LCTII Endgame Test

Martin Hertz · Post by **Martin Hertz** » Sun Feb 26, 2017 1:09 pm

A small endgame test with older DOS programs, using the LCTII endgame positions at 3.3 GHz and not emulated:

                       FIN.01     FIN.02     FIN.03     FIN.04     FIN.05     FIN.06     FIN.07     FIN.08     FIN.09    Points

RexChess 2.30         ?? 00&#58;00   ?? 00&#58;00   ?? 00&#58;09   ?? 00&#58;06   ?? 00&#58;17   ?? 01&#58;05   ?? 00&#58;20   ?? --&#58;--   ??--&#58;--     190

Chess Champion 2175   ?? 00&#58;00   ?? 00&#58;00   13 00&#58;01   11 00&#58;05   14 --&#58;--   16 04&#58;50   13 03&#58;10   20 --&#58;--   19 --&#58;--    145

Fritz 1.00            13 00&#58;01   15 --&#58;--   20 --&#58;--   13 01&#58;45   12 00&#58;50   13 00&#58;13   12 01&#58;47   17 --&#58;--   17 --&#58;--    105

Chess Player 2150     08 00&#58;01   07 00&#58;03   10 --&#58;--   09 --&#58;--   09 15&#58;45   09 02&#58;30   08 --&#58;--   10 --&#58;--   10 --&#58;--     75

Turbo-Chess           13 00&#58;48   13 09&#58;30   13 00&#58;45   11 --&#58;--   11 11&#58;55   13 --&#58;--   10 08&#58;10   15 --&#58;--   12 --&#58;--     50

Sargon III            13 00&#58;30   13 --&#58;--   15 04&#58;30   12 --&#58;--   12 --&#58;--   13 --&#58;--   11 07&#58;25   16 --&#58;--   13 --&#58;--     35

Unfotunately I cannot see the reached ply of REX, becuase there is an issue regarding the display, but it did really well.
The positions 1-7 are too easy for later DOS programs. The position 9 is very hard but solvable for the most later progams.
Position 8 seems much too difficult and even not easy for nowadays engines. It's a joke that Turbo-Chess performs better
than Sargon III, If we take into consideration that Turbo-Chess is written in Pascal and has almost no special endgame
knowledge. Well done Kaare...

spacious_mind · Post by **spacious_mind** » Tue Feb 28, 2017 12:19 am

Hi Martin,

I just completed the test with Amiga running at 68060 with Sargon 3:

Test 1 = 0:0 minutes 30 Points
Test 2 = 0:0 minutes 30 Points
Test 3 = 0:27 minutes 25 Points
Test 4 = XXXXXX
Test 5 = 1:50 minutes 15 Points
Test 6 = 5:20 minutes 10 Points
Test 7 = XXXXXX 14:20 Minutes
Test 8 = XXXXXX
Test 9 = XXXXXX

Total: 110 points

It scored 110 points. I would say that the Amiga Sargon 3 version is better than the PC version.

Maybe I should also try Amiga Chess Player 2150 and Chess Champion 2175.

You could try my 5 rating test games with your Sargon 3 and see how it scores.

Best regards

Martin Hertz · Post by **Martin Hertz** » Tue Feb 28, 2017 12:15 pm

Thank you for testing Sargon on the 68060. If we compare the results of all the 9 positions, I would say,
that something went wrong while porting the code to the 8086. It's not only a question of speed, but it
seems that the Sargon III on PC is clearly not the same program. I didn't expect such differences for the
alleged same programs. I hope this is not also the case with the CP2150 or CC2175, but who knows.
I'm very disappointed from PC Sargon3 and will cancel any further testing on this porting failure.
If your tests on the Whittigton programs doesn't show similar issues, I can make some tests with your
rating games too. I've tested CC2175 with 8 MB hash (XMS needed).

Best regards

spacious_mind · Post by **spacious_mind** » Tue Feb 28, 2017 1:09 pm

Martin Hertz wrote:Thank you for testing Sargon on the 68060. If we compare the results of all the 9 positions, I would say,
that something went wrong while porting the code to the 8086. It's not only a question of speed, but it
seems that the Sargon III on PC is clearly not the same program. I didn't expect such differences for the
alleged same programs. I hope this is not also the case with the CP2150 or CC2175, but who knows.
I'm very disappointed from PC Sargon3 and will cancel any further testing on this porting failure.
If your tests on the Whittigton programs doesn't show similar issues, I can make some tests with your
rating games too. I've tested CC2175 with 8 MB hash (XMS needed).

Best regards

I agree, it is hard to say but if you look at Test 7 you show the double speed that favors your PC. But the PC version does seem to be missing a lot.

When I get a chance I will test 2150 and 2175. Hopefully Chris Whittington did a much better job converting them to PC.

Best regards

spacious_mind · Post by **spacious_mind** » Wed Mar 01, 2017 12:43 am

You know what I think may have happened is that the Sargon 3 6502 program that you had with Apple II, C64 and Atari 400/800 was used to convert to PC. Therefore the base program may have been close to what you have with Fidelity Avantgarde 2100 or Designer 2100, Par Excellence. Whereas the 68000 program was what was written for V2 onwards. Hence the PC conversion probably did not use the Sargon 3 68,000 version at all.

That might explain the two different Sargon 3's.

I don't know just throwing it out there.

Best regards

Nick

Martin Hertz · Post by **Martin Hertz** » Wed Mar 01, 2017 12:48 pm

Yes, I think this is a good explanation. The 68000 Sargon III uses hash tables and the PC version apparently not.
Also the chess knowledge seems to be significantly less. It's confusing to call them both Sargon III.

spacious_mind · Post by **spacious_mind** » Sat Mar 04, 2017 1:15 pm

I played a game with WinVice Amiga 68060 Sargon 3 against Fidelity V11. My computer runs at 4.3 GHz.

[Event "Computer Test Match"]
[Site "?"]
[Date "2017.03.03"]
[Round "1"]
[White "Amiga 68060 Sargon 3, 30S AV.."]
[Black "Fidelity V11 68060 72 MHz, 30S AV.."]
[Result "1-0"]
[ECO "A26"]
[WhiteElo "2334"]
[BlackElo "2334"]
[Annotator "SM"]
[PlyCount "113"]
[EventDate "2017.03.03"]
[EventType "match"]
[EventRounds "2"]

1. e4 c5 2. c3 Nf6 3. e5 Nd5 4. d4 cxd4 5. cxd4 {Fidelity V11 out of book} Nc6 {Amiga 68060 Sargon 3 out of book} 6. Nf3 d6 7. Bc4 Nb6 8. Bb5 dxe5 9. Bxc6+ bxc6 10. Nxe5 Ba6 11. Nc3 e6 12. Qf3 Qc7 13. Bf4 Nd5 14. Ng6 Nxf4 15. Nxh8 Nd3+ 16. Kd2 e5 17. Kc2 exd4 18. Qe4+ Be7 19. Qxd4 Kf8 20. Rad1 Rd8 21. Qe3 Kg8 22. Nxf7 Kxf7 23. Kb1 Kg8 24. Qe6+ Kh8 25. Rd2 Bb4 26. Rhd1 Bxc3 27. bxc3 Qa5 28. Qb3 Qe5 29. Qa3 Bc4 30. g3 Rb8+ 31. Ka1 a6 32. f4 Qd5 33. Qa4 Qe4 34. Qa5 c5 35. Qc7 Qe8 36. Qd6 Qb5 37. f5 Kg8 38. h4 h6 39. g4 Qb7 40. f6 Qb5 41. fxg7 Kxg7 42. g5 h5 43. Qh6+ Kg8 44. Qg6+ Kh8 45. Qxh5+ Kg7 46. Qh6+ Kg8 47. Qg6+ Kh8 48. Qf6+ Kg8 49. h5 Rb6 50. Qd8+ Kg7 51. Rf1 Bf7 52. h6+ Kg6 53. h7 Nf4 54. Qg8+ Bxg8 55. hxg8=Q+ Kf5 56. Qf7+ Kxg5 57. Qxf4+ 1-0

The game could have gone either way, but in the end V11 made the most crucial mistakes. My feeling is that both computer programs are equally strong when Sargon 3 is emulated to play with Motorola 68060 speed on my 4.3 GHZ computer.

Fidelity V11 is rated 2331 at 30 seconds per move active chess at Schachcomputer.Info. So Sargon 3 I think is similar in strength.

Well kudos to Selective Search for being one of the very few old publications reporting on old chess programs that were not dedicated computers.

Below is their rating list from December 1989:

Well it is better than nothing Sargon 3 (PC) showed up with a rating of 1692. It is hard to say if this is just the PC version or a mix of all the versions lumped into one. Especially when you look at Superchess that I think was only available on a Sinclair Spectrum or White Knight was only available on a BBC. But at least it was something.

For comparison to the above list, I played 38 games with Sargon 3 on C64 in a Commodore 64 (not emulators) tournament once and it performed at approx. ELO 1569. Colossus 4 which was only available on C64, Sinclair Spectrum, Atari 400 etc on C64 with 38 games scored ELO 1676 and Cyrus II with 88 games scored ELO 1525. Colossus 2 with 50 games scored ELO 1401, Chessmaster 2000 with 38 games scored 1552 ELO, Chessmaster 2100 with 38 games scored ELO 1621, Mychess with 38 games scored 1648 and Mastertronics Masterchess scored a measly 919 ELO in 50 Games.

Best regards

spacious_mind · Post by **spacious_mind** » Sat Mar 04, 2017 3:59 pm

I just played a second game between Amiga 68060 Sargon 3 against Fidelity 68060 V11.

[Event "Computer Test Match"]
[Site "?"]
[Date "2017.03.04"]
[Round "2"]
[White "Fidelity V11 68060 72 MHz, 30S AV.."]
[Black "Amiga 68060 Sargon 3, 30S AV.."]
[Result "0-1"]
[ECO "B79"]
[WhiteElo "2334"]
[BlackElo "2334"]
[Annotator "SM"]
[PlyCount "148"]
[EventDate "2017.03.04"]
[EventType "match"]
[EventRounds "2"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 g6 6. Be3 Bg7 7. f3 O-O 8. Qd2 Nc6 9. Bc4 Bd7 10. O-O-O Qa5 11. Bb3 Rfc8 12. Kb1 Ne5 13. Bg5 Nc4 14. Bxc4 Rxc4 15. Nb3 Qe5 16. Rhe1 Rxc3 17. bxc3 Be6 {Fidelity V11 out of book} 18. Bf4 {Amiga 68060 Sargon 3 out of book} Qb5 19. Bh6 Bxh6 20. Qxh6 a5 21. Ka1 Qc4 22. Re3 a4 23. Nc1 Ra5 24. Red3 Qc8 25. g4 Rc5 26. Qd2 Nd7 27. Re3 Ne5 28. Qe1 Rb5 29. Nd3 Nxf3 30. Rxf3 Bxg4 31. Rxf7 Kxf7 32. Rc1 Ke8 33. Rb1 Rxb1+ 34. Kxb1 Qc4 35. e5 Be6 36. Nb4 d5 37. Kb2 Qe4 38. Nd3 Qxe1 39. Nxe1 Bd7 40. Nd3 Kd8 41. Ka3 Kc7 42. Nc5 b5 43. Kb4 Be8 44. Ka5 Kc6 45. Ne6 Bd7 46. Nf8 Bf5 47. Nxh7 Kc5 48. Ng5 Kc4 49. Nf3 Bxc2 50. Nd4 Kxc3 51. Nxb5+ Kc4 52. Na3+ Kd3 53. Kb4 e6 54. Kc5 Bd1 55. Nb5 Bg4 56. Nd4 Ke4 57. Nc6 g5 58. a3 Bf5 59. Kb4 d4 60. Kc4 d3 61. Kc3 Kd5 62. Nb4+ Kxe5 63. Nxd3+ Bxd3 64. Kxd3 Kf4 65. Ke2 Ke4 66. Kf2 e5 67. Ke2 g4 68. Kf2 Kd3 69. h3 g3+ 70. Kxg3 e4 71. Kf2 Kd2 72. Kg2 e3 73. h4 e2 74. h5 e1=Q 0-1

Final Position

[fen]8/8/8/7P/p7/P7/3k2K1/4q3 w - - 0 75[/fen]

Amiga 68060 Sargon 3 wins again. Maybe it is a little better that V11.

Best regards

Martin Hertz · Post by **Martin Hertz** » Sat Mar 04, 2017 7:50 pm

Hi Nick,

according to Schachcomputer.Info the V11 72 MHz needs 256 seconds for the BT2450-4 test position. I've tried it on the
PC Sargon III and it founds the move Qf7 in ply 9 after 38 seconds. This is not bad, but much too slow for 3,3 GHz.
I'm pretty sure, that an accurate port of the 68000 Sargon to the PC would do it in a few seconds. Maybe your emulation
is even faster than the V11. Thanks for the interesting informations about the list from 1989.

Best regards

spacious_mind · Post by **spacious_mind** » Sat Mar 04, 2017 10:12 pm

Martin Hertz wrote:Hi Nick,

according to Schachcomputer.Info the V11 72 MHz needs 256 seconds for the BT2450-4 test position. I've tried it on the
PC Sargon III and it founds the move Qf7 in ply 9 after 38 seconds. This is not bad, but much too slow for 3,3 GHz.
I'm pretty sure, that an accurate port of the 68000 Sargon to the PC would do it in a few seconds. Maybe your emulation
is even faster than the V11. Thanks for the interesting informations about the list from 1989.

Best regards

Amga 68060 Sargon 3 does position 4 in 111 seconds so it seems faster than V11 72 MHz. If you assume they are similar programs then Sargon 3 would be approx. 68060 166 MHz.

Best regards

Martin Hertz · Post by **Martin Hertz** » Sun Mar 05, 2017 7:05 pm

I think it's difficult to say, if the programs are really identical. Maybe using another program like CC2175 or
CP2150 brings more clarity. I've also tried to test Sargon V, but it needs EMS for hashing and this is an issue
on many later PCs. Without EMS Sargon V needs 32 seconds and ply 9 for solving BT-2450-4, thats very close to
Sargon III. XMS is no problem, so I can execute CC2175 with 8MB hash. The most DOS programs using larger hash
don't need EMS or XMS. A good choice for testing a Whittington program is the BT2450-5 position with CP2150,
because there are no hash issues. CP2150 founds Ka6 in ply 9 after 30 seconds at 3.3 GHz.

Hashing regards

spacious_mind · Post by **spacious_mind** » Sun Mar 05, 2017 8:58 pm

Martin Hertz wrote:I think it's difficult to say, if the programs are really identical. Maybe using another program like CC2175 or
CP2150 brings more clarity. I've also tried to test Sargon V, but it needs EMS for hashing and this is an issue
on many later PCs. Without EMS Sargon V needs 32 seconds and ply 9 for solving BT-2450-4, thats very close to
Sargon III. XMS is no problem, so I can execute CC2175 with 8MB hash. The most DOS programs using larger hash
don't need EMS or XMS. A good choice for testing a Whittington program is the BT2450-5 position with CP2150,
because there are no hash issues. CP2150 founds Ka6 in ply 9 after 30 seconds at 3.3 GHz.

Hashing regards

Yes its hard to say. The programs are not 100% exact for sure. Amiga Sargon 3 completed BT-2450-4 in 111 seconds with a depth of 9/3. With my DB Sargon 5, it needed 320 seconds and it only found it 9/26. So the search is not the same between them. I am guessing at 30 seconds that Amiga Sargon 3 is still also about 1/2 ply faster than DOSBox Max and Dynamic setting on my computer. But I think in overall strength if both were capable of running at same speed, I would still consider Amiga Sargon 3 and PC Sargon 5 equally as strong.

Best regards

spacious_mind · Post by **spacious_mind** » Sun Mar 05, 2017 9:50 pm

I just ran Amiga CP2150 through BT-2450-4 test position. It can't do it 30 minutes and still running. CP2150 on Amiga can't find it.

Best regards

Martin Hertz · Post by **Martin Hertz** » Mon Mar 06, 2017 4:24 pm

Yes, the BT2450-4 ist too difficult, because it needs around 10 minutes at 3.3 GHz. This was the reason why
I tried BT2450-5 instead with 30 seconds in ply 9. It should take only a few minutes with your 68060 emu and
around 5 minutes with DOSbox, I guess.

Best reagrds