chess.com/articles/Sufi_17_-_Sadler.pdf ) .
Feb 3, 2022
0
#5
bernier96 wrote:
So, in the line above, Black's move 13... h5 looks suspect as regards quality. To gather more data, I started an engine match Stockfish vs Stockfish which always starts from a custom position:
Another simpler way to have an engine play itself from a custom position is with the Fritz feature "shootout mode" ( https://en.chessbase.com/post/the-fritz11-shootout-mode ), under the analysis tab. For example, one can take the position before the losing mistake 130... Qe6? in Carlsen-Nepomniachtchi WCC Game 6 :
At this point, it's a theoretical draw. In the shootout options, I selected Fixed Depth, and depths 3-25 skipping the even numbers, for a total of 12 games. Fritz ten has the engine play itself for the requested 12 games, and the results were:
Shoot out:
----- Ply = 3 -----
Fritz 16 ½-½
----- Ply = 5 -----
Fritz 16 ½-½
----- Ply = 7 -----
Fritz 16 1-0
----- Ply = 9 -----
Fritz 16 ½-½
----- Ply = 11 -----
Fritz 16 1-0
----- Ply = 13 -----
Fritz 16 ½-½
----- Ply = 15 -----
Fritz 16 1-0
----- Ply = 17 -----
Fritz 16 1-0
----- Ply = 19 -----
Fritz 16 ½-½
----- Ply = 21 -----
Fritz 16 ½-½
----- Ply = 23 -----
Fritz 16 ½-½
----- Ply = 25 -----
Fritz 16 ½-½
Score White: 66%, 12 Games
Done
So for 19, 21, 23 ,25 ply or half-moves, the game ended in a draw, while even 17-ply ended in a White win. It would be nice to be able to do, e.g. Fixed Depth 17 for 1000 games, but the shootout menu doesn't offer this ( http://help.chessbase.com/Fritz/16/Eng/index.html?000090.htm ) .
Feb 3, 2022
0
#6
Fritz also has a position analysis feature called Monte Carlo analysis : http://help.chessbase.com/Fritz/15/Eng/index.html?monte_carlo.htm . I can run a Monte Carlo analysis on a given chess position and navigate through the new game tree as it's being built, but I don't know how to use the game tree after the Monte Carlo analysis session is over. It would be nice if Chessbase could show more examples so we could know how to get the most out of this feature.
The TCEC superfinals imposes slightly unbalanced openings to prevent all draws.
9...Qa5? is inferior to 9...Bd7: 9...Qa5 commits the queen too soon, while 9...Bd7 connects and thus activates the rooks.
10 minute games do not come near the truth, you should give the engine more time, or look for some ICCF games.
13...h5 is perfectly natural: black must block a white 14 h5. 9...Qa5? is to blame.
Feb 3, 2022
0
#8
pfren wrote:
Sorry to say, but this is ***NOT*** mainline Soltis. far from it.
According to chess.com's description , I think I got the move order mixed up:
https://www.chess.com/openings/Sicilian-Defense-Open-Dragon-Main-Line-Soltis-Variation
This is a suspect sideline for Black, which anyway may be defensible after 14...Kf8 (which was played by King Loek). Still it's clear that Black has problems.
The idea of using an engine to evaluate a tabiya where thousands of OTB and CC games have been played , is plain stupid, and won't get you anywhere.
With respect to the position after 14. Bg5, if one looks at chess.com's database of master games, there's 22 games. My view is that generating more game data with engine-engine matches is worth looking into.
9...Bd7 is by far the most common move, and for a very good reason. 9...Qa5 used to be the mainline some 40 years ago, but now we do know that it is inferior.
I'm inclined to accept this, coming from a titled player. After 9. Bc4, the scores of these moves at chess.com is close. I'd like to know how the inferiority of 9...Qa5 was discovered.
Wikipedia mentions an old main line with a move 10... Qa5 (Chris Ward): https://en.wikipedia.org/wiki/Sicilian_Defence,_Dragon_Variation,_Yugoslav_Attack,_9.Bc4#Black_...Qa5_lines
but no move 9... Qa5.
I played against the bot Komodo25/Maximum, doing a lot of memorization of Stockfish lines, and that got me interested in opening theory. In TCEC Season 17 there was a Dragon game of Leela vs. Stockfish where Leela won (game 36), and I went looking for mistakes by Stockfish, with unclear conclusions [difficult problem]. But I did find an opening line in Game 36 where the win/draw/loss score for White is impressive (about 85% score). This line is named by cheess.com Sicilian Defense: Open, Dragon, Main Line, Soltis Variation and consists of the moves:
1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 g6 6. Be3 Bg7 7. f3 O-O 8. Qd2
Nc6 9. Bc4 Qa5 10. O-O-O Bd7 11. Bb3 Rfc8 12. h4 Ne5 13. Kb1 h5 14. Bg5 . In the openings tab of the Game Review, we find win/draw/loss statistics of 80% White wins, 9% draws and 11% Black wins This gives White a score of 0.85 . Two Master games with this line are : Karjakin vs van Wely, 2016.01.17 and Naroditsky vs Nakamura 2015.04.05.
So, in the line above, Black's move 13... h5 looks suspect as regards quality. To gther more data, Istarted an engine match Stockfish vs Stockfish which always starts from a custom position:
Currently, the match of 10 minute games has White with 13 wins, 2 draws which agrees quite well with the game explorer statistics.
An engine match starting from a custom position isn't difficult to set up, but one has to know about opening databases and opening book for Shredder/Fritz/Arena GUI.
The Leela vs Stockfish game 36 from TCEC Season 17 Superfinal is copied below.