Your Country Needs You!

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
Post Reply
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Here is an update with results from computers that I have tested so far on all 5 test games. Out of curiosity I added a Gavon chess program named CPW v. 1.1 written by PAWEL KOZIOL & EDMUND MOSHAMMER. Gavon rates this program at ELO 1997.

Image

From my tests you can see that I score the program much higher at a rating of 2498.

Well as a test I set up Mephisto Berlin 68000 this morning, which per Active rating list has a rating of ELO 2182. With this kind or rating Berlin should have no problems beating CPW v. 1.1 if I were to take the results of rating lists at face value.

Here is the game:

[Event "Computer Test Match"]
[Site "USA"]
[Date "2015.02.22"]
[Round "?"]
[White "Mephisto Berlin 68000, LV 4 30S."]
[Black "Gavon CPW V.1.1, AT30."]
[Result "0-1"]
[WhiteElo "2182"]
[BlackElo "1997"]
[PlyCount "86"]

1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6. Be2 {Gavon CPW V.1.1 out
of book} Bg4 7. Be3 Nc6 {Mephisto Berlin 68000 out of book} 8. h3 Bxf3 9. Bxf3
Nd7 10. O-O e5 11. d5 Nd4 12. Be2 Qh4 13. f3 f5 14. b4 a5 15. Rb1 fxe4 16. Nxe4
axb4 17. Rxb4 Rxa2 18. Bxd4 exd4 19. Rxb7 Rc8 20. Rf2 Be5 21. Kf1 Nc5 22. Nxc5
dxc5 23. Qb1 Rca8 24. Rb8+ Rxb8 25. Qxb8+ Kg7 26. Qb3 Qg5 27. Bd1 Rxf2+ 28.
Kxf2 Bg3+ 29. Ke2 Bh4 30. g4 Qf4 31. Qa3 Bg5 32. Kf1 Qh2 33. Bb3 Qxh3+ 34. Kf2
Bh4+ 35. Ke2 {Gavon CPW V.1.1 announces Mate-in-10} Qg2+ 36. Kd3 Qf1+ 37. Kc2
d3+ 38. Kc3 Bf6+ 39. Kd2 Bg5+ 40. f4 Bxf4+ 41. Kc3 Qe1+ 42. Kxd3 Qd2+ 43. Ke4
Qe3# 0-1

The game was a lot of fun and very active, which means that I am probably going to enjoy play CPW v. 1.1 a lot. But as you can see, even though CPW v. 1.1 missed several opportunities to win the game faster, the program really had no problem beating Mephisto Berlin in this exciting game.

So, what is right? is ELO 1997 correct?, or is Mephisto Berlin at 2182 correct? or is my rating of ELO 2498 too high? I personally suspect the answer lies somewhere in the middle. I also suspect that CPW v. 1.1 has the ability to give an R30 King or Gideon or a Mephisto Tournament machine a good challenge.

Hopefully this example shows more clearly why I am driven with my tests. There is nothing currently available that can compare and keep the integrity and perspective of relative strengths from 0 ELO to TOP program strength of 3300+ as shown in CCRL 40/40, whether you are human, an IPad program, Android, Pocket PC, DOS, Windows or dedicated program. I believe these tests will keep the perspectives better than any other test or rating system that is currently available and that is the goal that is driving me :wink:

Best regards
Nick
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Hi Nick,

As I am not (yet) able to post a picture, I will send you tomorrow my detailed testresults by email.
But I can give here the total results and the resulting rating for the
SAITEK COUGAR (LEVEL A8 30 SECONDS AVERAGE):
game1: 2382, game2: 2087, game3: 2403, game4: 2250, game5: 2089 >>RATING: 2242

