Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Najdorf revisited
- - By Hamster (**) Date 2020-09-30 15:40
With Stockfish NNUE being very strong and quite unique in its evals I am thinking about looking into the Najdorf - again. As we all know this is the most popular opening for computer chess as well as for correspondence chess (and even OTB I think). There I thought it makes sense to go with 5 separate trees for the "main variations", does that makes sense or are there too many transpositions?
Also, am I missing much if I omit 6.g3, 6.Nb3, 6.Bc4 and others? I never felt like playing them with White and have hardly ever faced them with Black either.

Tree 1: 6.h3
Tree 2: 6.f3
Tree 3: 6.Be2
Tree 4: 6.Be3
Tree 5: 6.Bg5
Parent - By user923005 (****) Date 2020-09-30 16:58
Here is some of my Najdorf  Analysis, ordered by how often the positions occurred in real games.
The field cce is my own estimation of the value of the position and is based on wins, losses, and draws for the position.
If there are a huge number of games, then cce is quite reliable.  If there is a very deep search (acd) then ce is more reliable.

https://drive.google.com/file/d/1b0RO-XAew7us33XqYeV1hb8Gp6__WMfl/view?usp=sharing

sample data row:
rnbqkb1r/1p2pppp/p2p1n2/8/3NP3/2N5/PPP2PPP/R1BQKB1R w KQkq - acd 60; bm Be3; ce 0; cce -14;
pm Be3 {9880} Bg5 {7280} Be2 {4989} Bc4 {2664} h3 {1383} f3 {1367} f4 {1248} g3 {964} a4 {466} Bd3 {182} Nb3 {110} Qf3 {110} Rg1 {87} a3 {31} h4 {25} Qd3 {22} Qe2 {13} Nf3 {8} g4 {3} b3 {1} Bf4 {1} e5 {1} Nd5 {1}; pv Be3 e5 Nb3 Be6 f4 exf4 Bxf4 Nc6 Qd2 d5 exd5 Nxd5 Nxd5 Qxd5 Qxd5 Bxd5 O-O-O O-O-O Rg1 Be7 g3 Be6 Bg2 h5 h4 Bg4 Rxd8+ Rxd8 Re1 Kd7 c3 f6 a3 Rc8 Kc2 Bf5+ Kc1; white_wins 79256; black_wins 87217; draws 103270; Opening B90a Sicilian: Najdorf;
Parent - - By bformless (**) Date 2020-10-01 17:54 Upvotes 1
Maybe I can help you a little in your decision by showing you a screenshot of my ESGTB opening book among many other opening books.

Parent - - By Hamster (**) Date 2020-10-02 16:05
:smile: Impressive table, will take me a while to get something similar.

Did you include lc0 there too? Would be interesting, she likes 6.f3 most (at least with the J92-145 net and 200k nodes).
Parent - By bformless (**) Date 2020-10-02 20:36 Edited 2020-10-02 20:40
I have been using Stockfish mainly for a few years. Currently Stockfish NNUE. From time to time I also used asmFish and at the very beginning almost 9 years ago Houdini. Unfortunately the IA evaluation is not correct because I render a new tree and had to delete the IA trees from time to time because they got too big. I could also use Leela (the CUDA version), but I need uniform evaluations and different engines in the same tree is not so optimal. You could also call my ESGTB opening book the Stockfish opening book, because I've been rendering the tree exclusively with this engine for almost 8 years now and currently have almost 7 million positions with about 20+ alternatives per position calculated from the starting position.
Parent - - By user923005 (****) Date 2020-10-03 20:27
Here is some LC0 multi-pv=3 analysis of Najdorf positions:
https://drive.google.com/file/d/1VkYb22-IxdZD5VCvtSkpWtfR_gV5l3NV/view?usp=sharing
Parent - - By Hamster (**) Date 2020-10-06 08:43
Thanks. What is the best way to view the epd-file? Is there a good way to look at them with Aquarium or is Arena (or another program) better suited?
Parent - - By user923005 (****) Date 2020-10-06 09:37
Aquarium can import EPD.
You can even build a tree directly from it. (Or maybe I am thinking of ChessAssistant?)
Anyway, all the commerical chess GUIs know how to read EPD, even ChessBase.

Arena is not so good at editing and perusing EPD.
Scid can do it.
Winboard can do it.
If you just want to look at them as pictures, EPD2DIAG can do it.

I use my own tools and manipulate the data with SQL commands.
I don't use any of the commercial tools for EPD processing much anymore, except for non-stockfish based engines analysis for which I usually use Arena to run the engine and then my own tools to stitch together the analysis with the EPD.

I modify the source code of any Stockfish based engine to write the analysis directly to disk.
Then I have other tools to read the analysis and populate my SQL database.
Parent - - By user923005 (****) Date 2020-10-06 09:46
For Arena, go to tree utils and the EPD processing is found there.
Parent - By user923005 (****) Date 2020-10-08 15:20
I meant Aquarium, not Arena
Parent - - By Hamster (**) Date 2020-10-05 22:24
Another question, how did you get the same depth (20) for all the positions in the IA-column? I have never seen that before but love it as this makes the IA values much more comparable between each other.
Parent - By user923005 (****) Date 2020-10-06 09:39
Maybe he used Bojung Guo's data, which consists of well over a billion rows analyzed to a depth of 20 plies.
Parent - By bformless (**) Date 2020-10-08 14:43
I add the moves to analyze manually step by step going through the tree and subtree positions. You can choose the time: 10s to 60s "or" depth 20 for example and the alternatives are 20+ for every position. The engine will then calculate up to depth 20. My main tree consists of depth 28 for every alternative in a position as you can see in my screenshot. Some positions contain over 40 alternatives. The depth20 analyzisation is for a new NNUE tree that i let automatically grow by IDeA. I have chosen 40 alternatives for black and white per position and "wide" as the tree shape and 100% of the leafs to explore.
Parent - By MrKris (***) Date 2020-10-08 16:24
A little off topic. Apparently for suprise value and/or if white is higher rated.

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Rg1
ChessTempo.com, Freak Attack, both 2400+:
76 games, 1-0 40.8% , draw 32.9% , 0-1 26.3% (white 57.25%)

Engines and conditions in the game heading: 
Book end*:  Sf - Lc0     Lc0 - Sf
6...e5       draw         draw
6...b6       draw         draw
6...Nc6      draw         draw
6...g6       draw         draw
6...e6       draw         draw
6...h6       draw         draw
6...h5       draw         draw
*all black 6... played CT both 2400+


[Event "G/15'+5'' Sy.6 8GBh"]
[Site "R7 2700X | RTX 2060  LR ~1"]
[Date "2020.10.08"]
[Round "2"]
[White "Stockfish_20100519"]
[Black "Lc0.24.1_LS14.3"]
[Result "1/2-1/2"]
[ECO "B90"]
[Opening "Sicilian"]
[Variation "Najdorf, 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6"]
[TimeControl "900+5"]
[Termination "normal"]
[PlyCount "62"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} c5 {book} 2. Nf3 {book} d6 {book} 3. d4 {book} cxd4 {book} 4.
Nxd4 {book} Nf6 {book} 5. Nc3 {book} a6 {book} 6. Rg1 {book} b5 {book} 7.
a4 {+0.16/37 0 78s} b4 {+0.08/19 0 11s} 8. Nd5 {+0.24/33 0 7.4s} Nxe4
{+0.10/20 0 12s} 9. Bc4 {+0.38/34 0 15s} e6 {0.0022 22s} 10. Qe2 {0.0037
24s} Nc5 {-0.01/23 0 13s} 11. Bd2 {0.0041 25s} g6 {+0.38/21 0 28s} 12. h4
{0.0042 11s} h6 {+0.40/20 0 12s} 13. h5 {0.0041 14s} Bg7 {+0.20/20 0 33s}
14. Be3 {0.0042 15s} g5 {+0.04/24 0 29s} 15. O-O-O {0.0042 17s} O-O
{+0.02/23 0 14s} 16. Nf5 {0.0046 16s}

[FEN "rnbq1rk1/5pb1/p2pp2p/2nN1NpP/PpB5/4B3/1PP1QPP1/2KR2R1 b - - 3 16"]
16... Bxb2+ {+0.02/22 0 15s} 17. Kxb2
{0.0046 14s} Nxa4+ {+0.03/29 0 0.23s} 18. Kc1 {0.0045 14s} exf5 {+0.02/17 0
37s} 19. Bd4 {0.0048 21s} Nc6 {+0.08/16 0 38s} 20. Qe3 {0.0054 34s} Nc3
{+0.08/20 0 0.26s} 21. Nf6+ {0.0051 70s} Kh8 {+0.03/11 0 80s} 22. Nd7+
{0.0053 18s} Kg8 {+0.02/9 0 0.23s} 23. Nxf8 {0.0057 68s} Nxd4 {+0.05/18 0
33s} 24. Qxd4 {0.0050 7.6s} Qxf8 {+0.04/16 0 17s} 25. Qxd6 {0.0053 11s}
Nxd1 {+0.04/14 0 14s} 26. Qg6+ {0.0051 18s} Kh8 {+0.02/11 0 18s} 27. Bxf7
{0.0054 18s} Ra7 {+0.03/9 0 12s} 28. Qf6+ {0.0059 11s} Kh7 {+0.02/7 0 17s}
29. Qg6+ {0.0071 11s} Kh8 {0.003s} 30. Qf6+ {0.0070 13s} Kh7 {0.003 56s}
31. Qg6+ {0.0078 25s} Kh8 {0s, Draw by 3-fold repetition} 1/2-1/2

[Event "RUS vs. ROW"]
[Site "Moscow RUS"]
[Date "2002-09-08"]
[Round "1"]
[White "Ivanchuk, Vassily 2711"]
[Black "Kasparov, Garry 2838"]
[Result "1-0"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Rg1 g6 7. g4 Bg7 8. Be3 Nc6 9. f3 e5 10. Nxc6 bxc6 11. Qd2 Be6 12. O-O-O Bf8 13. Na4 h5 14. h3 Nd7 15. Qc3 hxg4 16. hxg4 d5 17. Qxc6 d4 18. Bd2 Rc8 19. Qb7 Rb8 20. Qxa6 Ra8 21. Qb5 Bxa2 22. Bc4 Bxc4 23. Qxc4 Qf6 24. g5 Qd6 25. Kb1 Rh3 26. Rgf1 Be7 27. b3 Qa3 28. Bc1 Qb4 29. Qxb4 Bxb4 30. f4 Rh4 31. Rh1 Rxh1 32. Rxh1 Ke7 33. f5 Ra6 34. Rh7 Nc5 35. Bd2 Rxa4 36. fxg6 Bxd2 37. Rxf7+ Ke6 38. Rf6+ Ke7 39. bxa4 Nxe4 40. Rf5 1-0
[Event "Chess24 Banter Blitz Cup"]
[Site "chess24.com INT"]
[Date "2020-04-07"]
[Round "6.3"]
[White "Carlsen, Magnus 2863"]
[Black " Sjugirov, Sanan 2675"]
[Result "1-0"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Rg1 b5 7. g4 Bb7 8. f3 g6 9. a3 Bg7 10. Be3 Nbd7 11. Qd2 Rc8 12. O-O-O Ne5 13. Be2 Nc4 14. Bxc4 Rxc4 15. Kb1 O-O 16. h4 Nd7 17. h5 Ne5 18. hxg6 fxg6 19. f4 Rxd4 20. Qxd4 Nc4 21. Qd3 Nxe3 22. Qxe3 a5 23. e5 b4 24. Nd5 bxa3 25. Qxa3 Bxd5 26. Rxd5 Qb6 27. Rgd1 Rxf4 28. exd6 exd6 29. Rxd6 Qf2 30. Qb3+ 1-0
Attachment: NadjorfFreakAttack.pgn - Engine games (33k)
Up Topic The Rybka Lounge / Computer Chess / Najdorf revisited

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill