Dragon Vs Delfi Dragon Elo = Set to 2860 Next

[Chessqueen]

Since i can NOT find Movei I will start with Delfi5.4 CCRL 2688 which should be close to 2860 FIDE
This is what Mr. Kaufman wrote in a previous post. " I ran a test match (one thread, Rapid 15' + 10") of Dragon 3 with elo set to 2850 vs. Pharaon 3.5.1 (which is 2613 CCRL Rapid on one thread). After 400 games, Pharaon leads by 39 elo, which suggests that Dragon 3 elo 2850 is about 2574 CCRL Rapid. But although I think the elo ratings for Dragon 3 (above 2000 at least) are reasonably accurate for Rapid by FIDE scale, there is a lot of evidence that they are tougher for humans to play against than normal (non-NNUE) engines that score 50% against them. So this test doesn't mean that a 2574 CCRL Rapid engine would be even with Nakamura, although it probably does mean that my 2700 estimate was too high, maybe 2650 CCRL would be about right for an even match with Nakamura. So at these super-GM levels a rough guide for Rapid is that CCRL = FIDE - 200. However at lower levels the gap grows, maybe 2000 CCRL is like FIDE 2300 for example. At really low levels it shrinks again, maybe 1000 CCRL is no better than 1000 FIDE, not much data on this though"

I just finished testing Komodo Dragon set to Elo 2850 Vs Pharaon 3.5.1 both using i thread or 1 core, and it ended 51 to 49 in favor of Pharaon 3.5.1, I did NOT tested them for 500 games since Mr. Kaufman already tested them for another 400 games. It seems clear to me that CRRL rated engines at 15/10 if you add 240 Elo it would give you the estimated FIDE rating. Therefore, If you want to get any idea of any CCRL engines playing close to T/C 15/10 what their estimated FIDE rating is simply add 240 Elo

[/quote]

[lkaufman]

Since i can NOT find Movei I will start with Delfi5.4 CCRL 2688 which should be close to 2860 FIDE
This is what Mr. Kaufman wrote in a previous post. " I ran a test match (one thread, Rapid 15' + 10") of Dragon 3 with elo set to 2850 vs. Pharaon 3.5.1 (which is 2613 CCRL Rapid on one thread). After 400 games, Pharaon leads by 39 elo, which suggests that Dragon 3 elo 2850 is about 2574 CCRL Rapid. But although I think the elo ratings for Dragon 3 (above 2000 at least) are reasonably accurate for Rapid by FIDE scale, there is a lot of evidence that they are tougher for humans to play against than normal (non-NNUE) engines that score 50% against them. So this test doesn't mean that a 2574 CCRL Rapid engine would be even with Nakamura, although it probably does mean that my 2700 estimate was too high, maybe 2650 CCRL would be about right for an even match with Nakamura. So at these super-GM levels a rough guide for Rapid is that CCRL = FIDE - 200. However at lower levels the gap grows, maybe 2000 CCRL is like FIDE 2300 for example. At really low levels it shrinks again, maybe 1000 CCRL is no better than 1000 FIDE, not much data on this though"

I just finished testing Komodo Dragon set to Elo 2850 Vs Pharaon 3.5.1 both using i thread or 1 core, and it ended 51 to 49 in favor of Pharaon 3.5.1, I did NOT tested them for 500 games since Mr. Kaufman already tested them for another 400 games. It seems clear to me that CRRL rated engines at 15/10 if you add 240 Elo it would give you the estimated FIDE rating. Therefore, If you want to get any idea of any CCRL engines playing close to T/C 15/10 what their estimated FIDE rating is simply add 240 Elo

[/quote]

Another good simulation trick is that if you have a strong engine (in the GM range) and set it to play blitz (CCRL tc of 2' + 1"), it should play about the same strength as a human GM of the same FIDE rating would at Rapid (15' + 10"). The difference between those time limits is around 200 elo or so for engines. So to simulate Nakamura playing Rapid, if you pick a 2850 CCRL Rapid rated engine and set it to play blitz (2' + 1"), that should be fairly close. That has the advantage of allowing you to play eight games in about the same time as one would normally take with a weaker engine playing Rapid.

[Chessqueen]

Since i can NOT find Movei I will start with Delfi5.4 CCRL 2688 which should be close to 2860 FIDE
This is what Mr. Kaufman wrote in a previous post. " I ran a test match (one thread, Rapid 15' + 10") of Dragon 3 with elo set to 2850 vs. Pharaon 3.5.1 (which is 2613 CCRL Rapid on one thread). After 400 games, Pharaon leads by 39 elo, which suggests that Dragon 3 elo 2850 is about 2574 CCRL Rapid. But although I think the elo ratings for Dragon 3 (above 2000 at least) are reasonably accurate for Rapid by FIDE scale, there is a lot of evidence that they are tougher for humans to play against than normal (non-NNUE) engines that score 50% against them. So this test doesn't mean that a 2574 CCRL Rapid engine would be even with Nakamura, although it probably does mean that my 2700 estimate was too high, maybe 2650 CCRL would be about right for an even match with Nakamura. So at these super-GM levels a rough guide for Rapid is that CCRL = FIDE - 200. However at lower levels the gap grows, maybe 2000 CCRL is like FIDE 2300 for example. At really low levels it shrinks again, maybe 1000 CCRL is no better than 1000 FIDE, not much data on this though"

I just finished testing Komodo Dragon set to Elo 2850 Vs Pharaon 3.5.1 both using i thread or 1 core, and it ended 51 to 49 in favor of Pharaon 3.5.1, I did NOT tested them for 500 games since Mr. Kaufman already tested them for another 400 games. It seems clear to me that CRRL rated engines at 15/10 if you add 240 Elo it would give you the estimated FIDE rating. Therefore, If you want to get any idea of any CCRL engines playing close to T/C 15/10 what their estimated FIDE rating is simply add 240 Elo

Another good simulation trick is that if you have a strong engine (in the GM range) and set it to play blitz (CCRL tc of 2' + 1"), it should play about the same strength as a human GM of the same FIDE rating would at Rapid (15' + 10"). The difference between those time limits is around 200 elo or so for engines. So to simulate Nakamura playing Rapid, if you pick a 2850 CCRL Rapid rated engine and set it to play blitz (2' + 1"), that should be fairly close. That has the advantage of allowing you to play eight games in about the same time as one would normally take with a weaker engine playing Rapid.
[/quote]

That works if you want to test a GM CCRL engine against a human, but if I set Komodo Dragon Elo = 2835 to play versus Dumb 1.9 which is equal to CCRL 2720 or about 2835 FIDE if you let it play using T/C 15/10, since Komodo Dragon will play instantly at the same strength, but if you limit Dumb 1.9 to a tc of 2'+1", by NOT using its full time of 15/10 to search, it will NOT be able to equalize the Elo strength of Komodo Dragon set = 2835 which is about 115 Elo difference from CCRL for instance Dumb 1.9 rated by CCRL = 2720 + 115 will be equal to 2835 FIDE . I already tried it in a few games.


Note: Mr. Kaufman you should find a formula to convert from CCRL to FIDE at least in t/c of 15/10 and CCRL should use the equivalent rating of FIDE next to the CCRL rating, so human players can practice against engines equivalent to their FIDE ratings.

[Chessqueen]

Komodo Dragon Elo set = 2845 Vs Dumb 1.9 rated 2720 by CCRL is losing 43 to 46 or should I say Dumb 1.9 is winning by 3 points so far. I believe that if I set Komodo Dragon Elo = 2850 FIDE it should be equal to Dumb 1.9 CCRL = 2720 at T/C 15/10 which is 130 Elo difference.

[pgn][Event "Testing.-2838"]
[Site "DESKTOP-OFQ3C0P"]
[Date "2022.05.28"]
[Round "89"]
[White "Dragon-2.6.1-64bit-avx2"]
[Black "Dumb-1.9"]
[Result "0-1"]
[BlackElo "2720"]
[ECO "B23"]
[Opening "Sicilian"]
[Time "10:40:17"]
[Variation "Closed, 1.e4 c5 2.Nc3 d6 3.f4"]
[WhiteElo "2845"]
[TimeControl "900+10"]
[Termination "adjudication"]
[PlyCount "97"]
[WhiteType "program"]
[BlackType "program"]

1. e4 c5 2. Nc3 d6 3. f4 g6 4. d4 Bg7 5. dxc5 Nc6 6. cxd6 Bxc3+ 7. bxc3
exd6 8. Nf3 Qa5 9. Bd3 Nf6 10. O-O O-O 11. Qe1 Re8 12. c4 Qc5+ 13. Be3 Qh5
14. Nd2 Ng4 15. h3 Nxe3 16. Qxe3 Qa5 17. Rab1 Qa3 18. Rb5 a6 19. Rd5 Ne7
20. e5 Nxd5 21. cxd5 Bf5 22. Nc4 Qxa2 23. Qf3 dxe5 24. fxe5 Rf8 25. Ne3
Bxd3 26. cxd3 Qa4 27. Ng4 Qd4+ 28. Kh1 f5 29. exf6 h5 30. f7+ Kh8 31. Nh6
Rad8 32. Qg3 Rxd5 33. Qxg6 Qg7 34. Qb6 Rb5 35. Qe6 Qe5 36. Qd7 Rb6 37. Ng8
Kg7 38. Kg1 Rd6 39. Qxb7 Qd4+ 40. Kh2 Rd7 41. Qxa6 Qe5+ 42. Kh1 Rdxf7 43.
Rg1 Rxg8 44. d4 Qxd4 45. Re1 Rf2 46. Qc6 Kh7 47. Rg1 Rgxg2 48. Qxg2 Rxg2
49. Rxg2 {User Adjudication} 0-1[/pgn]

[Chessqueen]

Komodo Dragon Elo set = 2845 Vs Dumb 1.9 rated 2720 by CCRL is losing 43 to 47 or should I say Dumb 1.9 is winning by 4 points so far. I believe that if I set Komodo Dragon Elo = 2850 FIDE it should be equal to Dumb 1.9 CCRL = 2720 at T/C 15/10 which is 130 Elo difference. I do not believe that I need to run 500 games, but yes most of you agree that 90 or 100 games is NOT enough to determine anything, all that I wanted to find out was if an engine rated around 2715 to 2720 by CCRL is equal to FIDE 2845 to 2850, and I am satisfied with what I found out


[pgn][Event "Testing.-2838"]
[Site "DESKTOP-OFQ3C0P"]
[Date "2022.05.28"]
[Round "90"]
[White "Dumb-1.9"]
[Black "Dragon-2.6.1-64bit-avx2"]
[Result "1-0"]
[BlackElo "2845"]
[ECO "B33"]
[Opening "Sicilian"]
[Time "11:03:36"]
[Variation "Pelikan, Chelyabinsk, 9.Bxf6"]
[WhiteElo "2720"]
[TimeControl "900+10"]
[Termination "adjudication"]
[PlyCount "90"]
[WhiteType "program"]
[BlackType "program"]

1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 Nc6 6. Ndb5 d6 7. Bf4 e5
8. Bg5 a6 9. Na3 b5 10. Bxf6 Qxf6 11. Nd5 Qd8 12. c4 b4 13. Qa4 Bd7 14. Nb5
axb5 15. Qxa8 Qxa8 16. Nc7+ Kd8 17. Nxa8 Nd4 18. Rc1 b3 19. axb3 d5 20.
cxb5 Bb4+ 21. Kd1 dxe4 22. Rc7 g6 23. Bc4 f5 24. Rb7 Bc5 25. Kc1 f4 26. Re1
Bc8 27. Rf7 Rf8 28. Rxh7 f3 29. gxf3 Nxf3 30. Re2 Nd4 31. Rxe4 Bd6 32. b6
Rxf2 33. Rh8+ Rf8 34. Reh4 e4 35. Rxf8+ Bxf8 36. Rh8 Ke7 37. Rh7+ Kd6 38.
Rc7 Bb7 39. Rxb7 e3 40. Rf7 Bh6 41. Kd1 e2+ 42. Bxe2 Ke6 43. Rc7 Nf5 44. b7
Ne3+ 45. Ke1 Ng2 {Arena Adjudication} 1-0[/pgn]