We are reaching the Max Odds that a top engine can give the top 3 GM's

[Chessqueen]

[pgn]
[Event "knight odds match"]
[Site "TalkChess"]
[Date "2021.05.29"]
[Round "1"]
[White "KomodoDragon2, MCTS."]
[Black "Whittington, Chris"]
[Result "*"]
[Annotator "Kaufman,Larry"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "73"]
[EventDate "2021.05.20"]
[EventType "match"]

1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 Bg4 5. Be2 Nc6 6. c4 Nb6 7. e6 fxe6 8. O-O
g6 9. a4 a5 10. d5 Bxf3 11. Bxf3 exd5 12. cxd5 Ne5 13. Be4 Bg7 14. f4 Ned7 15.
f5 O-O 16. Ra3 Ne5 17. Rg3 Qe8 18. b3 c6 19. dxc6 bxc6 20. Rf2 e6 21. fxg6 Rxf2
22. gxh7+ Kh8 23. Kxf2 Qf7+ 24. Kg1 Rf8 25. h3 d5 26. Bf3 c5 27. Qe2 Nc6 28.
Bb2 Nd4 29. Qe1 Nd7 30. Bd1 e5 31. Rg5 e4 32. Qg3 Nf3+ 33. Bxf3 Bxb2 34. Rg8+
Kxh7 35. Rg5 Bg7 36. Bh5 Qf6 37. Bg6+ Qxg6 [/pgn]

Yes, I expect that any knight odds matches with GMs will be with Armaggedon rule, because it's easy to get a draw against the engine at knight odds, since the engine (even with any plausible Contempt setting) will be aiming for a draw itself unless the human drops significant material. Having opposing aims where one side must prevail makes it more interesting.

Qxg6

I think that’s over-optimistic. At knight odds the engine is going to lose unless the human player is stressed out and errors. I still might of course.

[pgn]
[Event "knight odds match"]
[Site "TalkChess"]
[Date "2021.05.29"]
[Round "1"]
[White "KomodoDragon2, MCTS."]
[Black "Whittington, Chris"]
[Result "*"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "75"]
[EventDate "2021.05.20"]
[EventType "match"]

1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 Bg4 5. Be2 Nc6 6. c4 Nb6 7. e6 fxe6 8. O-O
g6 9. a4 a5 10. d5 Bxf3 11. Bxf3 exd5 12. cxd5 Ne5 13. Be4 Bg7 14. f4 Ned7 15.
f5 O-O 16. Ra3 Ne5 17. Rg3 Qe8 18. b3 c6 19. dxc6 bxc6 20. Rf2 e6 21. fxg6 Rxf2
22. gxh7+ Kh8 23. Kxf2 Qf7+ 24. Kg1 Rf8 25. h3 d5 26. Bf3 c5 27. Qe2 Nc6 28.
Bb2 Nd4 29. Qe1 Nd7 30. Bd1 e5 31. Rg5 e4 32. Qg3 Nf3+ 33. Bxf3 Bxb2 34. Rg8+
Kxh7 35. Rg5 Bg7 36. Bh5 Qf6 37. Bg6+ Qxg6 38. Qh4+ * [\pgn]

Well, I'm thinking in terms of a low-rated but still active GM (say 2400-2450 FIDE) and the standard Rapid time control of 15' + 10", which is roughly 30" per move for an average length game. So some time stress, but not severe. With Armageddon rule, I think this might be about balanced.
[/quote]
.[/pgn]

[Chessqueen]

Well at 30 second per move if it take an average of only around 45 to 55 moves it is less than an hour, why don't you accept to play it out in one sitting Chrisw and show us what you are capable of with the Armaggedon for the Human. Come on Chrisw accept the challenge, it will only take you less than an hour

[lkaufman]

[Event "knight odds match"]
[Site "TalkChess"]
[Date "2021.05.29"]
[Round "1"]
[White "KomodoDragon2, MCTS."]
[Black "Whittington, Chris"]
[Result "*"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "75"]
[EventDate "2021.05.20"]
[EventType "match"]

1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 Bg4 5. Be2 Nc6 6. c4 Nb6 7. e6 fxe6 8. O-O
g6 9. a4 a5 10. d5 Bxf3 11. Bxf3 exd5 12. cxd5 Ne5 13. Be4 Bg7 14. f4 Ned7 15.
f5 O-O 16. Ra3 Ne5 17. Rg3 Qe8 18. b3 c6 19. dxc6 bxc6 20. Rf2 e6 21. fxg6 Rxf2
22. gxh7+ Kh8 23. Kxf2 Qf7+ 24. Kg1 Rf8 25. h3 d5 26. Bf3 c5 27. Qe2 Nc6 28.
Bb2 Nd4 29. Qe1 Nd7 30. Bd1 e5 31. Rg5 e4 32. Qg3 Nf3+ 33. Bxf3 Bxb2 34. Rg8+
Kxh7 35. Rg5 Bg7 36. Bh5 Qf6 37. Bg6+ Qxg6 38. Qh4+ Qh6 [/pgn]

Well, I'm thinking in terms of a low-rated but still active GM (say 2400-2450 FIDE) and the standard Rapid time control of 15' + 10", which is roughly 30" per move for an average length game. So some time stress, but not severe. With Armageddon rule, I think this might be about balanced.

Qh6

Inverted backslash character btw
[/quote]

[pgn]
[Event "knight odds match"]
[Site "TalkChess"]
[Date "2021.05.29"]
[Round "1"]
[White "KomodoDragon2, MCTS."]
[Black "Whittington, Chris"]
[Result "*"]
[Annotator "Kaufman,Larry"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "77"]
[EventDate "2021.05.20"]
[EventType "match"]

1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 Bg4 5. Be2 Nc6 6. c4 Nb6 7. e6 fxe6 8. O-O
g6 9. a4 a5 10. d5 Bxf3 11. Bxf3 exd5 12. cxd5 Ne5 13. Be4 Bg7 14. f4 Ned7 15.
f5 O-O 16. Ra3 Ne5 17. Rg3 Qe8 18. b3 c6 19. dxc6 bxc6 20. Rf2 e6 21. fxg6 Rxf2
22. gxh7+ Kh8 23. Kxf2 Qf7+ 24. Kg1 Rf8 25. h3 d5 26. Bf3 c5 27. Qe2 Nc6 28.
Bb2 Nd4 29. Qe1 Nd7 30. Bd1 e5 31. Rg5 e4 32. Qg3 Nf3+ 33. Bxf3 Bxb2 34. Rg8+
Kxh7 35. Rg5 Bg7 36. Bh5 Qf6 37. Bg6+ Qxg6 38. Qh4+ Qh6 39. Rh5 * [/pgn]

[Chessqueen]

[Event "knight odds match"]
[Site "TalkChess"]
[Date "2021.05.29"]
[Round "1"]
[White "KomodoDragon2, MCTS."]
[Black "Whittington, Chris"]
[Result "*"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "75"]
[EventDate "2021.05.20"]
[EventType "match"]

1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 Bg4 5. Be2 Nc6 6. c4 Nb6 7. e6 fxe6 8. O-O
g6 9. a4 a5 10. d5 Bxf3 11. Bxf3 exd5 12. cxd5 Ne5 13. Be4 Bg7 14. f4 Ned7 15.
f5 O-O 16. Ra3 Ne5 17. Rg3 Qe8 18. b3 c6 19. dxc6 bxc6 20. Rf2 e6 21. fxg6 Rxf2
22. gxh7+ Kh8 23. Kxf2 Qf7+ 24. Kg1 Rf8 25. h3 d5 26. Bf3 c5 27. Qe2 Nc6 28.
Bb2 Nd4 29. Qe1 Nd7 30. Bd1 e5 31. Rg5 e4 32. Qg3 Nf3+ 33. Bxf3 Bxb2 34. Rg8+
Kxh7 35. Rg5 Bg7 36. Bh5 Qf6 37. Bg6+ Qxg6 38. Qh4+ Qh6 [/pgn]

Well, I'm thinking in terms of a low-rated but still active GM (say 2400-2450 FIDE) and the standard Rapid time control of 15' + 10", which is roughly 30" per move for an average length game. So some time stress, but not severe. With Armageddon rule, I think this might be about balanced.

Qh6

Inverted backslash character btw

[pgn]
[Event "knight odds match"]
[Site "TalkChess"]
[Date "2021.05.29"]
[Round "1"]
[White "KomodoDragon2, MCTS."]
[Black "Whittington, Chris"]
[Result "*"]
[Annotator "Kaufman,Larry"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[PlyCount "77"]
[EventDate "2021.05.20"]
[EventType "match"]

1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 Bg4 5. Be2 Nc6 6. c4 Nb6 7. e6 fxe6 8. O-O
g6 9. a4 a5 10. d5 Bxf3 11. Bxf3 exd5 12. cxd5 Ne5 13. Be4 Bg7 14. f4 Ned7 15.
f5 O-O 16. Ra3 Ne5 17. Rg3 Qe8 18. b3 c6 19. dxc6 bxc6 20. Rf2 e6 21. fxg6 Rxf2
22. gxh7+ Kh8 23. Kxf2 Qf7+ 24. Kg1 Rf8 25. h3 d5 26. Bf3 c5 27. Qe2 Nc6 28.
Bb2 Nd4 29. Qe1 Nd7 30. Bd1 e5 31. Rg5 e4 32. Qg3 Nf3+ 33. Bxf3 Bxb2 34. Rg8+
Kxh7 35. Rg5 Bg7 36. Bh5 Qf6 37. Bg6+ Qxg6 38. Qh4+ Qh6 39. Rh5 * [/pgn]
[/quote]


I gave Snowy which is rate 1977 the Black Position starting from this point on against Stockfish 13 NNUE using my Intel 17 4770 3.40 GHZ and snowy beat Stockfish 13 NNUE

448‑449 Snowy 0.2 64-bit 1977 +13 −13 47.2% +19.0 16.9% 2267
51.2%
450 Slibo 0.5.1 1976 +20 −20 50.9% −12.9 18.7% 990
75.2%
451 La Dame Blanche 2.0c 1968 +12 −12 52.1% −17.9 18.6% 2675
59.3%

[lkaufman]

I gave Snowy which is rate 1977 the Black Position starting from this point on against Stockfish 13 NNUE using my Intel 17 4770 3.40 GHZ and snowy beat Stockfish 13 NNUE

448‑449 Snowy 0.2 64-bit 1977 +13 −13 47.2% +19.0 16.9% 2267
51.2%
450 Slibo 0.5.1 1976 +20 −20 50.9% −12.9 18.7% 990
75.2%
451 La Dame Blanche 2.0c 1968 +12 −12 52.1% −17.9 18.6% 2675
59.3%

Since you often quote results for engines in the general ballpark of 2000 elo or so, I wonder if you have any results for any engines against strong human players, either blitz or rapid? I'm working on a better way to estimate human ratings for engines or skill levels of engines, and any data for midrange engines vs. strong human players would be helpful, as long as time limit and number of threads used are specified.

[Chessqueen]

I gave Snowy which is rate 1977 the Black Position starting from this point on against Stockfish 13 NNUE using my Intel 17 4770 3.40 GHZ and snowy beat Stockfish 13 NNUE

448‑449 Snowy 0.2 64-bit 1977 +13 −13 47.2% +19.0 16.9% 2267
51.2%
450 Slibo 0.5.1 1976 +20 −20 50.9% −12.9 18.7% 990
75.2%
451 La Dame Blanche 2.0c 1968 +12 −12 52.1% −17.9 18.6% 2675
59.3%

Since you often quote results for engines in the general ballpark of 2000 elo or so, I wonder if you have any results for any engines against strong human players, either blitz or rapid? I'm working on a better way to estimate human ratings for engines or skill levels of engines, and any data for midrange engines vs. strong human players would be helpful, as long as time limit and number of threads used are specified.

http://pigeonengine.com/
Pigeon has a low CCRL rating of 1800, but is rated about 2000 against humans online at both ICC and FICS, which tell me that you probably have to add 200 Elo to the CCRL rating to get the human equivalent. That's pretty good, you know, for a bird.
You can probably find other Engines rating against human on ICC and FICS or simply offer Knight Odds to human with Dragon2 MCTS ON ICC or FICS.

[lkaufman]

I gave Snowy which is rate 1977 the Black Position starting from this point on against Stockfish 13 NNUE using my Intel 17 4770 3.40 GHZ and snowy beat Stockfish 13 NNUE

448‑449 Snowy 0.2 64-bit 1977 +13 −13 47.2% +19.0 16.9% 2267
51.2%
450 Slibo 0.5.1 1976 +20 −20 50.9% −12.9 18.7% 990
75.2%
451 La Dame Blanche 2.0c 1968 +12 −12 52.1% −17.9 18.6% 2675
59.3%

Since you often quote results for engines in the general ballpark of 2000 elo or so, I wonder if you have any results for any engines against strong human players, either blitz or rapid? I'm working on a better way to estimate human ratings for engines or skill levels of engines, and any data for midrange engines vs. strong human players would be helpful, as long as time limit and number of threads used are specified.

http://pigeonengine.com/
Pigeon has a low CCRL rating of 1800, but is rated about 2000 against humans online at both ICC and FICS, which tell me that you probably have to add 200 Elo to the CCRL rating to get the human equivalent. That's pretty good, you know, for a bird.
You can probably find other Engines rating against human on ICC and FICS or simply offer Knight Odds to human with Dragon2 MCTS ON ICC or FICS.

