Resurrection Test Scores

[spacious_mind]

Here are the test scores for the Resurrection programs.

PHOENIX SYSTEMS RESURRECTION



COLOR KEY



Trying to keep the same color key to allow me to include non dedicated programs in future dedicated tournaments with a good degree of success.


RESURRECTION TEST SCORES - 30 SECONDS AVERAGE TIME

I had to do this test twice because playing the average time setting was a bit a problem. Reason being that with Resurrection the clock continues to count forwards when you take back moves. This means that the first couple or three moves take longer than 30 seconds per move average time and thereafter you end up with moves being played at around 15 to 20 seconds per move. This results in test scores that are less than optimal.



Therefore as a second test I played 30 seconds fixed time per move.


RESURRECTION TEST SCORES - 30 SECONDS FIXED TIME

As you will see from the table below, fixed time overall scored better than average time with the only exception being Fruit 2.1 that scored worse. This seems to indicate that even with stronger more modern programs, they still have their sweet spots which in Fruit 2.1's case was a time less than 30 seconds per move. This of course with more time would very likely change again.



In both tests Rybka, Fruit and Toga performed well below expectations. Deepsjeng slightly above and Ruffian just about right.

All the programs struggled with Test Game 4 and to some degree Test Game 5.

Obviously based on the above results that were different for each program, none of them were clones of each other. But I did get a distinct feeling based on the ideas and paths that Toga, Fruit and Rybka followed, that these programs have a relationship. Therefore those past heated debates about origin may not be 100% unfounded.

Since these tests are based on old historical human games, you can also see that these programs were not exactly written with human opponents in mind but more likely to beat other computer programs. Hence these erratic performances.

As an opponent for humans, Deepsjeng seems to have performed best, which is a surprise considering it's low ratings in dedicated computer matches.

Lots of food for thought.

PS.. Philidor with his game style (Game 4) and his opponent would probably have beaten all of the above

Super GM for his time period regards!

[spacious_mind]

Well, after the lower than what I expected results by most of the Resurrection programs in the Rating Test, I wanted to try Resurrection Rybka v. 2.2 against some Gavon programs.

Rybka v. 2.2 32 Bit performs at ELO 2893 at CCRL's site playing 40/4 and ELO 2888 at 40/40 (1 minute per move average). CCRL plays their tests using Athlon 64 X2 4600+ (2.4 GHz).

An Athlon 64 X2 4600+ (2.4 GHz) runs at about 17,500 Drhystone MIPS whereas the StrongArm used in the Resurrection running at 203 MHz runs at about 203 Drhystone MIPS which is about 1 Drhystone MIPS per MHz.

This would be a factor of @ 6.35 and if using a speed doubling of 70 ELO per doubling, Rybka v. 2.2 playing on Resurrection's StrongArm processor would amount to a reduction of 444 ELO from CCRL's 40/40 rating making an approximate ELO strength of ELO 2444 ELO.

Using a speed doubling of 60 ELO points per doubling you would get a 381 ELO reduction from ELO 2888 = ELO 2507 for Rybka v. 2.2 on Resurrection.

Therefore Rybka's ELO should be somewhere between ELO 2507 and ELO 2444 with Resurrection.

With that in mind I wanted to play a couple of Gavon opponents.


RESURRECTION RYBKA V. 2.2 - GAVON HAKKAPELIITTA 1

Gavon's Raspberry PI ARM11-76JZ(F)-S at 700 MHz has about 965 DMPIPS. At CCRL 40/40 Hakkapeliitta 1 is rated at ELO 2515. Using a factor of 4.13 its adjusted rating for Gavon would be somewhere between 2267 - 2226 ELO which theoretically should make Resurrection Rybka v. 2.2 the stronger player. In my rating test Gavon Hakkapeliitta 1 scored ELO 2491

GAME 1

[Event "Computer Test Match"]
[Site "Alabama"]
[Date "2015.08.08"]
[Round "1"]
[White "Resurrection Rybka v. 2.2, 30S Av"]
[Black "Gavon Hakkapeliitta 1, AT30."]
[Result "1/2-1/2"]
[ECO "D90"]
[WhiteElo "2321"]
[BlackElo "2491"]
[PlyCount "127"]
[EventDate "2015.08.08"]
[EventType "match"]
[EventRounds "2"]
[EventCountry "USA"]

1. d4 Nf6 2. c4 g6 3. Nc3 d5 4. Nf3 Bg7 5. Qa4+ Bd7 6. Qb3 dxc4 7. Qxc4 O-O 8.
Bf4 Na6 {Resurrection Rybka v. 2.2 out of book} 9. e3 {Gavon Hakkapeliitta 1
out of book} c5 10. Rd1 Qa5 11. dxc5 Rac8 12. Nd4 Qxc5 13. Qxc5 Nxc5 14. Be2
Nfe4 15. Nd5 Rce8 16. b4 e5 17. bxc5 exf4 18. Nxf4 Re5 19. Nb3 Ba4 20. Rd4 Bc6
21. Bf3 Ree8 22. Rd1 Bb5 23. Nd5 Re5 24. Bxe4 Rxe4 25. g3 Ra4 26. Nd4 Bxd4 27.
Rxd4 Rxa2 28. Nc3 Ra1+ 29. Rd1 Rxd1+ 30. Kxd1 Ba6 31. Kc2 Rc8 32. Rd1 Rxc5 33.
e4 Kg7 34. Rd7 Kf6 35. f4 Ke6 36. Rd8 Bc4 37. e5 Rc7 38. g4 Ke7 39. Rb8 Rd7 40.
Ra8 b6 41. Kb2 f6 42. exf6+ Kxf6 43. Ne4+ Kg7 44. Re8 Bd5 45. h3 Bxe4 46. Rxe4
Rc7 47. Re6 Rf7 48. Re4 Kf6 49. Re8 Rc7 50. Kb3 Kf7 51. Rh8 Ke6 52. Rd8 Ke7 53.
Rd1 Rc5 54. Re1+ Kd6 55. Re8 Rc7 56. Rf8 Ke7 57. Rh8 Kd6 58. Rf8 Ke7 59. Rh8
Ke6 60. Rd8 Ke7 61. Rd1 Rc5 62. Re1+ Kf7 63. Rd1 Ke7 64. Re1+ {3-fold
repetition of position} 1/2-1/2

The game ended in a draw by repetition of moves.

[fen]8/p3k2p/1p4p1/2r5/5PP1/1K5P/8/4R3 w - - 0 54[/fen]

Hakka 1 might have won this game, it had the better opportunities and as you can see from the above position was leading by a pawn. But it ran out of ideas in the endgame.

GAME 2

[Event "Computer Test Match"]
[Site "Alabama"]
[Date "2015.08.08"]
[Round "2"]
[White "Gavon Hakkapeliitta 1, AT30."]
[Black "Resurrection Rybka v. 2.2, 30S Av"]
[Result "0-1"]
[ECO "B21"]
[WhiteElo "2491"]
[BlackElo "2321"]
[PlyCount "94"]
[EventDate "2015.08.08"]
[EventType "match"]
[EventRounds "2"]
[EventCountry "USA"]

1. e4 c5 2. f4 {Resurrection Rybka v. 2.2 out of book} Nf6 3. Nc3 d5 4. exd5
Nxd5 {Gavon Hakkapeliitta 1 out of book} 5. Bc4 Nxf4 6. d3 Nxg2+ 7. Kf1 Nh4 8.
Ne4 e6 9. Nf3 Nxf3 10. Qxf3 Nc6 11. Qh5 g6 12. Qg5 Be7 13. Qh6 f5 14. Ng5 Bf8
15. Qh4 Nd4 16. Qg3 b5 17. Qe5 bxc4 18. Qxh8 Qd5 19. Rg1 h6 20. c3 Nc2 21. Kf2
hxg5 22. dxc4 Qe4 23. Bxg5 Bb7 24. Qh3 Nxa1 25. Re1 Qc2+ 26. Re2 Qa4 27. Qh7
Be4 28. Qxg6+ Kd7 29. b3 Qc6 30. Qf7+ Kc8 31. Qxf8+ Kb7 32. Qe7+ Qc7 33. Rd2
Qxe7 34. Bxe7 Rh8 35. Kg3 e5 36. Bxc5 Nc2 37. Re2 Rh7 38. h3 Kc6 39. Bf2 Rg7+
40. Kh4 a5 41. Rd2 Rg2 42. c5 f4 43. Rd6+ Kc7 44. Rd2 Kc8 45. c4 Nd4 46. Rxd4
exd4 47. Bxd4 f3 0-1

This game has a wild opening which pretty much gave Rybka 2.2 the advantage throughout.

But in the position below Rybka almost blundered the win away.

[fen]r3kb2/pb6/4p1p1/2p2pB1/2P1q3/2P4Q/PP3K1P/n3R3 w q - 0 25[/fen]

This position arises after 25. Re1. Best move here for Rybka was to take the pawn with 25. ... Qxc4. Instead Rybka plays 35. ... Qc2+?

[fen]r3kb2/p6Q/4p1p1/2p2pB1/q1P1b3/2P5/PP2RK1P/n7 w q - 0 28[/fen]

In this position 28. Rd2+! would have assured Hakka 1 a draw. Hakka 1 missed it and allows Rybka to regain its composure and the win.

Final Score:

Resurrection Rybka v. 2.2 - 1.1/2 Points
Gavon Hakkapeliitta 1 - 1/2 Point

Resurrection Rybka wins the match!


RESURRECTION RYBKA V. 2.2 - GAVON DISCOCHECK V.3.7

For the next test match I played Gavon DiscoCheck v. 3.7 that scored ELO 2501 in my rating test. At CCRL 40/40 DiscoCheck v. 3.7 has a rating of ELO 2737.

Making the same allowances as previously DiscoCheck v. 3.7 rating with Gavon would be somewhere between 2489 - 2448 ELO.

Therefore this should be a much tougher challenge for Resurrection Rybka v. 2.2.

GAME 1

[Event "Computer Test Match"]
[Site "Alabama"]
[Date "2015.08.08"]
[Round "1"]
[White "Gavon Discocheck 3.7, AT30."]
[Black "Resurrection Rybka v. 2.2, 30S Av"]
[Result "1/2-1/2"]
[ECO "D01"]
[WhiteElo "2501"]
[BlackElo "2321"]
[PlyCount "140"]
[EventDate "2015.08.08"]
[EventType "match"]
[EventRounds "2"]
[EventCountry "USA"]

1. d4 Nf6 2. Nc3 d5 3. Bg5 Nbd7 4. Nf3 h6 5. Bf4 c6 {Gavon Discocheck 3.7 out
of book} 6. Qd3 {Resurrection Rybka v. 2.2 out of book} e6 7. O-O-O Ng4 8. Bg3
Bd6 9. Nd2 Qc7 10. Qf3 Bxg3 11. hxg3 Ndf6 12. e4 e5 13. exd5 exd4 14. Re1+ Kd8
15. Ne2 Re8 16. dxc6 bxc6 17. Nb3 Ne5 18. Qf4 c5 19. Kb1 a5 20. Nec1 a4 21. Nd2
a3 22. b3 Nd5 23. Qh4+ f6 24. Nd3 Nc3+ 25. Ka1 Bb7 26. Qh5 Bd5 27. Nf4 Bc6 28.
Bc4 Qd7 29. Nd3 Bxg2 30. Nxe5 fxe5 31. Rhg1 Bb7 32. Rxe5 Rxe5 33. Qxe5 Qe7 34.
Re1 Qxe5 35. Rxe5 Kd7 36. Be6+ Kc6 37. Bf5 Kb5 38. Re7 Bd5 39. Bg6 Rf8 40. f4
Rf6 41. f5 Kb4 42. g4 Nd1 43. Rd7 Ne3 44. Kb1 Bc6 45. Rxg7 Nxg4 46. Ra7 Ne5 47.
Bh7 h5 48. Re7 Ng4 49. Bg6 h4 50. Rh7 Ne5 51. Rh6 h3 52. Rxh3 Nxg6 53. fxg6
Rxg6 54. c3+ dxc3 55. Rh4+ Kb5 56. Nc4 Rg1+ 57. Kc2 Rg2+ 58. Kb1 Bd5 59. Nxa3+
Kc6 60. Nc4 Re2 61. Rh6+ Kc7 62. Rh7+ Kd8 63. Nd6 Rb2+ 64. Kc1 Rxa2 65. Nb7+
Bxb7 66. Rxb7 Rf2 67. Rb5 Rf5 68. b4 Rf1+ 69. Kc2 cxb4 70. Rxb4 Rf3 1/2-1/2

Rybka v. 2.2 had the advantage for most of the game but just could not break through and the game ended in a draw.

GAME 2

[Event "Computer Test Match"]
[Site "Alabama"]
[Date "2015.08.09"]
[Round "2"]
[White "Resurrection Rybka v. 2.2, 30S Av"]
[Black "Gavon Discocheck 3.7, AT30."]
[Result "0-1"]
[ECO "B90"]
[WhiteElo "2501"]
[BlackElo "2321"]
[PlyCount "162"]
[EventDate "2015.08.09"]
[EventType "match"]
[EventRounds "2"]
[EventCountry "USA"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. f3 {Gavon Discocheck 3.
7 out of book} e5 7. Nb3 Be7 8. Be3 O-O 9. Qd2 Be6 10. O-O-O h6 {Resurrection
Rybka v. 2.2 out of book} 11. h4 Nh5 12. Qe1 Nd7 13. Kb1 Rc8 14. g3 b5 15. a3
Nhf6 16. g4 Nh7 17. a4 bxa4 18. Nxa4 Rb8 19. Bxa6 Bxb3 20. cxb3 Rxb3 21. Bc4
Rb8 22. Rh2 Re8 23. Qc3 Nhf8 24. Rc2 Ne6 25. Bxe6 fxe6 26. Qc4 Kf7 27. g5 hxg5
28. hxg5 g6 29. Rh1 Rh8 30. Rch2 Rxh2 31. Rxh2 Bf8 32. Rh7+ Kg8 33. Rh1 Kf7 34.
Qc6 Be7 35. Qc2 Nf8 36. Qc3 Rb7 37. Qd2 d5 38. exd5 exd5 39. Nc5 Rb5 40. Qc1
Qc8 41. Nd3 Qf5 42. Qc3 Rb4 43. Qc2 Rc4 44. Qe2 Kg8 45. Ka2 Ne6 46. b3 Rc7 47.
b4 Nxg5 48. Bxg5 Bxg5 49. b5 Ra7+ 50. Kb3 Be7 51. Rd1 e4 52. fxe4 dxe4 53. Nb2
Ra3+ 54. Kc2 e3+ 55. Rd3 Bf6 56. Qxe3 Ra2 57. Qe8+ Kg7 58. Qd7+ Qxd7 59. Rxd7+
Kh6 60. Kd3 Rxb2 61. Rb7 g5 62. Kc4 g4 63. Rd7 g3 64. Rd1 g2 65. Kd3 Be5 66.
Kc4 Bh2 67. Kc3 Rxb5 68. Kc2 Rc5+ 69. Kb3 Rb5+ 70. Kc2 Rf5 71. Ra1 Rf1 72. Ra6+
Kg7 73. Ra7+ Kg6 74. Ra8 Be5 75. Rg8+ Bg7 76. Ra8 {Gavon Discocheck 3.7
announces Mate in 12.} g1=Q {Resurrection Rybka v. 2.2 announces Mate in 7.}
77. Ra6+ {Gavon Discocheck 3.7 announces Mate in 7.} Kf7 78. Ra1 Qc5+ 79. Kd3
Qd5+ 80. Kc2 Qc4+ 81. Kd2 Bh6# 0-1

Rybka v. 2.2 also had some early advantage in this game immediately after opening book was completed.

DiscoCheck v. 3.7 however defended well and Rybka v. 2.2 missed some best moves. In the end however DiscoCheck v. 3.7 ended up outplaying Resurrection Rybka v. 2.2

[fen]1r1qr1k1/3nbppn/3p3p/4p3/N1B1P1PP/4BP2/1P5R/1K1RQ3 w - - 0 23[/fen]

It's an amalgamation of tiny errors of not finding best moves that eventually cost Rybka the game as in the above example where the move 23. Nc3 would have been a stronger move than the played move 23. Qc3.

Final Score:

Resurrection Rybka v. 2.2 - 1/2 Point
Gavon DiscoCheck v. 3.7 - 1.1/2 Points


Interesting games, you should try and play them back on your computers.

Best regards