A few remarks resulted from observing the machine computing the moves:
1. I had the feeling that overall it used more then 30 secs per move. For instance with game4 I noticed it used a few minutes more time for white than for black. Exact figures are hard to give, you allways lose some time taking back moves and moving again.
2. With game4, at move 19 it played white Ra2, but that didn't exist in your movelist, so I put Ra3 there instead. That took me some time too of course.
3. Given the results of the machines you tested, I am amazed about this score for the Cougar, comparing it to the Saitek Risc 2500, because on www.schach-computer.info (the wiki elo liste) the Risc 2500 is rated about 200 elo points higher than the Cougar. On the other hand, on the same list the Legend/Concerto is rated 200 elo points less, which is the same outcome as this test.
4. In game1, it moves 18. .. Rg8 for black. I wonder if that is a good move, or that it is caused by losing the right for O-O by all the move forward and take back actions.

All in all, it is indeed a lot of fun doing these tests! And now it comes to me that in the dutch chess association (KNSB) monthly paper, also these type of tests are given for human players. It will be interesting to try some of those tests and compare the outcome!

Late night and bedtime regards (00:10 over here),
Paul.
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

paulwise3 wrote:Hi Nick,

As I am not (yet) able to post a picture, I will send you tomorrow my detailed testresults by email.
But I can give here the total results and the resulting rating for the
SAITEK COUGAR (LEVEL A8 30 SECONDS AVERAGE):
game1: 2382, game2: 2087, game3: 2403, game4: 2250, game5: 2089 >>RATING: 2242

A few remarks resulted from observing the machine computing the moves:
1. I had the feeling that overall it used more then 30 secs per move. For instance with game4 I noticed it used a few minutes more time for white than for black. Exact figures are hard to give, you allways lose some time taking back moves and moving again.
2. With game4, at move 19 it played white Ra2, but that didn't exist in your movelist, so I put Ra3 there instead. That took me some time too of course.
3. Given the results of the machines you tested, I am amazed about this score for the Cougar, comparing it to the Saitek Risc 2500, because on www.schach-computer.info (the wiki elo liste) the Risc 2500 is rated about 200 elo points higher than the Cougar. On the other hand, on the same list the Legend/Concerto is rated 200 elo points less, which is the same outcome as this test.
4. In game1, it moves 18. .. Rg8 for black. I wonder if that is a good move, or that it is caused by losing the right for O-O by all the move forward and take back actions.

All in all, it is indeed a lot of fun doing these tests! And now it comes to me that in the dutch chess association (KNSB) monthly paper, also these type of tests are given for human players. It will be interesting to try some of those tests and compare the outcome!

Late night and bedtime regards (00:10 over here),
Paul.
Hi Paul,

Thanks, yes please send me the moves of the tests and I will update the spreadsheet and repost it.

Below is the result for Saitek Cougar. The way I rate is different to just dividing 5 games total by five, because I take into consideration every single move that was played therefore the rating is different and normally lower than dividing by 5.

Image

Yes some computers are performing better than expected and some worse. For example Glass Chess should be lower too. But I think as more games are added through the Generations there will be more complex games as well and the ratings will balance out. In game 4 for example who would have thought that Mephisto Berlin would only score 1576? A bad test game for Lang's I guess.

I will recheck game 4 to see if I missed something and repost the test sheet, thanks for spotting any errors.

In game 1 18. ... Rg8 is not a good move :)

Best regards
Nick
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Hi Nick, as you will have noticed I mailed you the details of the Cougar test.
Today I started testing the CXG Sphinx Comet, and because of the amazing result in the first testgame I wanted it to share right away: With white it scored 2632 elo points...!!
COMPUTER TEST RESULT
Sphinx
Comet
Kaare Danielsen
Level A6 - 30 seconds average - Normal
Hitachi HD6301 - 8 Bit - 1.79 Mhz - 4 KB ROM - 256 Bytes RAM
WHITE MOVES
5.Nf3
6.0-0
7.Qxf3
8.Qxb7
9.Nb5
10.Nxa7
11.Qc6+
12.0-0
13.Qc6+
14.Qxb5+
15.0-0
16.Qxd5
17.Qxd5
18.Rxd5
19.Rg5
20.Qxd7+
21.Qd8+
WHITE SCORE:
2632
No wonder the Concerto has problems with it ;-)
With black it scored a modest 1408 elo, so the total for game 1 is 2020 elo.

Also, I tried the Agressive and Defensive styles. With Agressive the white score is the same, and the black score slightly better: 1416 elo. With Defensive it scores also the same with white, but a poor 1276 with black...

I'm not sure I have the patience for testing these styles also for the other games. But we'll see ;-)

Comet regards,
Paul.
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

paulwise3 wrote:Hi Nick, as you will have noticed I mailed you the details of the Cougar test.
Today I started testing the CXG Sphinx Comet, and because of the amazing result in the first testgame I wanted it to share right away: With white it scored 2632 elo points...!!
COMPUTER TEST RESULT
Sphinx
Comet
Kaare Danielsen
Level A6 - 30 seconds average - Normal
Hitachi HD6301 - 8 Bit - 1.79 Mhz - 4 KB ROM - 256 Bytes RAM
WHITE MOVES
5.Nf3
6.0-0
7.Qxf3
8.Qxb7
9.Nb5
10.Nxa7
11.Qc6+
12.0-0
13.Qc6+
14.Qxb5+
15.0-0
16.Qxd5
17.Qxd5
18.Rxd5
19.Rg5
20.Qxd7+
21.Qd8+
WHITE SCORE:
2632
No wonder the Concerto has problems with it ;-)
With black it scored a modest 1408 elo, so the total for game 1 is 2020 elo.

Also, I tried the Agressive and Defensive styles. With Agressive the white score is the same, and the black score slightly better: 1416 elo. With Defensive it scores also the same with white, but a poor 1276 with black...

I'm not sure I have the patience for testing these styles also for the other games. But we'll see ;-)

Comet regards,
Paul.
Hi Paul,

I had also done Sphinx Comet. but not posted it because of not having completed all 5 games. Below are my results with the different styles for Game 1. It will be interesting for you to compare yours and see if there is much randomness in the moves:

Code: Select all

COMPUTER TEST RESULT	COMPUTER TEST RESULT	COMPUTER TEST RESULT	COMPUTER TEST RESULT
CXG	CXG	CXG	CXG
SPHINX COMET CXG – 902 NORMAL	SPHINX COMET CXG – 902 AGGRESSIVE	SPHINX COMET CXG – 902 DEFENSIVE	SPHINX COMET CXG – 902 RANDOM
KAARE DANIELSEN	KAARE DANIELSEN	KAARE DANIELSEN	KAARE DANIELSEN
30S – LEVEL A6 NORMAL	30S – LEVEL C6 AGGRESSIVE	30S – LEVEL E6 DEFENSIVE	30S – LEVEL G6 RANDOM
HITACHI HD6301 8 BIT 1.78 MHZ  4 KB RAM 256 BYTES	HITACHI HD6301 8 BIT 1.78 MHZ  4 KB RAM 256 BYTES	HITACHI HD6301 8 BIT 1.78 MHZ  4 KB RAM 256 BYTES	HITACHI HD6301 8 BIT 1.78 MHZ  4 KB RAM 256 BYTES
WHITE MOVES	WHITE MOVES	WHITE MOVES	WHITE MOVES
5.d4	5.Nf3	5.d4	5.Nf3
6.d4	6.0-0	6.0-0	6.0-0
7.Qxf3	7.Qxf3	7.Qxf3	7.Qxf3
8.Qxb7	8.Qxb7	8.Qxb7	8.Qxb7
9.Nb5	9.Nb5	9.Nb5	9.Nb5
10.Nxa7	10.Nxa7	10.Nxa7	10.Nxa7
11.Qc6+	11.Qc6+	11.Qc6+	11.Qc6+
12.0-0	12.d4	12.Qb5+	12.0-0
13.Qc6+	13.Qc6+	13.Qc6+	13.Qc6+
14.Qxb5+	14.Qxb5+	14.Qxb5+	14.Qxb5+
15.0-0	15.0-0	15.0-0	15.0-0
16.Qxd5	16.Qxd5	16.Qxd5	16.Qxd5
17.Qxd5	17.Qxd5	17.Qxd5	17.Qxd5
18.Rxd5	18.Rxd5	18.Rxd5	18.Rxd5
19.Rg5	19.Rg5	19.Rg5	19.Rg5
20.Qxd7+	20.Qxd7+	20.Qxd7+	20.Qxd7+
21.Qd8+	21.Qd8+	21.Qd8+	21.Qd8+
WHITE SCORE:	WHITE SCORE:	WHITE SCORE:	WHITE SCORE:
2632	2596	2468	2632
BLACK MOVES	BLACK MOVES	BLACK MOVES	BLACK MOVES
4. ... e5	4. ... e5	4. ... e5	4. ... e5
5. ... Be6	5. ... Qd6	5. ... Be6	5. ... Be6
6. ... Be6	6. ... Bf5	6. ... Be6	6. ... Be6
7. ... Nc6	7. ... Nc6	7. ... Nc6	7. ... Nc6
8. ... Nbd7	8. ... Nbd7	8. ... Nbd7	8. ... Nbd7
9. ... Bd6	9. ... Bd6	9. ... Bd6	9. ... Bd6
10. ... Rb8	10. ... Rb8	10. ... Rb8	10. ... Rb8
11. ... Qxc8	11. ... Qxc8	11. ... Qxc8	11. ... Qxc8
12. ... Nd6	12. ... Nd6	12. ... Nd6	12. ... Nb6
13. ... Ke7	13. ... Ke7	13. ... Ke7	13. ... Ke7
14. ... Qd7	14. ... Qd7	14. ... Qd7	14. ... Qd7
15. ... exd5	15. ... exd5	15. ... exd5	15. ... exd5
16. ... Qf6	16. ... Qf6	16. ... Qb8	16. ... Qb8
17. ... c6	17. ... c6	17. ... c6	17. ... c6
18. ... Qg6	18. ... Qg6	18. ... Qg6	18. ... Qg6
19. ... Qxg2	19. ... Qe4+	19. ... Qxg2	19. ... Qxg2
20. ... Kf8	20. ... Kf8	20. ... Kf8	20. ... Kf8
BLACK SCORE:	BLACK SCORE:	BLACK SCORE:	BLACK SCORE:
1276	1464	1276	1216
TOTAL SCORE:	TOTAL SCORE:	TOTAL SCORE:	TOTAL SCORE:
1954	2030	1872	1924
best regards
Nick
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Hi Nick,
Ok, there seems to be some randomness, but I think that should go for every machine, at least it should be. I did testgame 1 for the Concerto, and there it made only one move different from your test. But the bottomline is that the Comet does very well with white in testgame 1. What strikes me, is that you have a lot of differences in the white moves for the various styles, and I have hardly one. Could there be a difference in how the starting position for the test is reached? For instance, you may have used the multimove level to do the starting moves. I just let it play and corrected the moves if they were different. Once it is out of openingbook, after a few seconds I press the Move button. If the move differs, I correct it. And so on, until the starting position is reached. I admit it is clumsy, but I do not have to change level then.
Here are my moves for the Normal, Aggressive and Defensive styles.

Code: Select all

COMPUTER TEST RESULT	COMPUTER TEST RESULT	COMPUTER TEST RESULT
Sphinx	Sphinx	Sphinx
Comet	Comet	Comet
Kaare Danielsen	Kaare Danielsen	Kaare Danielsen
Level A6 - 30 seconds average - Normal	Level C6 - 30 seconds average - Agressive	Level E6 - 30 seconds average - Defensive
Hitachi HD6301 - 8 Bit - 1.79 Mhz - 4 KB ROM - 256 Bytes RAM	Hitachi HD6301 - 8 Bit - 1.79 Mhz - 4 KB ROM - 256 Bytes RAM	Hitachi HD6301 - 8 Bit - 1.79 Mhz - 4 KB ROM - 256 Bytes RAM
WHITE MOVES	WHITE MOVES	WHITE MOVES
5.Nf3	5.Nf3	5.d4
6.0-0	6.0-0	6.0-0
7.Qxf3	7.Qxf3	7.Qxf3
8.Qxb7	8.Qxb7	8.Qxb7
9.Nb5	9.Nb5	9.Nb5
10.Nxa7	10.Nxa7	10.Nxa7
11.Qc6+	11.Qc6+	11.Qc6+
12.0-0	12.0-0	12.0-0
13.Qc6+	13.Qc6+	13.Qc6+
14.Qxb5+	14.Qxb5+	14.Qxb5+
15.0-0	15.0-0	15.0-0
16.Qxd5	16.Qxd5	16.Qxd5
17.Qxd5	17.Qxd5	17.Qxd5
18.Rxd5	18.Rxd5	18.Rxd5
19.Rg5	19.Rg5	19.Rg5
20.Qxd7+	20.Qxd7+	20.Qxd7+
21.Qd8+	21.Qd8+	21.Qd8+
WHITE SCORE:	WHITE SCORE:	WHITE SCORE:
2632	2632	2632
BLACK MOVES	BLACK MOVES	BLACK MOVES
4. ... e5	4. ... e5	4. ... Be6
5. ... Nc6	5. ... Nc6	5. ... Be6
6. ... Be6	6. ... Bf5	6. ... Be6
7. ... Nc6	7. ... Nc6	7. ... Nc6
8. ... Nbd7	8. ... Nbd7	8. ... Nbd7
9. ... Bd6	9. ... Bd6	9. ... Bd6
10. ... Rb8	10. ... Rb8	10. ... Rb8
11. ... Qxc8	11. ... Qxc8	11. ... Qxc8
12. ... Nd6	12. ... Nd6	12. ... Nd6
13. ... Ke7	13. ... Ke7	13. ... Ke7
14. ... Qd7	14. ... Qd7	14. ... Qd7
15. ... exd5	15. ... exd5	15. ... exd5
16. ... Qb8	16. ... Qf6	16. ... Qb8
17. ... c6	17. ... c6	17. ... c6
18. ... Qg6	18. ... Qg6	18. ... Qg6
19. ... Qxg2	19. ... Qxg2	19. ... Qxg2
20. ... Kf8	20. ... Kf8	20. ... Kf8
BLACK SCORE:	BLACK SCORE:	BLACK SCORE:
1408	1416	1276
TOTAL SCORE:	TOTAL SCORE:	TOTAL SCORE:
2020	2024	1954
Regards, Paul.
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

paulwise3 wrote:Hi Nick,
Ok, there seems to be some randomness, but I think that should go for every machine, at least it should be. I did testgame 1 for the Concerto, and there it made only one move different from your test. But the bottomline is that the Comet does very well with white in testgame 1. What strikes me, is that you have a lot of differences in the white moves for the various styles, and I have hardly one. Could there be a difference in how the starting position for the test is reached? For instance, you may have used the multimove level to do the starting moves. I just let it play and corrected the moves if they were different. Once it is out of openingbook, after a few seconds I press the Move button. If the move differs, I correct it. And so on, until the starting position is reached. I admit it is clumsy, but I do not have to change level then.
Here are my moves for the Normal, Aggressive and Defensive styles.

Regards, Paul.
Hi Paul,

Yes I use Multimove whenever that is available for the computers, so maybe that creates a little variation. But no matter multiple tests just means that the scores across the tests can be averaged out.

