Will Komodo Dragon3 have a level based on ratings ?

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

Moderators: hgm, Rebel, chrisw

Chessqueen
Posts: 5578
Joined: Wed Sep 05, 2018 2:16 am
Location: Moving
Full name: Jorge Picado

Re: Will Komodo Dragon3 have a level based on ratings ?

Post by Chessqueen »

lkaufman wrote: Mon Sep 20, 2021 10:30 pm
Chessqueen wrote: Mon Sep 20, 2021 8:24 pm
lkaufman wrote: Mon Sep 20, 2021 5:45 pm
Chessqueen wrote: Mon Sep 20, 2021 5:16 pm
lkaufman wrote: Mon Sep 20, 2021 6:25 am
Chessqueen wrote: Mon Sep 20, 2021 4:41 am
lkaufman wrote: Mon Sep 20, 2021 3:02 am
Chessqueen wrote: Sat Sep 18, 2021 2:52 am
Chessqueen wrote: Sat Sep 18, 2021 2:02 am This time I set up the Skill Level to 18 in Dragon2, there must be something wrong with the Skill Level 114 x 18 = 2052 ? Another question is when I set the skill level to 18 is that with 1 core or with 4 cores on my intel i7 3.40GHz ? So far I have been playing those two game with only 1 core :roll:


I think that you played rather well in these two games for a 1950 player at 15' + 10", but you are probably right that the formula is a bit too generous in rating Dragon around levels 16 thru 19. You already know that your friend Jorge Hasbun was even with levels 21 and 22 combined at that level, so the formula seems about right at those levels, and probably Ruiz at 2286 would be somewhere between levels 20 and 21 at that time control. For knight odds instead perhaps try the b1 knight for variety this time, with depth 22 and four threads.
So why cc2150dx Jason Coombs stated this?
Skill levels 1-24 are set to play with 1 core/thread regardless of how many cores/threads you set.
I imagine he read that the Skill levels were calibrated based on one thread and wrongly assumed that threads were just set to 1 automatically.
I'm really having a tough time deciding what CCRL Rapid rating corresponds to a FIDE human rating of let's say 2000, assuming both players are playing 15' + 10" rapid on the reference i7 hardware. Some people seem to think that the CCRL ratings are way too low, several hundred below 2000 in this example, while others think they are just slightly low in this range. What is your own experience as to which engine or engines are a fair match for you in 15' + 10" rapid?
This program Snowy is a fair match for me, I really had a hard time surviving the opening since I was stuck trying to get out alive, Snowy prevented me from castling but I managed to get out of all snowy attacks and got a draw :roll:

[pgn][Event "Finding a Challenger"]
[Date "2021.09.20"]
[Round "?"]
[White "Snowy_0_2_x64"]
[Black "ChessQueen"]
[Result "1/2-1/2"]
[BlackElo "1975"]
[ECO "C67"]
[Opening "Spanish"]
[Time "08:42:26"]
[Variation "Open Berlin, 5.d4 Nd6 Queenswap, 9.Nc3 Ke8 10.h3"]
[WhiteElo "1998"]
[TimeControl "900+10"]
[Termination "normal"]
[PlyCount "132"]
[WhiteType "program"]
[BlackType "human"]

1. e4 e5 2. Nf3 Nc6 3. Bb5 Nf6 4. O-O Nxe4 5. d4 Nd6 6. Bxc6 dxc6 7. dxe5
Nf5 8. Qxd8+ Kxd8 9. h3 Ke8 10. Nc3 h5 11. b3 Be6 12. Ne4 Rd8 13. c4 b5 14.
Neg5 Bd7 15. a4 a6 16. Ne4 bxc4 17. bxc4 Be6 18. Bg5 Rc8 19. Rac1 Bb4 20.
Rfd1 Ra8 21. Bf4 Be7 22. Neg5 Bc8 23. c5 Rb8 24. g3 Rb4 25. e6 f6 26. Nf7
Rh7 27. Re1 Rb3 28. Bxc7 Rxf3 29. Nd6+ Bxd6 30. cxd6 Rd3 31. Rb1 Rxd6 32.
Rb8 Rd8 33. Bxd8 Kxd8 34. g4 Ne7 35. Rb4 hxg4 36. hxg4 g6 37. Rc4 Kc7 38.
Rd4 g5 39. f4 gxf4 40. Rxf4 Nd5 41. Rfe4 Re7 42. a5 Kd6 43. Rc4 Rxe6 44.
Rec1 Kc7 45. Rd4 Re5 46. Kf2 Rg5 47. Rg1 Ne7 48. Rf4 Ng8 49. Rh1 Rxg4 50.
Rh7+ Kd6 51. Rf3 Be6 52. Ra7 Bc4 53. Ra8 Rg5 54. Rc3 Kd5 55. Rc1 Nh6 56.
Kf3 Ng4 57. Rd8+ Kc5 58. Rd2 Ne5+ 59. Ke4 Rh5 60. Rdc2 Kd6 61. Rc3 Bd5+ 62.
Ke3 Rh3+ 63. Ke2 Rh2+ 64. Ke3 Rh3+ 65. Ke2 Rh2+ 66. Ke3 Rh3+ {3-fold
repetition} 1/2-1/2[/pgn]
You were clearly winning at the end, I suppose you took the draw due to being down to increment, although even at ten seconds per move you might have won. You say "Snowy prevented you from castling" but that's just the result of choosing the Berlin defense, it's the main line. I suspect you are now stronger than 1950 FIDE strength. So you see my dilemma; your results seem to suggest that the CCRL Rapid ratings around 2000 are not too low at all, while others say they are way too low, humans have no chance against CCRL engines of the same rating they say. I would like to get the levels right for next Dragon to match FIDE ratings at Rapid, but I don't know how far off the CCRL ratings really are from FIDE in the 2000 ballpark.
I was forced to repeat since I was running very low on time with only 1 minute left, well exactly 55 seconds left and Snowy had 5 minutes left, show me how would you play this ending against snowy with only 55 seconds left to snowy 5 minutes. You can run several games with Skill Level 18 with your Komodo Dragon2 vs snowy on your machine http://ccrl.chessdom.com/ccrl/4040/cgi/ ... 0_2_64-bit :roll:
55 seconds or one second, doesn't matter if you have a ten second increment for each move, that's enough to win an easily won endgame. Of course what constitutes an easily won endgame depends on the level of the player. I thought your endgame was sufficiently winning for a 1950 FIDE player to win at ten seconds per move, but I could be mistaken about that, it might require a higher elo than that. Or at least, if you played it out taking nine seconds per move so as not to flag, you would still probably draw if you failed to win.
I decided to see what would happened if I play fast and try to convert, since i realized that I had plenty of time with the 10 seconds increment, well my thought was just that I could not move as fast as Snowy with only 55 seconds left, I totally forgot about the 10 seconds increment so at that moment I panic, but here it is, all that I had to do is to trade a Rook and push my pawn :roll:

[pgn][Event "Finding a Challenger"]
[Date "2021.09.20"]
[Round "?"]
[White "Snowy_0_2_x64"]
[Black "ChessQueen"]
[Result "0-1"]
[BlackElo "1975"]
[ECO "C67"]
[Opening "Spanish"]
[Time "08:42:26"]
[Variation "Open Berlin, 5.d4 Nd6 Queenswap, 9.Nc3 Ke8 10.h3"]
[WhiteElo "1998"]
[TimeControl "900+10"]
[Termination "normal"]
[PlyCount "132"]
[WhiteType "program"]
[BlackType "human"]


1. e4 e5 2. Nf3 Nc6 3. Bb5 Nf6 4. O-O Nxe4 5. d4 Nd6 6. Bxc6 dxc6 7. dxe5
Nf5 8. Qxd8+ Kxd8 9. h3 Ke8 10. Nc3 h5 11. b3 Be6 12. Ne4 Rd8 13. c4 b5 14.
Neg5 Bd7 15. a4 a6 16. Ne4 bxc4 17. bxc4 Be6 18. Bg5 Rc8 19. Rac1 Bb4 20.
Rfd1 Ra8 21. Bf4 Be7 22. Neg5 Bc8 23. c5 Rb8 24. g3 Rb4 25. e6 f6 26. Nf7
Rh7 27. Re1 Rb3 28. Bxc7 Rxf3 29. Nd6+ Bxd6 30. cxd6 Rd3 31. Rb1 Rxd6 32.
Rb8 Rd8 33. Bxd8 Kxd8 34. g4 Ne7 35. Rb4 hxg4 36. hxg4 g6 37. Rc4 Kc7 38.
Rd4 g5 39. f4 gxf4 40. Rxf4 Nd5 41. Rfe4 Re7 42. a5 Kd6 43. Rc4 Rxe6 44.
Rec1 Kc7 45. Rd4 Re5 46. Kf2 Rg5 47. Rg1 Ne7 48. Rf4 Ng8 49. Rh1 Rxg4 50.
Rh7+ Kd6 51. Rf3 Be6 52. Ra7 Bc4 53. Ra8 Rg5 54. Rc3 Kd5 55. Rc1 Nh6 56.
Kf3 Ng4 57. Rd8+ Kc5 58. Rd2 Ne5+ 59. Ke4 Rh5 60. Rdc2 Kd6 61. Rc3 Bd5+ 62.
Ke3 Rh3+ 63. Kd2 Rxc3 64. Kxc3 f5 65. Rf1 Ke6 66. Kd4 Nc4 67. Kc3 Ne3 68.
Rf4 Ke5 69. Rh4 f4 70. Rh5+ Kf6 71. Rh6+ Kg5 72. Rh2 Ng4 73. Rc2 f3 74. Rc1
f2 75. Kd3 Bg2 76. Ke2 Kf4 77. Rc4+ Ke5 78. Rc5+ Bd5 79. Rc3 Kd4 80. Rd3+
Kc5 81. Rd1 Bc4+ 82. Kd2 f1=Q 83. Rxf1 Bxf1 *[/pgn]
Do NOT worry and be happy, we all live a short life :roll: