Deep Fritz 8 (Quad Core) Vs Stockfish (IPad)

[mwyoung]

Deep Fritz 8
Q6600 2.4 Ghz Quad Core
1 GB Hash
6 Million Nodes per Second search speed.
SSDF Rating Deep Fritz 8 2GB Q6600 2,4 GHz 2906

Vs.

StockFish
IPad
Hash Unknown
60,000 Nodes per Second search speed.
Rating Unknown.

4 game match @ 15 mins per game.

Deep Fritz 8 ......0 0 0 1 1
StockFish...........1 1 1 0 3

[Event "Blitz 15m"]
[Site "?"]
[Date "2010.12.03"]
[Round "1"]
[White "Stockfish IPad"]
[Black "Deep Fritz 8"]
[Result "1-0"]
[ECO "E12"]
[PlyCount "79"]
[TimeControl "900"]

{1056MB, 11-8-2010.ctg, MWYOUNG-PC} 1. d4 {0} Nf6 {B 0} 2. c4 {4} e6 {B 0} 3.
Nf3 {5} b6 {B 0} 4. Nc3 {4} Bb7 {B 0} 5. a3 {4} g6 {B 0} 6. Qc2 {4} d5 {0.52/
13 27} 7. cxd5 {31} exd5 {0.62/14 0} 8. Bf4 {53} a6 {0.60/14 0} 9. Rc1 {31}
Nbd7 {0.54/13 21} 10. e3 {76} Bg7 {0.31/15 0} 11. b4 {28} O-O {0.34/14 20} 12.
Be2 {39} c6 {0.20/14 23} 13. O-O {77} Qe7 {0.20/14 20} 14. Ne5 {45} Rfc8 {0.30/
14 5} 15. Na4 {49} Qd8 {0.31/13 25} 16. Qb2 {40} Nf8 {0.31/13 20} 17. Bg5 {23}
Qc7 {0.57/14 5} 18. Bf3 {40} h6 {0.71/15 0} 19. Bh4 {27} Bh8 {0.90/13 23} 20.
Qb3 {43} g5 {1.12/14 25} 21. Bg3 {24} Qd8 {1.20/14 0} 22. Ng4 {14} Bg7 {2.00/
14 19} 23. Nxf6+ {6} Bxf6 {2.11/16 20} 24. Bg4 {5} Ne6 {2.23/15 46} 25. Qc2 {28
} Ng7 {2.09/14 30} 26. f4 {19} Be7 {2.59/12 16} 27. fxg5 {10} Bxg5 {2.76/13 16}
28. Rce1 {7} h5 {3.68/13 17} 29. Bxc8 {20} Rxc8 {4.05/15 6} 30. Qf2 {12} Qe7 {
4.20/15 26} 31. h4 {20} Bh6 {4.01/15 12} 32. Be5 {23} f5 {3.99/15 13} 33. Nxb6
{24} Rf8 {4.17/15 0} 34. Qg3 {21} Qf7 {4.18/15 0} 35. Na4 {13} Kh7 {4.30/15 0}
36. Nc5 {24} Bc8 {4.27/15 0} 37. e4 {4} a5 {4.45/13 15} 38. bxa5 {27} dxe4 {5.
29/13 0} 39. Nxe4 {22} Ba6 {5.54/14 2} 40. Rf2 {17} 1-0

[Event "Blitz 15m"]
[Site "?"]
[Date "2010.12.03"]
[Round "2"]
[White "Deep Fritz 8"]
[Black "Stockfish IPad"]
[Result "0-1"]
[ECO "E03"]
[PlyCount "72"]
[TimeControl "900"]

{1056MB, 11-8-2010.ctg, MWYOUNG-PC} 1. c4 {B 0} Nf6 {9} 2. d4 {B 0} e6 {4} 3.
g3 {B 0} d5 {8} 4. Bg2 {B 0} dxc4 {4} 5. Qa4+ {B 0} Nbd7 {7} 6. Qxc4 {B 0} a6 {
8} 7. Nf3 {0.10/14 16} b5 {4} 8. Qc6 {0.02/14 16} Rb8 {6} 9. O-O {-0.17/14 15}
Bb7 {5} 10. Qc3 {-0.22/14 28} c5 {41} 11. dxc5 {-0.27/13 23} Bxc5 {43} 12. Be3
{-0.29/14 0} Rc8 {41} 13. Nbd2 {-0.32/14 38} O-O {29} 14. Bxc5 {-0.35/14 30}
Nxc5 {5} 15. Rfd1 {-0.30/13 14} Qb6 {47} 16. Rac1 {-0.32/13 16} Rfd8 {37} 17.
Qb4 {-0.34/14 0} Nd5 {26} 18. Qh4 {-0.46/14 16} Na4 {5} 19. Rxc8 {-0.55/13 11}
Rxc8 {9} 20. Rb1 {-0.55/13 8} Rc2 {11} 21. Ng5 {0.00/14 22} Nf6 {32} 22. Nde4 {
0.00/14 18} Bxe4 {51} 23. Nxe4 {0.01/16 0} Qc7 {27} 24. Nxf6+ {0.47/15 22} gxf6
{10} 25. Be4 {0.38/14 3} Rxb2 {23} 26. Qxh7+ {0.09/14 28} Kf8 {6} 27. Re1 {0.
05/14 18} Nc3 {24} 28. Bf3 {-0.26/14 19} Qc4 {6} 29. a4 {0.15/12 16} bxa4 {31}
30. Bh5 {-0.37/12 21} e5 {28} 31. Ra1 {-0.77/13 0} Nxe2+ {26} 32. Kg2 {-1.79/
14 0} Nc3 {25} 33. Kg1 {-2.72/13 0} Ke7 {27} 34. h4 {-3.91/14 0} Qd5 {21} 35.
Re1 {-5.09/15 0} a3 {21} 36. Bxf7 {-7.73/15 16} Qxf7 {10} 0-1

[Event "Blitz 15m"]
[Site "?"]
[Date "2010.12.03"]
[Round "3"]
[White "Stockfish IPad"]
[Black "Deep Fritz 8"]
[Result "1-0"]
[ECO "C65"]
[PlyCount "101"]
[TimeControl "900"]

{1056MB, 11-8-2010.ctg, MWYOUNG-PC} 1. e4 {0} e5 {B 0} 2. Nf3 {4} Nc6 {B 0} 3.
Bb5 {4} Nf6 {B 0} 4. O-O {5} Bc5 {B 0} 5. c3 {6} O-O {B 0} 6. d4 {5} Bb6 {B 0}
7. Bg5 {4} h6 {B 0} 8. Bh4 {3} d6 {B 0} 9. a4 {4} a5 {B 0} 10. Re1 {4} exd4 {
B 0} 11. Bxc6 {5} bxc6 {B 0} 12. Nxd4 {4} Re8 {B 0} 13. Nd2 {4} c5 {0.60/14 15}
14. Nf5 {44} Bxf5 {0.42/16 0} 15. exf5 {38} Rxe1+ {0.23/17 20} 16. Qxe1 {6} d5
{0.25/16 20} 17. Rd1 {6} Qe8 {0.18/16 24} 18. Bxf6 {11} gxf6 {0.15/16 5} 19.
Nf1 {98} Qxe1 {0.07/17 19} 20. Rxe1 {6} c4 {0.08/18 20} 21. Ng3 {36} Kf8 {0.25/
18 0} 22. Nh5 {26} Ra6 {0.21/18 0} 23. Rd1 {25} c6 {0.27/19 7} 24. Nxf6 {19}
Bd8 {0.30/18 17} 25. Nh5 {10} Ra7 {0.33/17 6} 26. Kf1 {14} Rd7 {0.41/16 3} 27.
Re1 {27} Rd6 {0.50/18 0} 28. g3 {29} Bc7 {0.61/17 27} 29. Ke2 {26} Bd8 {0.57/
17 0} 30. Kd1 {34} Rd7 {0.62/17 17} 31. f4 {16} Re7 {0.64/17 2} 32. Re5 {14}
Bb6 {0.80/17 24} 33. Nf6 {23} Rxe5 {1.14/17 23} 34. Nd7+ {6} Ke7 {1.29/19 19}
35. Nxe5 {10} c5 {1.18/18 6} 36. Ke2 {12} Bc7 {1.46/17 13} 37. Ng4 {7} h5 {1.
47/20 16} 38. Ne3 {5} Kd6 {1.58/20 16} 39. Kf3 {43} Bd8 {2.07/19 13} 40. h3 {5}
Be7 {2.05/19 12} 41. g4 {10} hxg4+ {2.05/19 12} 42. Nxg4 {6} Bh4 {1.84/20 17}
43. Ne5 {7} Ke7 {1.54/19 6} 44. Kg4 {13} Bf6 {1.81/19 3} 45. h4 {19} Bg7 {2.01/
18 0} 46. Kf3 {17} Bh8 {1.94/18 0} 47. h5 {20} Bg7 {2.44/19 2} 48. Ng4 {18} f6
{2.75/20 0} 49. h6 {22} Bf8 {5.22/21 0} 50. h7 {18} Bg7 {5.61/21 0} 51. Ne3 {13
} 1-0

[Event "Blitz 15m"]
[Site "?"]
[Date "2010.12.03"]
[Round "4"]
[White "Deep Fritz 8"]
[Black "Stockfish IPad"]
[Result "1-0"]
[ECO "D34"]
[PlyCount "143"]
[TimeControl "900"]

{1056MB, 11-8-2010.ctg, MWYOUNG-PC} 1. Nf3 {B 0} d5 {8} 2. g3 {B 0} c5 {4} 3.
Bg2 {B 0} Nc6 {4} 4. d4 {B 0} Nf6 {4} 5. O-O {B 0} e6 {4} 6. c4 {B 0} Be7 {6}
7. cxd5 {0.34/14 26} exd5 {6} 8. Nc3 {B 0} O-O {5} 9. dxc5 {B 0} Bxc5 {4} 10.
Bg5 {B 0} d4 {4} 11. Bxf6 {B 0} Qxf6 {4} 12. Nd5 {B 0} Qd8 {3} 13. Nd2 {B 0}
Re8 {4} 14. Rc1 {B 0} Bb6 {4} 15. Nxb6 {0.05/14 39} Qxb6 {43} 16. Nc4 {0.07/14
0} Qd8 {26} 17. e3 {0.16/14 33} Be6 {33} 18. Bxc6 {0.30/15 31} bxc6 {32} 19.
Qxd4 {0.46/16 0} Qxd4 {24} 20. exd4 {0.52/16 0} Rad8 {31} 21. Na5 {0.47/16 0}
Rxd4 {32} 22. Nxc6 {0.43/16 0} Rd2 {26} 23. Nxa7 {0.27/16 0} Rxb2 {33} 24. a3 {
0.15/16 0} g6 {47} 25. Rb1 {0.37/15 21} Reb8 {26} 26. Rxb2 {0.27/18 24} Rxb2 {5
} 27. Ra1 {0.09/18 50} Kg7 {7} 28. Nc6 {0.89/16 20} Kf6 {28} 29. Nb4 {0.96/15 0
} Bc8 {25} 30. h4 {1.04/15 0} Bb7 {25} 31. Nd3 {1.05/17 17} Rd2 {21} 32. Ne1 {
0.98/15 0} Rd5 {26} 33. Nc2 {0.99/16 0} Rd2 {21} 34. Ne3 {1.19/16 24} Ke6 {9}
35. a4 {1.19/16 21} h5 {7} 36. a5 {1.18/16 12} Ba6 {6} 37. Rb1 {1.09/15 9} Bd3
{11} 38. Rb3 {1.13/15 20} Be2 {4} 39. Ra3 {1.06/15 17} Ba6 {8} 40. Rc3 {0.98/
15 9} Kd6 {22} 41. Rb3 {0.54/16 27} Kc5 {26} 42. Ra3 {0.65/16 0} Kb4 {22} 43.
Ra1 {0.66/16 15} Kc5 {39} 44. g4 {0.98/15 20} hxg4 {17} 45. Nxg4 {0.93/15 5}
Rd5 {13} 46. Kh2 {0.94/15 23} f5 {24} 47. Ne3 {1.18/15 14} Rd4 {6} 48. Kh3 {1.
23/15 12} Rd6 {8} 49. Kg3 {1.50/16 26} Rd4 {15} 50. Ng2 {1.53/15 0} Rg4+ {21}
51. Kh3 {1.61/16 0} Rb4 {14} 52. f3 {1.70/15 10} Rb3 {13} 53. Kg3 {1.84/16 3}
Rd3 {9} 54. Nf4 {2.01/14 4} Rd6 {8} 55. Re1 {2.31/16 6} Bc4 {19} 56. Re7 {2.57/
16 0} Rd1 {40} 57. Rg7 {3.03/13 4} Kb5 {21} 58. Nxg6 {3.72/14 11} Kxa5 {14} 59.
Kf4 {3.94/14 0} Rh1 {13} 60. Kxf5 {3.98/13 5} Kb4 {5} 61. f4 {4.09/12 4} Bd3+ {
6} 62. Kg5 {4.17/13 4} Rg1+ {19} 63. Kf6 {4.94/15 0} Rh1 {21} 64. Rh7 {5.21/14
0} Ra1 {16} 65. f5 {5.13/12 2} Ra6+ {15} 66. Kg5 {5.53/13 2} Ra5 {6} 67. Rf7 {
6.29/13 0} Ra1 {15} 68. h5 {6.42/12 2} Rg1+ {9} 69. Kh6 {6.81/14 0} Rh1 {21}
70. f6 {7.38/15 0} Be2 {11} 71. Rb7+ {8.57/11 3} Kc5 {6} 72. f7 {10.96/13 0}
1-0

[mcostalba]

Deep Fritz 8
6 Million Nodes per Second search speed.

Vs.

StockFish
60,000 Nodes per Second search speed.

This is impressive is 2 orders of magnitude of speed difference (X100) !

This is the power of software against hardware.

[Tord Romstad]

StockFish
IPad
Hash Unknown

It's 8 MB, in case you wonder.

60,000 Nodes per Second search speed.

That sounds a little too slow. I usually see around 100 kN/s in the opening and middle game on my iPad. If the speed is consistently as low as 60 kN/s, I suspect that something is wrong. Try rebooting your iPad and see if the speed increases.

Thanks for running this match! Impressive results, but I think Fritz would win a longer match. Stockfish on the iPad is strong, but I don't think it's quite that strong. But in any case, just being able to give Fritz decent resistance with such a big hardware handicap is really good.

[mwyoung]

StockFish
IPad
Hash Unknown

It's 8 MB, in case you wonder.

60,000 Nodes per Second search speed.

That sounds a little too slow. I usually see around 100 kN/s in the opening and middle game on my iPad. If the speed is consistently as low as 60 kN/s, I suspect that something is wrong. Try rebooting your iPad and see if the speed increases.

Thanks for running this match! Impressive results, but I think Fritz would win a longer match. Stockfish on the iPad is strong, but I don't think it's quite that strong. But in any case, just being able to give Fritz decent resistance with such a big hardware handicap is really good.

I did nps for the start position, in the middle game I am sure it is 100,000 NPS. Just wanted to give some idea of the hardware speed difference.

I don't know if I agree that Deep Fritz 8 would win a longer match. The games were pretty crushing by Stockfish. Positionally Deep Fritz 8 can not keep up with Stockfish moves, even with the hardware advantage.

I picked Deep Fritz 8 for its rating, It seemed a good match up 2900 vs Stockfish IPad, I guessed a rating of around 2800. It seems I was wrong.

I will match Deep Fritz 11 Vs. Stockfish Ipad to see how StockFish plays. That maybe asking to much of Stockfish.

Little doubt that Stockfish IPad plays at a very High GM Level. Much higher then the 2500 Elo rating given in the Stockfish Ipad GUI.

[MikeB]

StockFish
IPad
Hash Unknown

It's 8 MB, in case you wonder.

60,000 Nodes per Second search speed.

That sounds a little too slow. I usually see around 100 kN/s in the opening and middle game on my iPad. If the speed is consistently as low as 60 kN/s, I suspect that something is wrong. Try rebooting your iPad and see if the speed increases.

Thanks for running this match! Impressive results, but I think Fritz would win a longer match. Stockfish on the iPad is strong, but I don't think it's quite that strong. But in any case, just being able to give Fritz decent resistance with such a big hardware handicap is really good.

I did nps for the start position, in the middle game I am sure it is 100,000 NPS. Just wanted to give some idea of the hardware speed difference.

I don't know if I agree that Deep Fritz 8 would win a longer match. The games were pretty crushing by Stockfish. Positionally Deep Fritz 8 can not keep up with Stockfish moves, even with the hardware advantage.

I picked Deep Fritz 8 for its rating, It seemed a good match up 2900 vs Stockfish IPad, I guessed a rating of around 2800. It seems I was wrong.

I will match Deep Fritz 11 Vs. Stockfish Ipad to see how StockFish plays. That maybe asking to much of Stockfish.

Little doubt that Stockfish IPad plays at a very High GM Level. Much higher then the 2500 Elo rating given in the Stockfish Ipad GUI.

I agree with your last comment just based on my experience with Droidfish on the HTC Incredible. It sees easily 100K nps, I have it set for 60 moves in 15 seconds. After around 50 games against it. I have one draw to show for it. We need a sd setting to allow us mortals to beat it. Normally for device like a phone or pda with a setting that it makes all 60 moves in 15 seconds - I know I would do better than this. In fact it has received negative ratings on the Android Market for being too strong - although it is a free app anyway. Go figure.

[Tord Romstad]

I did nps for the start position,

Yes, but the point is that the N/s count for the opening position should be much higher than 60kN/s. I get 101kN/s after 60 seconds on my iPad. If you get something significantly below 100kN/s, something is probably wrong.

I don't know if I agree that Deep Fritz 8 would win a longer match. The games were pretty crushing by Stockfish. Positionally Deep Fritz 8 can not keep up with Stockfish moves, even with the hardware advantage.

In these four games, yes. I'm still pretty sure Fritz 8 would get the upper hand in a long match. But as I wrote in my previous e-mail, just putting up a good fight against Fritz 8 on a quad is very good.

I will match Deep Fritz 11 Vs. Stockfish Ipad to see how StockFish plays. That maybe asking to much of Stockfish.

I think so, too, and your first few games seem to confirm this.

Little doubt that Stockfish IPad plays at a very High GM Level. Much higher then the 2500 Elo rating given in the Stockfish Ipad GUI.

A strength setting of 2500 isn't supposed to correspond to a FIDE rating of 2500. This would be impossible to achieve in practice, because it would require calibrating the strength by playing lots of serious tournament games against strong human players. Stockfish at a strength setting of 2500 is supposed to play 100 Elo points stronger than Stockfish at the same hardware at a strength setting of 2400 (and 200 Elo stronger than at 2300, and so on), nothing more and nothing less. Comparing the actual numbers on Stockfish's own rating scale to any other rating list (like the FIDE, CEGT or CCRL lists) is meaningless. Elo ratings are relative, not absolute.

I'll probably adjust the rating scale in the next public version, and let the scale range from 0 to 2000 rather than the current 500 to 2500. Hopefully, this will stop people from confusing the numbers with human ratings.

As to just how strong Stockfish on the iPad is in human terms, nobody knows, and opinions differ. Some strong players think it has super grandmaster strength, others think it plays like an average international master. We'll probably never know the truth.

[Uri Blass]

Scale range from 0 to 2000 is not the best idea to
stop people from confusing the numbers with human ratings because there are people with fide rating of 1800 or fide rating of 2000

It may be better to have a scale rate from minus 3000 to minus 1000 or from 20000 to 22000 and you can be sure that nobody is going to confuse the numbers with human rating.

[mwyoung]

I did nps for the start position,

Yes, but the point is that the N/s count for the opening position should be much higher than 60kN/s. I get 101kN/s after 60 seconds on my iPad. If you get something significantly below 100kN/s, something is probably wrong.

I don't know if I agree that Deep Fritz 8 would win a longer match. The games were pretty crushing by Stockfish. Positionally Deep Fritz 8 can not keep up with Stockfish moves, even with the hardware advantage.

In these four games, yes. I'm still pretty sure Fritz 8 would get the upper hand in a long match. But as I wrote in my previous e-mail, just putting up a good fight against Fritz 8 on a quad is very good.

I will match Deep Fritz 11 Vs. Stockfish Ipad to see how StockFish plays. That maybe asking to much of Stockfish.

I think so, too, and your first few games seem to confirm this.

Little doubt that Stockfish IPad plays at a very High GM Level. Much higher then the 2500 Elo rating given in the Stockfish Ipad GUI.

A strength setting of 2500 isn't supposed to correspond to a FIDE rating of 2500. This would be impossible to achieve in practice, because it would require calibrating the strength by playing lots of serious tournament games against strong human players. Stockfish at a strength setting of 2500 is supposed to play 100 Elo points stronger than Stockfish at the same hardware at a strength setting of 2400 (and 200 Elo stronger than at 2300, and so on), nothing more and nothing less. Comparing the actual numbers on Stockfish's own rating scale to any other rating list (like the FIDE, CEGT or CCRL lists) is meaningless. Elo ratings are relative, not absolute.

I'll probably adjust the rating scale in the next public version, and let the scale range from 0 to 2000 rather than the current 500 to 2500. Hopefully, this will stop people from confusing the numbers with human ratings.

As to just how strong Stockfish on the iPad is in human terms, nobody knows, and opinions differ. Some strong players think it has super grandmaster strength, others think it plays like an average international master. We'll probably never know the truth.

I just tested my Ipad, and Iphone with stockfish. Both systems show about 60Kns when running stockfish. So not sure why you are getting 100Kns. Maybe I have an older system....

[M ANSARI]

Wow ... that is indeed impressive, especially when you consider the fact that this is an even more powerful setup than was used in the Fritz in Bahrain match against Kramnik where Kramnik was totally outplayed even though the match was setup in his favor (he could see engine's book moves and what moves the engine was pondering and would automatically get a draw if there was an EGTB draw etc...). That would mean that we have today on the Ipad and maybe the Iphone, something that is stronger than any human player. This was mentioned by some people jokingly that we will reach a point where a phone would play stronger than the strongest human ... but not in their wildest dreams would they have expected that time to come so soon! A really incredible job by Tord and the Stockfish team!

[Milos]

Thanks for running this match! Impressive results, but I think Fritz would win a longer match. Stockfish on the iPad is strong, but I don't think it's quite that strong. But in any case, just being able to give Fritz decent resistance with such a big hardware handicap is really good.

I would say they should be quite the same in a longer run. On the same hardware as Fritz, SF is around 2^6 times faster than on iPod which is between 300 and 400 elo difference.
Difference between Fritz 8 and SF1.9 on the same hardware is around 350 elo...