I just played another Gavon test, this time with Gavon JFresh v. 0.1a written by Christian Daley and rated by Gavon as 1537 ELO. So far I played two test games and it scored:

Gavon JFresh v. 0.1a


Game 1 - 2278
Game 2 - 1948
-----------------
AV: 2096 for two games

I then compared this to CXG Sphinx Legend.

CXG Sphinx Legend

Game 1 - 2228
Game 2 - 2068
-----------------
AV: 2140 for two games

After two tests the rating is in favor of CXG Sphinx Legend by 46 points. Therefore instead of playing the other 3 tests, I felt like playing them against each other in a game. Here is the game:

[Event "Computer Test Match"]
[Site "Alabama"]
[Date "2015.02.22"]
[Round "?"]
[White "CXG Sphinx Legend, LV 53 30S."]
[Black "Gavon JFresh v. 0.1a, AT30."]
[Result "1-0"]
[ECO "E15"]
[WhiteElo "1792"]
[BlackElo "1537"]
[PlyCount "83"]
[EventCountry "USA"]

1. d4 Nf6 2. c4 e6 3. Nf3 b6 4. g3 Ba6 5. b3 Bb4+ 6. Bd2 {Gavon JFresh v. 0.1a
out of book} Bxd2+ {CXG Sphinx Legend out of book} 7. Nbxd2 d6 8. e3 Nc6 9. Bg2
O-O 10. Ne5 dxe5 11. Bxc6 exd4 12. Bxa8 Qxa8 13. O-O d3 14. Qb1 Rd8 15. Qb2 Bb7
16. Qe5 c5 17. Rad1 Bg2 18. Rfe1 Bc6 19. e4 Rd7 20. f3 a6 21. Re3 Rd4 22. Qc7
b5 23. Qe7 Nd7 24. Nf1 bxc4 25. bxc4 Qf8 26. Qxf8+ Kxf8 27. Rdxd3 Ne5 28. Rxd4
cxd4 29. Ra3 Nxc4 30. Rxa6 Bd7 31. Ra7 Bc6 32. Kf2 d3 33. Rc7 Ne5 34. f4 Ng4+
35. Kf3 Nxh2+ 36. Nxh2 d2 37. Ke2 Bxe4 38. Kxd2 h5 39. a4 Ke8 40. a5 Bg6 41. a6
Be4 42. a7 1-0

CXG Sphinx Legend wins the game. But what an exciting game it was as well. The style setting was the standard style 5. Gavon's opening book is interesting there are 20 opening books. For my tests I use the book ALL (10) which goes 10 half moves deep, meaning after move 5 the computer is out of book. I don't like opening books to go too deep as it often gives a computer with the bigger opening an advantage and I am not really that interested in artificial advantages.

•alekhine.
•botvinnik.
•capablanca.
•carlsen..
•euwe.
•fischer.
•karpov.
•kasparov.
•keres.
•lasker.
•morphy.
•petrosian.
•smyslov.
•spassky.
•steinitz.
•tal.
•all.
•all(10).
•all(20).
•all(30).

Anyway the opening was wild and the game started wild. But CXG Legend a supposedly passive player coped well and did not look too passive in this game.

Interesting stuff regards,
Nick
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

I just completed a second game between Gavon JFresh v. 0.1a and CXG Sphinx Legend.

[Event "Computer Test Match"]
[Site "Alabama"]
[Date "2015.02.22"]
[Round "?"]
[White "Gavon JFresh v. 0.1a, AT30."]
[Black "CXG Sphinx Legend, LV 53 30S."]
[Result "1/2-1/2"]
[ECO "B42"]
[WhiteElo "1537"]
[BlackElo "1792"]
[PlyCount "97"]
[EventCountry "USA"]

1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 a6 5. Bd3 Nf6 {Gavon JFresh v. 0.1a out
of book} 6. Nc3 {CXG Sphinx Legend out of book} Nc6 7. Nxc6 dxc6 8. Be3 e5 9.
Qd2 Be6 10. O-O-O Be7 11. Qe2 Qa5 12. Bc4 Bxc4 13. