When PC double speed how do we calculate the Elo Gained?

[Chessqueen]

I have some of these old Engines like Shredder 8 2792 (SSDF Athlon 1200) and Rebel 8 , how do we calculate the rating gained of these old engines by using newer PC ?
forum3/viewtopic.php?t=59497

[lkaufman]

I have some of these old Engines like Shredder 8 2792 (SSDF Athlon 1200) and Rebel 8 , how do we calculate the rating gained of these old engines by using newer PC ?
forum3/viewtopic.php?t=59497

A good way to estimate the elo value of a speed doubling on a rating list is to take advantage of the fact that for the past decade, it has been a pretty reliable rule for most good engines that four threads is about as good as one thread running three times as fast. Using the CEGT blitz list (because CEGT uses normal Elo rather than Bayes, which shrinks the differences), most engines listed with one and four CPUs prior to NNUE showed roughly a hundred elo difference; it's less with the NNUE engines, perhaps just because they are so much stronger that the draw percentage between them goes way up. So roughly speaking, pre-NNUE, triple speed is 100 true elo points in blitz, which works out to 63 elo per doubling (just divide the logarithms to see why). The gains would be smaller on longer TC lists, and also smaller on CCRL due to BayesElo. I don't know how an i7 at say 3.6 Ghz compares to an Athlon at 1.2 Ghz; probably it's more than just triple, but I don't know that, perhaps someone who's more of a hardware expert can make that comparison more accurately. If it is just triple, then you would add 100 for Blitz, perhaps 85 for Rapid or 70 for classical TC, just guessing from my experience on these. I too am interested in this issue, as I believe the SSDF list is pretty accurate for the specified hardware and time limit in human terms.

[Chessqueen]

I have some of these old Engines like Shredder 8 2792 (SSDF Athlon 1200) and Rebel 8 , how do we calculate the rating gained of these old engines by using newer PC ?
forum3/viewtopic.php?t=59497

A good way to estimate the elo value of a speed doubling on a rating list is to take advantage of the fact that for the past decade, it has been a pretty reliable rule for most good engines that four threads is about as good as one thread running three times as fast. Using the CEGT blitz list (because CEGT uses normal Elo rather than Bayes, which shrinks the differences), most engines listed with one and four CPUs prior to NNUE showed roughly a hundred elo difference; it's less with the NNUE engines, perhaps just because they are so much stronger that the draw percentage between them goes way up. So roughly speaking, pre-NNUE, triple speed is 100 true elo points in blitz, which works out to 63 elo per doubling (just divide the logarithms to see why). The gains would be smaller on longer TC lists, and also smaller on CCRL due to BayesElo. I don't know how an i7 at say 3.6 Ghz compares to an Athlon at 1.2 Ghz; probably it's more than just triple, but I don't know that, perhaps someone who's more of a hardware expert can make that comparison more accurately. If it is just triple, then you would add 100 for Blitz, perhaps 85 for Rapid or 70 for classical TC, just guessing from my experience on these. I too am interested in this issue, as I believe the SSDF list is pretty accurate for the specified hardware and time limit in human terms.

Therefore, Shredder8 rated 2792 (SSDF Athlon 1200) would be close to Magnus on my intel i7 3.40GHz (2792 + 80 ) = 2872 Rapid, but for some reason I would put my money on Shredder8. Now that is matter that Shredder8 only does average of 23 plies at TC 15'+10" on my intel i7 3.40GHz ?

Note: A Beautiful Swindle played by an engine

[lkaufman]

I have some of these old Engines like Shredder 8 2792 (SSDF Athlon 1200) and Rebel 8 , how do we calculate the rating gained of these old engines by using newer PC ?
forum3/viewtopic.php?t=59497

A good way to estimate the elo value of a speed doubling on a rating list is to take advantage of the fact that for the past decade, it has been a pretty reliable rule for most good engines that four threads is about as good as one thread running three times as fast. Using the CEGT blitz list (because CEGT uses normal Elo rather than Bayes, which shrinks the differences), most engines listed with one and four CPUs prior to NNUE showed roughly a hundred elo difference; it's less with the NNUE engines, perhaps just because they are so much stronger that the draw percentage between them goes way up. So roughly speaking, pre-NNUE, triple speed is 100 true elo points in blitz, which works out to 63 elo per doubling (just divide the logarithms to see why). The gains would be smaller on longer TC lists, and also smaller on CCRL due to BayesElo. I don't know how an i7 at say 3.6 Ghz compares to an Athlon at 1.2 Ghz; probably it's more than just triple, but I don't know that, perhaps someone who's more of a hardware expert can make that comparison more accurately. If it is just triple, then you would add 100 for Blitz, perhaps 85 for Rapid or 70 for classical TC, just guessing from my experience on these. I too am interested in this issue, as I believe the SSDF list is pretty accurate for the specified hardware and time limit in human terms.

Therefore, Shredder8 rated 2792 (SSDF Athlon 1200) would be close to Magnus on my intel i7 3.40GHz (2792 + 80 ) = 2872 Rapid, but for some reason I would put my money on Shredder8. Now that is matter that Shredder8 only does average of 23 plies at TC 15'+10" on my intel i7 3.40GHz ?

Note: A Beautiful Swindle played by an engine

You are overlooking that Athlon only had one thread, this calculation assumes you just use one thread, but of course you would use all four (assuming it is four core) (we only used one thread for knight odds, but that's not relevant). But good news, we don't have to guess or even to know anything at all about ancient hardware speeds, because there are enough engines rated on two different hardwares on the SSDF to directly calculate the elo gain (at their standard TC) for varying hardware. Comparing a Q6600 at 2.4 Ghz running four threads with the Athlon 1200 running one, the average elo gain for seven engines was 112 elo. I estimated 70 elo for using four threads instead of one at classical TC, and if the double clock speed actually means it ran twice as fast, that would be an additional 70 * 63% = 44 elo. 70 + 44 = 114, so my guess was within two elo of the actual observed result (which has a significant margin of error). Quite funny! Anyway SSDF also has the AMD Rizen 7 1800x processor with eight cores at 3.6 Ghz, with an average further elo gain of 51 elo for 15 engines. So to get the estimated standard rating of any engine on the SSDF list that was tested on the old Athlon 1200, if it were run on a modern 8 core 3.6 Ghz processor just add 163 elo. Your own i7 is probably about midway between the q6600 and the AMD 1800x (it's only four cores like the former, but about as fast per core as the later), so perhaps just add 112 + 26 = 138 elo to Athlon ratings to get what it would be on your quad. So a 2792 would become 2930, indeed quite a bit above Magnus. Now at Rapid, we know that humans perform worse against engines, but it's not clear just how much worse. I think if you use 100 elo for this, you won't be far from the truth. So in Rapid, Shredder 8 should be about 3030 on your quad on the FIDE Rapid rating list. Don't worry about number of plies, that varies greatly from engine to engine. Anyway I think this gives me reliable enough information to be able to estimate ratings for Skill levels in Dragon for the next release, at least for the ones in the range of reasonably strong human players, say over 1500 or so.

[Chessqueen]

I have some of these old Engines like Shredder 8 2792 (SSDF Athlon 1200) and Rebel 8 , how do we calculate the rating gained of these old engines by using newer PC ?
forum3/viewtopic.php?t=59497

A good way to estimate the elo value of a speed doubling on a rating list is to take advantage of the fact that for the past decade, it has been a pretty reliable rule for most good engines that four threads is about as good as one thread running three times as fast. Using the CEGT blitz list (because CEGT uses normal Elo rather than Bayes, which shrinks the differences), most engines listed with one and four CPUs prior to NNUE showed roughly a hundred elo difference; it's less with the NNUE engines, perhaps just because they are so much stronger that the draw percentage between them goes way up. So roughly speaking, pre-NNUE, triple speed is 100 true elo points in blitz, which works out to 63 elo per doubling (just divide the logarithms to see why). The gains would be smaller on longer TC lists, and also smaller on CCRL due to BayesElo. I don't know how an i7 at say 3.6 Ghz compares to an Athlon at 1.2 Ghz; probably it's more than just triple, but I don't know that, perhaps someone who's more of a hardware expert can make that comparison more accurately. If it is just triple, then you would add 100 for Blitz, perhaps 85 for Rapid or 70 for classical TC, just guessing from my experience on these. I too am interested in this issue, as I believe the SSDF list is pretty accurate for the specified hardware and time limit in human terms.

Therefore, Shredder8 rated 2792 (SSDF Athlon 1200) would be close to Magnus on my intel i7 3.40GHz (2792 + 80 ) = 2872 Rapid, but for some reason I would put my money on Shredder8. Now that is matter that Shredder8 only does average of 23 plies at TC 15'+10" on my intel i7 3.40GHz ?

Note: A Beautiful Swindle played by an engine

You are overlooking that Athlon only had one thread, this calculation assumes you just use one thread, but of course you would use all four (assuming it is four core) (we only used one thread for knight odds, but that's not relevant). But good news, we don't have to guess or even to know anything at all about ancient hardware speeds, because there are enough engines rated on two different hardwares on the SSDF to directly calculate the elo gain (at their standard TC) for varying hardware. Comparing a Q6600 at 2.4 Ghz running four threads with the Athlon 1200 running one, the average elo gain for seven engines was 112 elo. I estimated 70 elo for using four threads instead of one at classical TC, and if the double clock speed actually means it ran twice as fast, that would be an additional 70 * 63% = 44 elo. 70 + 44 = 114, so my guess was within two elo of the actual observed result (which has a significant margin of error). Quite funny! Anyway SSDF also has the AMD Rizen 7 1800x processor with eight cores at 3.6 Ghz, with an average further elo gain of 51 elo for 15 engines. So to get the estimated standard rating of any engine on the SSDF list that was tested on the old Athlon 1200, if it were run on a modern 8 core 3.6 Ghz processor just add 163 elo. Your own i7 is probably about midway between the q6600 and the AMD 1800x (it's only four cores like the former, but about as fast per core as the later), so perhaps just add 112 + 26 = 138 elo to Athlon ratings to get what it would be on your quad. So a 2792 would become 2930, indeed quite a bit above Magnus. Now at Rapid, we know that humans perform worse against engines, but it's not clear just how much worse. I think if you use 100 elo for this, you won't be far from the truth. So in Rapid, Shredder 8 should be about 3030 on your quad on the FIDE Rapid rating list. Don't worry about number of plies, that varies greatly from engine to engine. Anyway I think this gives me reliable enough information to be able to estimate ratings for Skill levels in Dragon for the next release, at least for the ones in the range of reasonably strong human players, say over 1500 or so.

I got my old AMD Athlon 1200 from my basement that was collecting dust and match Shredder 8 vs my Intel i7 3.40 GHz Dragon2 at Knight Odds
I was making both moves going back and forth between the 2 computers and engines, but at one point Shredder 8 with my old AMD Athlon 1200 got too low on time with 45 seconds left that I decided to adjust the game to a draw since Shredder8 was making moves too quickly for me to keep up inputting into Dragon2 or my 2nd PC i7 3.40 GHz. I was using my 4 cores with Contempt = 125 on my i7 for Dradon2. I believe that the rating at the time of the the AMD Athlon 1200 was a little inflated, where Shredder8 in my opinion was no more than 2650.

[pgn][Event "Knight Odds"]
[Site "?"]
[Date "2021.09.14"]
[Round "?"]
[White "Dragon-2-64bit-avx2"]
[Black "Shredder8"]
[Result "1/2-1/2"]
[BlackElo "2792"]
[Time "04:33:17"]
[WhiteElo "3590"]
[TimeControl "1800+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "adjudication"]
[PlyCount "234"]
[WhiteType "program"]
[BlackType "program"]

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. g3 d5 5. Bg2 Kf8 6. O-O dxc4 7. Qa4 c5 8.
Qb5 Qa5 9. Qxc4 Nc6 10. a3 Bxc3 11. bxc3 Nd5 12. a4 Kg8 13. Ba3 cxd4 14.
cxd4 Bd7 15. Rfb1 Rc8 16. Qd3 b6 17. e4 Nc3 18. Rc1 Nb4 19. Qd2 Nba2 20.
Rc2 Qxa4 21. Raxa2 Nxa2 22. Rxa2 Qb3 23. h4 a5 24. Kh2 h6 25. Bd6 Qc3 26.
Qf4 Qxd4 27. Be5 Qc5 28. Rd2 Bc6 29. Rd6 Rh7 30. Bh3 b5 31. Bd4 Qc4 32. Qe5
Re8 33. Bg2 h5 34. Be3 f6 35. Qa1 a4 36. Rd4 Qb3 37. Qc1 Ba8 38. Qd2 e5 39.
Rb4 Qf7 40. Rxb5 g5 41. hxg5 h4 42. g4 h3 43. Bf1 fxg5 44. Be2 Bc6 45. Rb6
Bxe4 46. Rb4 Bd5 47. Rxa4 Rg7 48. Ra7 Bb7 49. Qd6 Qd5 50. Qxd5+ Bxd5 51.
Ra5 Bg2 52. Bc4+ Kh7 53. Bd3+ Kg8 54. Kg3 Bb7 55. Kxh3 Kf7 56. Rb5 Bc8 57.
Be4 Rgg8 58. Kg3 Kf6 59. Bc1 Rh8 60. Rb6+ Be6 61. Bf5 Rh1 62. Bxe6 Rxe6 63.
Bxg5+ Kf7 64. Rb4 Kg6 65. Be3 Rg1+ 66. Kh3 Re1 67. Ra4 Kf6 68. Ra3 Rb1 69.
Ra8 Kg6 70. Kg2 Rc6 71. Ra5 Rb4 72. Rxe5 Rxg4+ 73. Kf3 Rcc4 74. Rd5 Kf6 75.
Ra5 Ra4 76. Rh5 Rge4 77. Rb5 Ke6 78. Rc5 Rac4 79. Rg5 Rh4 80. Kg3 Rhe4 81.
Bb6 Re2 82. Be3 Kf6 83. Rb5 Ke6 84. Rb6+ Kd5 85. Rb1 Ra2 86. Rb5+ Ke4 87.
Rb8 Rca4 88. Rb3 Re2 89. f3+ Kf5 90. Rb5+ Ke6 91. Rb8 Kf5 92. Rf8+ Kg6 93.
Bb6 Rc4 94. Bd8 Rec2 95. Rf6+ Kg7 96. Rb6 Rc1 97. Bf6+ Kf7 98. Be5 Re1 99.
Bb8 Rcc1 100. Bd6 Re6 101. Rb7+ Kg6 102. Bb8 Rg1+ 103. Kf2 Ree1 104. Ba7
Kf5 105. Rb5+ Ke6 106. f4 Rb1 107. Ra5 Rbe1 108. Kf3 Rgf1+ 109. Kg4 Re4
110. Ra6+ Kd5 111. Bb8 Rb4 112. Rd6+ Ke4 113. Re6+ Kd3 114. Be5 Rc4 115.
Kf5 Rc8 116. Rf6 Rh1 117. Rd6+ Ke3 1/2-1/2[/pgn]

[lkaufman]

I got my old AMD Athlon 1200 from my basement that was collecting dust and match Shredder 8 vs my Intel i7 3.40 GHz Dragon2 at Knight Odds
I was making both moves going back and forth between the 2 computers and engines, but at one point Shredder 8 with my old AMD Athlon 1200 got too low on time with 45 seconds left that I decided to adjust the game to a draw since Shredder8 was making moves too quickly for me to keep up inputting into Dragon2 or my 2nd PC i7 3.40 GHz. I was using my 4 cores with Contempt = 125 on my i7 for Dradon2. I believe that the rating at the time of the the AMD Athlon 1200 was a little inflated, where Shredder8 in my opinion was no more than 2650.

[pgn][Event "Knight Odds"]
[Site "?"]
[Date "2021.09.14"]
[Round "?"]
[White "Dragon-2-64bit-avx2"]
[Black "Shredder8"]
[Result "1/2-1/2"]
[BlackElo "2792"]
[Time "04:33:17"]
[WhiteElo "3590"]
[TimeControl "1800+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "adjudication"]
[PlyCount "234"]
[WhiteType "program"]
[BlackType "program"]

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. g3 d5 5. Bg2 Kf8 6. O-O dxc4 7. Qa4 c5 8.
Qb5 Qa5 9. Qxc4 Nc6 10. a3 Bxc3 11. bxc3 Nd5 12. a4 Kg8 13. Ba3 cxd4 14.
cxd4 Bd7 15. Rfb1 Rc8 16. Qd3 b6 17. e4 Nc3 18. Rc1 Nb4 19. Qd2 Nba2 20.
Rc2 Qxa4 21. Raxa2 Nxa2 22. Rxa2 Qb3 23. h4 a5 24. Kh2 h6 25. Bd6 Qc3 26.
Qf4 Qxd4 27. Be5 Qc5 28. Rd2 Bc6 29. Rd6 Rh7 30. Bh3 b5 31. Bd4 Qc4 32. Qe5
Re8 33. Bg2 h5 34. Be3 f6 35. Qa1 a4 36. Rd4 Qb3 37. Qc1 Ba8 38. Qd2 e5 39.
Rb4 Qf7 40. Rxb5 g5 41. hxg5 h4 42. g4 h3 43. Bf1 fxg5 44. Be2 Bc6 45. Rb6
Bxe4 46. Rb4 Bd5 47. Rxa4 Rg7 48. Ra7 Bb7 49. Qd6 Qd5 50. Qxd5+ Bxd5 51.
Ra5 Bg2 52. Bc4+ Kh7 53. Bd3+ Kg8 54. Kg3 Bb7 55. Kxh3 Kf7 56. Rb5 Bc8 57.
Be4 Rgg8 58. Kg3 Kf6 59. Bc1 Rh8 60. Rb6+ Be6 61. Bf5 Rh1 62. Bxe6 Rxe6 63.
Bxg5+ Kf7 64. Rb4 Kg6 65. Be3 Rg1+ 66. Kh3 Re1 67. Ra4 Kf6 68. Ra3 Rb1 69.
Ra8 Kg6 70. Kg2 Rc6 71. Ra5 Rb4 72. Rxe5 Rxg4+ 73. Kf3 Rcc4 74. Rd5 Kf6 75.
Ra5 Ra4 76. Rh5 Rge4 77. Rb5 Ke6 78. Rc5 Rac4 79. Rg5 Rh4 80. Kg3 Rhe4 81.
Bb6 Re2 82. Be3 Kf6 83. Rb5 Ke6 84. Rb6+ Kd5 85. Rb1 Ra2 86. Rb5+ Ke4 87.
Rb8 Rca4 88. Rb3 Re2 89. f3+ Kf5 90. Rb5+ Ke6 91. Rb8 Kf5 92. Rf8+ Kg6 93.
Bb6 Rc4 94. Bd8 Rec2 95. Rf6+ Kg7 96. Rb6 Rc1 97. Bf6+ Kf7 98. Be5 Re1 99.
Bb8 Rcc1 100. Bd6 Re6 101. Rb7+ Kg6 102. Bb8 Rg1+ 103. Kf2 Ree1 104. Ba7
Kf5 105. Rb5+ Ke6 106. f4 Rb1 107. Ra5 Rbe1 108. Kf3 Rgf1+ 109. Kg4 Re4
110. Ra6+ Kd5 111. Bb8 Rb4 112. Rd6+ Ke4 113. Re6+ Kd3 114. Be5 Rc4 115.
Kf5 Rc8 116. Rf6 Rh1 117. Rd6+ Ke3 1/2-1/2[/pgn]

If this was 15' + 10", being "short of time" shouldn't be an issue, playing the endgame with just increment is quite normal at this tc, but anyway in this game a draw was the proper result. I think Dragon will actually do better giving knight odds limiting the search to about 22 plies, especially if you are playing with ponder on, but even without that. Beyond around that depth it just starts to see the flaw in aggressive continuations and essentially gives up when down a piece or more.

[Chessqueen]

I got my old AMD Athlon 1200 from my basement that was collecting dust and match Shredder 8 vs my Intel i7 3.40 GHz Dragon2 at Knight Odds
I was making both moves going back and forth between the 2 computers and engines, but at one point Shredder 8 with my old AMD Athlon 1200 got too low on time with 45 seconds left that I decided to adjust the game to a draw since Shredder8 was making moves too quickly for me to keep up inputting into Dragon2 or my 2nd PC i7 3.40 GHz. I was using my 4 cores with Contempt = 125 on my i7 for Dradon2. I believe that the rating at the time of the the AMD Athlon 1200 was a little inflated, where Shredder8 in my opinion was no more than 2650.

[pgn][Event "Knight Odds"]
[Site "?"]
[Date "2021.09.14"]
[Round "?"]
[White "Dragon-2-64bit-avx2"]
[Black "Shredder8"]
[Result "1/2-1/2"]
[BlackElo "2792"]
[Time "04:33:17"]
[WhiteElo "3590"]
[TimeControl "1800+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "adjudication"]
[PlyCount "234"]
[WhiteType "program"]
[BlackType "program"]

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. g3 d5 5. Bg2 Kf8 6. O-O dxc4 7. Qa4 c5 8.
Qb5 Qa5 9. Qxc4 Nc6 10. a3 Bxc3 11. bxc3 Nd5 12. a4 Kg8 13. Ba3 cxd4 14.
cxd4 Bd7 15. Rfb1 Rc8 16. Qd3 b6 17. e4 Nc3 18. Rc1 Nb4 19. Qd2 Nba2 20.
Rc2 Qxa4 21. Raxa2 Nxa2 22. Rxa2 Qb3 23. h4 a5 24. Kh2 h6 25. Bd6 Qc3 26.
Qf4 Qxd4 27. Be5 Qc5 28. Rd2 Bc6 29. Rd6 Rh7 30. Bh3 b5 31. Bd4 Qc4 32. Qe5
Re8 33. Bg2 h5 34. Be3 f6 35. Qa1 a4 36. Rd4 Qb3 37. Qc1 Ba8 38. Qd2 e5 39.
Rb4 Qf7 40. Rxb5 g5 41. hxg5 h4 42. g4 h3 43. Bf1 fxg5 44. Be2 Bc6 45. Rb6
Bxe4 46. Rb4 Bd5 47. Rxa4 Rg7 48. Ra7 Bb7 49. Qd6 Qd5 50. Qxd5+ Bxd5 51.
Ra5 Bg2 52. Bc4+ Kh7 53. Bd3+ Kg8 54. Kg3 Bb7 55. Kxh3 Kf7 56. Rb5 Bc8 57.
Be4 Rgg8 58. Kg3 Kf6 59. Bc1 Rh8 60. Rb6+ Be6 61. Bf5 Rh1 62. Bxe6 Rxe6 63.
Bxg5+ Kf7 64. Rb4 Kg6 65. Be3 Rg1+ 66. Kh3 Re1 67. Ra4 Kf6 68. Ra3 Rb1 69.
Ra8 Kg6 70. Kg2 Rc6 71. Ra5 Rb4 72. Rxe5 Rxg4+ 73. Kf3 Rcc4 74. Rd5 Kf6 75.
Ra5 Ra4 76. Rh5 Rge4 77. Rb5 Ke6 78. Rc5 Rac4 79. Rg5 Rh4 80. Kg3 Rhe4 81.
Bb6 Re2 82. Be3 Kf6 83. Rb5 Ke6 84. Rb6+ Kd5 85. Rb1 Ra2 86. Rb5+ Ke4 87.
Rb8 Rca4 88. Rb3 Re2 89. f3+ Kf5 90. Rb5+ Ke6 91. Rb8 Kf5 92. Rf8+ Kg6 93.
Bb6 Rc4 94. Bd8 Rec2 95. Rf6+ Kg7 96. Rb6 Rc1 97. Bf6+ Kf7 98. Be5 Re1 99.
Bb8 Rcc1 100. Bd6 Re6 101. Rb7+ Kg6 102. Bb8 Rg1+ 103. Kf2 Ree1 104. Ba7
Kf5 105. Rb5+ Ke6 106. f4 Rb1 107. Ra5 Rbe1 108. Kf3 Rgf1+ 109. Kg4 Re4
110. Ra6+ Kd5 111. Bb8 Rb4 112. Rd6+ Ke4 113. Re6+ Kd3 114. Be5 Rc4 115.
Kf5 Rc8 116. Rf6 Rh1 117. Rd6+ Ke3 1/2-1/2[/pgn]

