KomodoDragon 2.5 is out

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Damir
Posts: 2905
Joined: Mon Feb 11, 2008 3:53 pm
Location: Denmark
Full name: Damir Desevac

Re: KomodoDragon 2.5 is out

Post by Damir »

Cumnor wrote: Fri Sep 24, 2021 1:27 pm Please can we have the info link for faq updated that is sent in your email as it seems to point to outdated info.

It does not mention anything about the Dragon version?
I think you should stop using Internet Explorer as your Browser, as it is outdated and no longer updated. Try Microsoft Edge instead. :) :)
Andrew
Posts: 231
Joined: Thu Mar 09, 2006 12:51 am
Location: Australia

Re: KomodoDragon 2.5 is out

Post by Andrew »

Cumnor wrote: Fri Sep 24, 2021 1:27 pm Please can we have the info link for faq updated that is sent in your email as it seems to point to outdated info.

It does not mention anything about the Dragon version?
Was there an email? I've been subscribed to Komodo for many years, don't remember getting an email recently, all updates have been here!

Andrew
Uri Blass
Posts: 11139
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: KomodoDragon 2.5 is out

Post by Uri Blass »

pohl4711 wrote: Fri Sep 24, 2021 11:41 am I just purchased KomodoDragon 2.5 on the Komodo-website.

SPCC-Testruns of Dragon 2.5 and Dragon 2.5 MCTS will start in some hours, when the Stockfish 210921 testruns are finished...


From the readme-file:
"Dragon 2.5 is a huge improvement over all previous Komodo versions, including Dragon 1 and 2, in both standard and MCTS modes. Dragon supports multi-core computers and endgame tablebases. In standard mode, Dragon 2.5 came out 60 elo stronger than Dragon 2 on four threads and 72 elo ahead on one thread in direct matches at the CCRL blitz time control of 2’ + 1”, and in MCTS mode on four threads 73 elo ahead and on one thread 84 elo ahead."
I wonder if dragon2.5 is also better than previous version in odd games against weak engines.
Experience with stockfish show that you can be better in normal chess and weaker in odd games.
Chessqueen
Posts: 5685
Joined: Wed Sep 05, 2018 2:16 am
Location: Moving
Full name: Jorge Picado

Re: KomodoDragon 2.5 is out

Post by Chessqueen »

Uri Blass wrote: Fri Sep 24, 2021 2:37 pm
pohl4711 wrote: Fri Sep 24, 2021 11:41 am I just purchased KomodoDragon 2.5 on the Komodo-website.

SPCC-Testruns of Dragon 2.5 and Dragon 2.5 MCTS will start in some hours, when the Stockfish 210921 testruns are finished...


From the readme-file:
"Dragon 2.5 is a huge improvement over all previous Komodo versions, including Dragon 1 and 2, in both standard and MCTS modes. Dragon supports multi-core computers and endgame tablebases. In standard mode, Dragon 2.5 came out 60 elo stronger than Dragon 2 on four threads and 72 elo ahead on one thread in direct matches at the CCRL blitz time control of 2’ + 1”, and in MCTS mode on four threads 73 elo ahead and on one thread 84 elo ahead."
I wonder if dragon2.5 is also better than previous version in odd games against weak engines.
Experience with stockfish show that you can be better in normal chess and weaker in odd games.
What I really want to find out is if the newer version of Dragon 2.5 has a level option based on ratings, instead of having to go to configuration and set up the Skill level to 18, 19, 20, etc... depending on what rating you would like dragon to pay. Anyway a 2500 rated engine is too much for a Rook Odds, but a 2404 like Dragon 4.4 by Bruno is just the equal rating to Dtragon2 with a Rook Odds, which each program take turn winning and are about equal

[pgn][Event "Rook Odds"]
[Date "2021.09.24"]
[Round "2"]
[White "Dragon-2-64bit-avx2"]
[Black "Dragon_46"]
[Result "0-1"]
[BlackElo "2406"]
[Time "07:34:29"]
[WhiteElo "2000"]
[TimeControl "0+3"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "111"]
[WhiteType "program"]
[BlackType "program"]


1. d3 e5 2. Nf3 Nc6 3. Nbd2 d5 4. e4 Nf6 5. Be2 Bc5 6. c3 Ng4 7. O-O Bxf2+
8. Rxf2 Nxf2 9. Kxf2 Be6 10. Kg1 a5 11. a4 Qe7 12. Nf1 dxe4 13. dxe4 Rd8
14. Qe1 f6 15. Be3 O-O 16. h3 Rd7 17. Kh2 Rfd8 18. N3d2 b6 19. Qd1 Qd6 20.
Ng3 Qf8 21. Qe1 Ne7 22. Nh1 Ng6 23. Bb5 Rd6 24. Nf2 Qf7 25. b4 axb4 26.
cxb4 c6 27. Bf1 Ra8 28. a5 bxa5 29. bxa5 Rxa5 30. Qc1 Ra2 31. Nd3 Qc7 32.
Nc4 Rd7 33. Nc5 Nf8 34. Nb6 Re7 35. Bc4 Ra5 36. Nxe6 Nxe6 37. Qc3 Ra7 38.
Qb3 Kf7 39. Bc5 Rb7 40. h4 Re8 41. h5 Rbb8 42. h6 g6 43. Qb2 Qd8 44. g3 Rb7
45. Qf2 Qd1 46. Qb2 Qh5+ 47. Kg2 Qxh6 48. Qb3 Qd2+ 49. Bf2 Ree7 50. Kf1 Qa5
51. Qe3 Qb4 52. Ba6 Rb8 53. Kg2 Nd4 54. Qh6 Kg8 55. Nc8 Rf7 56. Qc1 Rc7 57.
Qa1 Kg7 58. Kh1 Ra8 59. Qf1 Qa3 60. Be1 Qxa6 61. Qxa6 Rxa6 62. Bc3 Rb7 63.
Bxd4 exd4 64. Kg2 {White resigns} *[/pgn]
Last edited by Chessqueen on Fri Sep 24, 2021 3:01 pm, edited 2 times in total.
Chessqueen
Posts: 5685
Joined: Wed Sep 05, 2018 2:16 am
Location: Moving
Full name: Jorge Picado

Re: KomodoDragon 2.5 is out

Post by Chessqueen »

A 2406 rated engine like Dragon 4.6 by Bruno is just equal to Dragon2 with a Rook Odds, in which each program take turn winning, very soon I will report based on the newer version against the same engine Dragon 4.6 rated 2406 By Bruno. I am almost positive that Dragon 2.5 play much better at Odds than Dragon 2, since the rating is higher :roll:

[pgn][Event "Rook Odds"]
[Date "2021.09.24"]
[Round "2"]
[White "Dragon-2-64bit-avx2"]
[Black "Dragon_46"]
[Result "0-1"]
[BlackElo "2406"]
[Time "07:34:29"]
[WhiteElo "2000"]
[TimeControl "0+3"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "111"]
[WhiteType "program"]
[BlackType "program"]

1. e3 d5 2. c4 e6 3. a3 Nf6 4. Nf3 b6 5. d4 Ba6 6. b3 Bd6 7. Bb2 dxc4 8.
bxc4 O-O 9. Bd3 Bb7 10. Nbd2 Qe7 11. Qb3 c5 12. O-O Nbd7 13. Bb1 Rfd8 14.
g3 Rac8 15. Re1 Ng4 16. Qd3 Nf8 17. h3 Nh6 18. d5 exd5 19. cxd5 Qd7 20. Kg2
b5 21. e4 Re8 22. Rd1 Ng6 23. Nf1 f6 24. Qc2 Rcd8 25. Ne3 Re7 26. a4 a6 27.
e5 fxe5 28. Ng5 Nf7 29. Ne4 c4 30. axb5 axb5 31. h4 Bc8 32. Rh1 Nf8 33. Bc1
Ba6 34. h5 h6 35. f3 Ree8 36. Rd1 Bc8 37. Ng4 Qc7 38. Qb2 Bxg4 39. fxg4 b4
40. g5 c3 41. Qe2 Nxg5 42. Bxg5 hxg5 43. Qd3 Qf7 44. h6 gxh6 45. Rf1 Qc7
46. Nf6+ Kg7 47. Qe4 Re7 48. Nh5+ Kh8 49. Rf6 b3 50. Rf2 Rf7 51. Re2 Qc5
52. Qd3 Be7 53. Qe3 Qxd5+ 54. Be4 Qd1 55. Bf3 c2 56. Kh3 {White resigns} *[/pgn]
User avatar
pohl4711
Posts: 2862
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: KomodoDragon 2.5 is out

Post by pohl4711 »

Chessqueen wrote: Fri Sep 24, 2021 2:48 pm What I really want to find out is if the newer version of Dragon 2.5 has a level option based on ratings, instead of having to go to configuration and set up the Skill level to 18, 19, 20, etc... depending on what rating you would like dragon to play.
New Dragon 2.5 has now 35 skill-levels and the readme says, the skill-level x100 = Human Fide Elo.
So skill-level 20 means 2000 Fide-Elo, for example.
Chessqueen
Posts: 5685
Joined: Wed Sep 05, 2018 2:16 am
Location: Moving
Full name: Jorge Picado

Re: KomodoDragon 2.5 is out

Post by Chessqueen »

I will give you the final score after 100 games at Rook odds, if anybody wants the pgn I will email it to them :roll:

[pgn][Event "Rooks odds"]
[Site "MININT-UB2PIMJ"]
[Date "2021.09.24"]
[Round "3"]
[White "Dragon-2-64bit-avx2"]
[Black "Dragon_46"]
[Result "1-0"]
[BlackElo "2406"]
[Time "08:07:41"]
[WhiteElo "2000"]
[TimeControl "0+3"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "adjudication"]
[PlyCount "156"]
[WhiteType "program"]
[BlackType "program"]

1. Nf3 d5 2. g3 Bf5 3. d3 e6 4. Bg2 Nc6 5. O-O Nf6 6. b3 Bb4 7. Nh4 Bg4 8.
h3 Bh5 9. Bb2 Qe7 10. Nf3 e5 11. Nh4 Nd4 12. g4 Nxg4 13. hxg4 Bxg4 14. Nf3
Nc6 15. a3 Bd6 16. Nc3 d4 17. Nd5 Qe6 18. e4 h5 19. Qd2 h4 20. Kh2 h3 21.
Bh1 O-O-O 22. Rg1 f6 23. Nb4 Ne7 24. Qe1 c5 25. Na2 Ng6 26. Nd2 Rdg8 27. b4
Be7 28. Nc4 b5 29. bxc5 bxc4 30. Qa5 cxd3 31. Nb4 d2 32. Nd5 Qd7 33. Qxd2
Bxc5 34. Qa5 Bd6 35. Qa6+ Kd8 36. Rxg4 Qxg4 37. Qxd6+ Qd7 38. Qb8+ Qc8 39.
Qd6+ Ke8 40. Bxd4 Kf7 41. Be3 Rd8 42. Qb4 a6 43. Qb3 Qe6 44. Bf3 Rh4 45. c4
Qc8 46. c5 Kf8 47. Qb6 Re8 48. c6 Kg8 49. c7 Ne7 50. Be2 Rg4 51. Qb3 Nxd5
52. exd5 Kh8 53. Bb6 Rg6 54. d6 Rg2+ 55. Kh1 Qd7 56. Qf3 Rg5 57. Be3 e4 58.
Qf4 Kg8 59. Qh4 Rd5 60. Bc4 Qf5 61. Kh2 Kf7 62. Qxh3 Qe5+ 63. Kg2 Qxd6 64.
Qh5+ Ke7 65. Qxe8+ Kxe8 66. c8=Q+ Ke7 67. Qb7+ Ke6 68. Qxg7 Qc6 69. Qg8+
Ke5 70. Bxd5 Qxd5 71. Qxd5+ Kxd5 72. Kg3 Kc4 73. Kf4 Kb3 74. Bc5 Kc4 75.
Be7 f5 76. Kxf5 Kd3 77. Bg5 Ke2 78. Be3 Kf3 1-0[/pgn]
Chessqueen
Posts: 5685
Joined: Wed Sep 05, 2018 2:16 am
Location: Moving
Full name: Jorge Picado

Re: KomodoDragon 2.5 is out

Post by Chessqueen »

I will give you the final score after 100 games at Rook odds, if anybody wants the pgn I will email it to them, as you can see these two programs are even at Rook Odds, according to me, but I will not make any prediction until the end of the 100 games, this is version Dragon2, Dragon2.5 will be tested later against the same Dragon 4.6 rated 2406 By Bruno :roll:

[pgn][Event "Rook Odds"]
[Site "MININT-UB2PIMJ"]
[Date "2021.09.24"]
[Round "4"]
[White "Dragon-2-64bit-avx2"]
[Black "Dragon_46"]
[Result "1-0"]
[BlackElo "2406"]
[Time "08:18:24"]
[WhiteElo "2000"]
[TimeControl "0+3"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w Kkq - 0 1"]
[Termination "normal"]
[PlyCount "136"]
[WhiteType "program"]
[BlackType "program"]

1. Nf3 d5 2. e3 Bf5 3. b3 e6 4. Bb2 Nc6 5. a3 Nh6 6. Be2 Rg8 7. O-O Bd6 8.
c4 dxc4 9. bxc4 Ng4 10. d4 b6 11. c5 bxc5 12. h3 Nxf2 13. Kxf2 Rb8 14. Qc1
Qd7 15. dxc5 Bxc5 16. Nbd2 Qd6 17. Nc4 Qe7 18. Bd1 Bd3 19. Ba4 Bxc4 20.
Bxc6+ Kf8 21. Ne5 Qd6 22. Nxc4 Qxc6 23. Ne5 Qb5 24. Kg1 Ke8 25. a4 Qb6 26.
Qd1 Bxe3+ 27. Kh1 Qd6 28. Qf3 Rf8 29. Rd1 Bd2 30. Ba3 c5 31. Nc4 Qc7 32.
Nxd2 Rd8 33. Rc1 Rxd2 34. Rxc5 Qd8 35. Qc6+ Rd7 36. Rb5 Rg8 37. Rb7 g5 38.
Rxa7 g4 39. Ra8 Qxa8 40. Qxa8+ Rd8 41. Qc6+ Rd7 42. hxg4 Rxg4 43. Qc8+ Rd8
44. Qc7 Rd7 45. Qb8+ Rd8 46. Qb5+ Rd7 47. a5 Rh4+ 48. Kg1 Rd4 49. a6 Rd3
50. Qb4 Rxa3 51. Qxa3 Rd1+ 52. Kh2 Rd8 53. a7 Ra8 54. Qc5 Rxa7 55. Qxa7 Kf8
56. Qc5+ Kg7 57. Qg5+ Kf8 58. g4 Ke8 59. Qg8+ Ke7 60. g5 e5 61. Qxh7 Ke8
62. Qh8+ Kd7 63. Qh7 Ke8 64. Qg8+ Ke7 65. Qg7 Ke6 66. Qf6+ Kd5 67. Qxf7+
Kc5 68. g6 e4 {Black resigns} *[/pgn]
Chessqueen
Posts: 5685
Joined: Wed Sep 05, 2018 2:16 am
Location: Moving
Full name: Jorge Picado

Re: KomodoDragon 2.5 is out

Post by Chessqueen »

pohl4711 wrote: Fri Sep 24, 2021 3:16 pm
Chessqueen wrote: Fri Sep 24, 2021 2:48 pm What I really want to find out is if the newer version of Dragon 2.5 has a level option based on ratings, instead of having to go to configuration and set up the Skill level to 18, 19, 20, etc... depending on what rating you would like dragon to play.
New Dragon 2.5 has now 35 skill-levels and the readme says, the skill-level x100 = Human Fide Elo.
So skill-level 20 means 2000 Fide-Elo, for example.
Very soon I will test it against myself to see how accurate it is, and also against some friends that i have with different FIDE ratings :roll:
mjlef
Posts: 1494
Joined: Thu Mar 30, 2006 2:08 pm

Re: KomodoDragon 2.5 is out

Post by mjlef »

AdminX wrote: Fri Sep 24, 2021 1:25 pm
mjlef wrote: Fri Sep 24, 2021 1:11 pm
Damir wrote: Fri Sep 24, 2021 1:00 pm Please don't smoke hash. No problem here, when logging in with my account.. Just downloaded Dragon 2.5... The file is 210 MB.. :shock: :shock:
Yes, the files are a lot bigger since the neural net is more than twice as big, and there are 7 compiles (Windows AVX2 and regular, Linux AVX2 and regular and MacOS AVX2, regular and a special m1 compile). Maybe we should go to a scheme where you can download just the versions for your operating system? Save people some time and bandwidth. Future nets are likely to be even bigger. The bigger, the more knowledge.

Mark
Any plans on selling or adding a Pi Version (64 & 32 Bit)? It might be weaker, but I would use it with my DGT Chessboard.
We have not discussed that yet. We have made some versions for embedded chess boards, so it is possible. Dragon 2.5 uses a lot of memory since it has a pretty big net. You could private me with specs on your Pi and I could see if it would even fit. Also, demand for this here definitely influences us.

Mark