Thanks, that is quite helpful! I'm surprised that the gap is only 200 elo at this level, I thought it would be more. Do you have any idea what time limit was the average for the human games on which the ratings were based? If you or anyone else knows of other similar online ratings for engines that are based on games with humans, please advise. I'm trying to determine ratings for Skill levels, that is unrelated to the rating of Dragon giving knight odds, about which I already know quite a bit. At about 6' + 1" it was pretty even with Alex Lenderman at knight odds with Armageddon scoring, so 2650 FIDE blitz level.

[amanjpro]

I gave Snowy which is rate 1977 the Black Position starting from this point on against Stockfish 13 NNUE using my Intel 17 4770 3.40 GHZ and snowy beat Stockfish 13 NNUE

448‑449 Snowy 0.2 64-bit 1977 +13 −13 47.2% +19.0 16.9% 2267
51.2%
450 Slibo 0.5.1 1976 +20 −20 50.9% −12.9 18.7% 990
75.2%
451 La Dame Blanche 2.0c 1968 +12 −12 52.1% −17.9 18.6% 2675
59.3%

Since you often quote results for engines in the general ballpark of 2000 elo or so, I wonder if you have any results for any engines against strong human players, either blitz or rapid? I'm working on a better way to estimate human ratings for engines or skill levels of engines, and any data for midrange engines vs. strong human players would be helpful, as long as time limit and number of threads used are specified.

http://pigeonengine.com/
Pigeon has a low CCRL rating of 1800, but is rated about 2000 against humans online at both ICC and FICS, which tell me that you probably have to add 200 Elo to the CCRL rating to get the human equivalent. That's pretty good, you know, for a bird.
You can probably find other Engines rating against human on ICC and FICS or simply offer Knight Odds to human with Dragon2 MCTS ON ICC or FICS.

Thanks, that is quite helpful! I'm surprised that the gap is only 200 elo at this level, I thought it would be more. Do you have any idea what time limit was the average for the human games on which the ratings were based? If you or anyone else knows of other similar online ratings for engines that are based on games with humans, please advise. I'm trying to determine ratings for Skill levels, that is unrelated to the rating of Dragon giving knight odds, about which I already know quite a bit. At about 6' + 1" it was pretty even with Alex Lenderman at knight odds with Armageddon scoring, so 2650 FIDE blitz level.

You can look at the lichess bots, there is a team called selfmade bots mostly for bots at the lower end of the ccrl rating list.

But it is problematic to use it as the source for your research blindly, as most engines there are under heavy development. and the games are not exclusively against humans.

For Zahak, even when he was barely around 1900, he has almost never lost to a human. Most games were bullets though, but some blitzes too.

I know he played a match against a guy around 2200 at some point, it was a 83 games long "match" , and Zahak won 80-3 or something around this, with one and only one loss.... But I'm not sure if you can really extract much conclusion from this though

[lkaufman]

I gave Snowy which is rate 1977 the Black Position starting from this point on against Stockfish 13 NNUE using my Intel 17 4770 3.40 GHZ and snowy beat Stockfish 13 NNUE

448‑449 Snowy 0.2 64-bit 1977 +13 −13 47.2% +19.0 16.9% 2267
51.2%
450 Slibo 0.5.1 1976 +20 −20 50.9% −12.9 18.7% 990
75.2%
451 La Dame Blanche 2.0c 1968 +12 −12 52.1% −17.9 18.6% 2675
59.3%

Since you often quote results for engines in the general ballpark of 2000 elo or so, I wonder if you have any results for any engines against strong human players, either blitz or rapid? I'm working on a better way to estimate human ratings for engines or skill levels of engines, and any data for midrange engines vs. strong human players would be helpful, as long as time limit and number of threads used are specified.

http://pigeonengine.com/
Pigeon has a low CCRL rating of 1800, but is rated about 2000 against humans online at both ICC and FICS, which tell me that you probably have to add 200 Elo to the CCRL rating to get the human equivalent. That's pretty good, you know, for a bird.
You can probably find other Engines rating against human on ICC and FICS or simply offer Knight Odds to human with Dragon2 MCTS ON ICC or FICS.

Thanks, that is quite helpful! I'm surprised that the gap is only 200 elo at this level, I thought it would be more. Do you have any idea what time limit was the average for the human games on which the ratings were based? If you or anyone else knows of other similar online ratings for engines that are based on games with humans, please advise. I'm trying to determine ratings for Skill levels, that is unrelated to the rating of Dragon giving knight odds, about which I already know quite a bit. At about 6' + 1" it was pretty even with Alex Lenderman at knight odds with Armageddon scoring, so 2650 FIDE blitz level.

You can look at the lichess bots, there is a team called selfmade bots mostly for bots at the lower end of the ccrl rating list.

But it is problematic to use it as the source for your research blindly, as most engines there are under heavy development. and the games are not exclusively against humans.

For Zahak, even when he was barely around 1900, he has almost never lost to a human. Most games were bullets though, but some blitzes too.

I know he played a match against a guy around 2200 at some point, it was a 83 games long "match" , and Zahak won 80-3 or something around this, with one and only one loss.... But I'm not sure if you can really extract much conclusion from this though

Thanks, but I don't see a "Zahak" listed on the CCRL rating list, is it on any rating list, perhaps under another name? Could it be that "pigeon" has a rating of 2000 on ICC based just on winning almost every game from 1300 or so rated players, and would actually crush 2000 rated humans most of the time? I just played two games with it on my pc at 2' + 1" and lost both pretty badly; I'm not very strong in blitz nowadays at age 73 but surely not weaker than 2000 (FIDE blitz is 2186).

[amanjpro]

I gave Snowy which is rate 1977 the Black Position starting from this point on against Stockfish 13 NNUE using my Intel 17 4770 3.40 GHZ and snowy beat Stockfish 13 NNUE

448‑449 Snowy 0.2 64-bit 1977 +13 −13 47.2% +19.0 16.9% 2267
51.2%
450 Slibo 0.5.1 1976 +20 −20 50.9% −12.9 18.7% 990
75.2%
451 La Dame Blanche 2.0c 1968 +12 −12 52.1% −17.9 18.6% 2675
59.3%

Since you often quote results for engines in the general ballpark of 2000 elo or so, I wonder if you have any results for any engines against strong human players, either blitz or rapid? I'm working on a better way to estimate human ratings for engines or skill levels of engines, and any data for midrange engines vs. strong human players would be helpful, as long as time limit and number of threads used are specified.

http://pigeonengine.com/
Pigeon has a low CCRL rating of 1800, but is rated about 2000 against humans online at both ICC and FICS, which tell me that you probably have to add 200 Elo to the CCRL rating to get the human equivalent. That's pretty good, you know, for a bird.
You can probably find other Engines rating against human on ICC and FICS or simply offer Knight Odds to human with Dragon2 MCTS ON ICC or FICS.

Thanks, that is quite helpful! I'm surprised that the gap is only 200 elo at this level, I thought it would be more. Do you have any idea what time limit was the average for the human games on which the ratings were based? If you or anyone else knows of other similar online ratings for engines that are based on games with humans, please advise. I'm trying to determine ratings for Skill levels, that is unrelated to the rating of Dragon giving knight odds, about which I already know quite a bit. At about 6' + 1" it was pretty even with Alex Lenderman at knight odds with Armageddon scoring, so 2650 FIDE blitz level.

You can look at the lichess bots, there is a team called selfmade bots mostly for bots at the lower end of the ccrl rating list.

But it is problematic to use it as the source for your research blindly, as most engines there are under heavy development. and the games are not exclusively against humans.

For Zahak, even when he was barely around 1900, he has almost never lost to a human. Most games were bullets though, but some blitzes too.

I know he played a match against a guy around 2200 at some point, it was a 83 games long "match" , and Zahak won 80-3 or something around this, with one and only one loss.... But I'm not sure if you can really extract much conclusion from this though

Thanks, but I don't see a "Zahak" listed on the CCRL rating list, is it on any rating list, perhaps under another name? Could it be that "pigeon" has a rating of 2000 on ICC based just on winning almost every game from 1300 or so rated players, and would actually crush 2000 rated humans most of the time? I just played two games with it on my pc at 2' + 1" and lost both pretty badly; I'm not very strong in blitz nowadays at age 73 but surely not weaker than 2000 (FIDE blitz is 2186).

Zahak is actually listed in CCRL 404 list (that is the 2m+1s TC): https://ccrl.chessdom.com/ccrl/404/cgi/ ... +opponents

Most of the games on lichess are played when Zahak was still around 1900-2000, as he has only today got the latest rating (~2400)