If this was 15' + 10", being "short of time" shouldn't be an issue, playing the endgame with just increment is quite normal at this tc, but anyway in this game a draw was the proper result. I think Dragon will actually do better giving knight odds limiting the search to about 22 plies, especially if you are playing with ponder on, but even without that. Beyond around that depth it just starts to see the flaw in aggressive continuations and essentially gives up when down a piece or more.

The time Control was "1800+10" or 30'+10" and giving more time to an engine around Shedder 8 strength is very tough and at the moment the only engine that can get a draw is Dragon2, when I purchased Shredder 8 back in 2004 I used to have the same machine as SSDF AMD Athlon 1200 and I never thought that in just 15 years+ any program would give Shredder 8 a Knight Odds, but the point here is that back then the SSDF was over inflating engine by as much as 150 rating points.

[lkaufman]

I got my old AMD Athlon 1200 from my basement that was collecting dust and match Shredder 8 vs my Intel i7 3.40 GHz Dragon2 at Knight Odds
I was making both moves going back and forth between the 2 computers and engines, but at one point Shredder 8 with my old AMD Athlon 1200 got too low on time with 45 seconds left that I decided to adjust the game to a draw since Shredder8 was making moves too quickly for me to keep up inputting into Dragon2 or my 2nd PC i7 3.40 GHz. I was using my 4 cores with Contempt = 125 on my i7 for Dradon2. I believe that the rating at the time of the the AMD Athlon 1200 was a little inflated, where Shredder8 in my opinion was no more than 2650.

[pgn][Event "Knight Odds"]
[Site "?"]
[Date "2021.09.14"]
[Round "?"]
[White "Dragon-2-64bit-avx2"]
[Black "Shredder8"]
[Result "1/2-1/2"]
[BlackElo "2792"]
[Time "04:33:17"]
[WhiteElo "3590"]
[TimeControl "1800+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "adjudication"]
[PlyCount "234"]
[WhiteType "program"]
[BlackType "program"]

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. g3 d5 5. Bg2 Kf8 6. O-O dxc4 7. Qa4 c5 8.
Qb5 Qa5 9. Qxc4 Nc6 10. a3 Bxc3 11. bxc3 Nd5 12. a4 Kg8 13. Ba3 cxd4 14.
cxd4 Bd7 15. Rfb1 Rc8 16. Qd3 b6 17. e4 Nc3 18. Rc1 Nb4 19. Qd2 Nba2 20.
Rc2 Qxa4 21. Raxa2 Nxa2 22. Rxa2 Qb3 23. h4 a5 24. Kh2 h6 25. Bd6 Qc3 26.
Qf4 Qxd4 27. Be5 Qc5 28. Rd2 Bc6 29. Rd6 Rh7 30. Bh3 b5 31. Bd4 Qc4 32. Qe5
Re8 33. Bg2 h5 34. Be3 f6 35. Qa1 a4 36. Rd4 Qb3 37. Qc1 Ba8 38. Qd2 e5 39.
Rb4 Qf7 40. Rxb5 g5 41. hxg5 h4 42. g4 h3 43. Bf1 fxg5 44. Be2 Bc6 45. Rb6
Bxe4 46. Rb4 Bd5 47. Rxa4 Rg7 48. Ra7 Bb7 49. Qd6 Qd5 50. Qxd5+ Bxd5 51.
Ra5 Bg2 52. Bc4+ Kh7 53. Bd3+ Kg8 54. Kg3 Bb7 55. Kxh3 Kf7 56. Rb5 Bc8 57.
Be4 Rgg8 58. Kg3 Kf6 59. Bc1 Rh8 60. Rb6+ Be6 61. Bf5 Rh1 62. Bxe6 Rxe6 63.
Bxg5+ Kf7 64. Rb4 Kg6 65. Be3 Rg1+ 66. Kh3 Re1 67. Ra4 Kf6 68. Ra3 Rb1 69.
Ra8 Kg6 70. Kg2 Rc6 71. Ra5 Rb4 72. Rxe5 Rxg4+ 73. Kf3 Rcc4 74. Rd5 Kf6 75.
Ra5 Ra4 76. Rh5 Rge4 77. Rb5 Ke6 78. Rc5 Rac4 79. Rg5 Rh4 80. Kg3 Rhe4 81.
Bb6 Re2 82. Be3 Kf6 83. Rb5 Ke6 84. Rb6+ Kd5 85. Rb1 Ra2 86. Rb5+ Ke4 87.
Rb8 Rca4 88. Rb3 Re2 89. f3+ Kf5 90. Rb5+ Ke6 91. Rb8 Kf5 92. Rf8+ Kg6 93.
Bb6 Rc4 94. Bd8 Rec2 95. Rf6+ Kg7 96. Rb6 Rc1 97. Bf6+ Kf7 98. Be5 Re1 99.
Bb8 Rcc1 100. Bd6 Re6 101. Rb7+ Kg6 102. Bb8 Rg1+ 103. Kf2 Ree1 104. Ba7
Kf5 105. Rb5+ Ke6 106. f4 Rb1 107. Ra5 Rbe1 108. Kf3 Rgf1+ 109. Kg4 Re4
110. Ra6+ Kd5 111. Bb8 Rb4 112. Rd6+ Ke4 113. Re6+ Kd3 114. Be5 Rc4 115.
Kf5 Rc8 116. Rf6 Rh1 117. Rd6+ Ke3 1/2-1/2[/pgn]

If this was 15' + 10", being "short of time" shouldn't be an issue, playing the endgame with just increment is quite normal at this tc, but anyway in this game a draw was the proper result. I think Dragon will actually do better giving knight odds limiting the search to about 22 plies, especially if you are playing with ponder on, but even without that. Beyond around that depth it just starts to see the flaw in aggressive continuations and essentially gives up when down a piece or more.

The time Control was "1800+10" or 30'+10" and giving more time to an engine around Shedder 8 strength is very tough and at the moment the only engine that can get a draw is Dragon2, when I purchased Shredder 8 back in 2004 I used to have the same machine as SSDF AMD Athlon 1200 and I never thought that in just 15 years+ any program would give Shredder 8 a Knight Odds, but the point here is that back then the SSDF was over inflating engine by as much as 150 rating points.

Are you saying that the current SSDF list is 150 elo inflated relative to FIDE for engines in the human GM range? If so please state your reasons for this belief, I thought they were roughly correct now.

[Chessqueen]

I got my old AMD Athlon 1200 from my basement that was collecting dust and match Shredder 8 vs my Intel i7 3.40 GHz Dragon2 at Knight Odds
I was making both moves going back and forth between the 2 computers and engines, but at one point Shredder 8 with my old AMD Athlon 1200 got too low on time with 45 seconds left that I decided to adjust the game to a draw since Shredder8 was making moves too quickly for me to keep up inputting into Dragon2 or my 2nd PC i7 3.40 GHz. I was using my 4 cores with Contempt = 125 on my i7 for Dradon2. I believe that the rating at the time of the the AMD Athlon 1200 was a little inflated, where Shredder8 in my opinion was no more than 2650.

[pgn][Event "Knight Odds"]
[Site "?"]
[Date "2021.09.14"]
[Round "?"]
[White "Dragon-2-64bit-avx2"]
[Black "Shredder8"]
[Result "1/2-1/2"]
[BlackElo "2792"]
[Time "04:33:17"]
[WhiteElo "3590"]
[TimeControl "1800+10"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKB1R w KQkq - 0 1"]
[Termination "adjudication"]
[PlyCount "234"]
[WhiteType "program"]
[BlackType "program"]

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. g3 d5 5. Bg2 Kf8 6. O-O dxc4 7. Qa4 c5 8.
Qb5 Qa5 9. Qxc4 Nc6 10. a3 Bxc3 11. bxc3 Nd5 12. a4 Kg8 13. Ba3 cxd4 14.
cxd4 Bd7 15. Rfb1 Rc8 16. Qd3 b6 17. e4 Nc3 18. Rc1 Nb4 19. Qd2 Nba2 20.
Rc2 Qxa4 21. Raxa2 Nxa2 22. Rxa2 Qb3 23. h4 a5 24. Kh2 h6 25. Bd6 Qc3 26.
Qf4 Qxd4 27. Be5 Qc5 28. Rd2 Bc6 29. Rd6 Rh7 30. Bh3 b5 31. Bd4 Qc4 32. Qe5
Re8 33. Bg2 h5 34. Be3 f6 35. Qa1 a4 36. Rd4 Qb3 37. Qc1 Ba8 38. Qd2 e5 39.
Rb4 Qf7 40. Rxb5 g5 41. hxg5 h4 42. g4 h3 43. Bf1 fxg5 44. Be2 Bc6 45. Rb6
Bxe4 46. Rb4 Bd5 47. Rxa4 Rg7 48. Ra7 Bb7 49. Qd6 Qd5 50. Qxd5+ Bxd5 51.
Ra5 Bg2 52. Bc4+ Kh7 53. Bd3+ Kg8 54. Kg3 Bb7 55. Kxh3 Kf7 56. Rb5 Bc8 57.
Be4 Rgg8 58. Kg3 Kf6 59. Bc1 Rh8 60. Rb6+ Be6 61. Bf5 Rh1 62. Bxe6 Rxe6 63.
Bxg5+ Kf7 64. Rb4 Kg6 65. Be3 Rg1+ 66. Kh3 Re1 67. Ra4 Kf6 68. Ra3 Rb1 69.
Ra8 Kg6 70. Kg2 Rc6 71. Ra5 Rb4 72. Rxe5 Rxg4+ 73. Kf3 Rcc4 74. Rd5 Kf6 75.
Ra5 Ra4 76. Rh5 Rge4 77. Rb5 Ke6 78. Rc5 Rac4 79. Rg5 Rh4 80. Kg3 Rhe4 81.
Bb6 Re2 82. Be3 Kf6 83. Rb5 Ke6 84. Rb6+ Kd5 85. Rb1 Ra2 86. Rb5+ Ke4 87.
Rb8 Rca4 88. Rb3 Re2 89. f3+ Kf5 90. Rb5+ Ke6 91. Rb8 Kf5 92. Rf8+ Kg6 93.
Bb6 Rc4 94. Bd8 Rec2 95. Rf6+ Kg7 96. Rb6 Rc1 97. Bf6+ Kf7 98. Be5 Re1 99.
Bb8 Rcc1 100. Bd6 Re6 101. Rb7+ Kg6 102. Bb8 Rg1+ 103. Kf2 Ree1 104. Ba7
Kf5 105. Rb5+ Ke6 106. f4 Rb1 107. Ra5 Rbe1 108. Kf3 Rgf1+ 109. Kg4 Re4
110. Ra6+ Kd5 111. Bb8 Rb4 112. Rd6+ Ke4 113. Re6+ Kd3 114. Be5 Rc4 115.
Kf5 Rc8 116. Rf6 Rh1 117. Rd6+ Ke3 1/2-1/2[/pgn]

If this was 15' + 10", being "short of time" shouldn't be an issue, playing the endgame with just increment is quite normal at this tc, but anyway in this game a draw was the proper result. I think Dragon will actually do better giving knight odds limiting the search to about 22 plies, especially if you are playing with ponder on, but even without that. Beyond around that depth it just starts to see the flaw in aggressive continuations and essentially gives up when down a piece or more.

The time Control was "1800+10" or 30'+10" and giving more time to an engine around Shedder 8 strength is very tough and at the moment the only engine that can get a draw is Dragon2, when I purchased Shredder 8 back in 2004 I used to have the same machine as SSDF AMD Athlon 1200 and I never thought that in just 15 years+ any program would give Shredder 8 a Knight Odds, but the point here is that back then the SSDF was over inflating engine by as much as 150 rating points.

Are you saying that the current SSDF list is 150 elo inflated relative to FIDE for engines in the human GM range? If so please state your reasons for this belief, I thought they were roughly correct now.

No, what I meant to say is that back in 2003 and 2004 during the time of AMD Athlon 1200 those ratings given to Shredder 8 and other engines were too high probably not 150 Elo higher but much higher than what should have been.

[JohnW]

I always wondered how to calculate the rating of the old Mephisto ROMS now running on PC. It's a lot more than doubling of speed I would assume comparing a modern day PC to 68000 architecture.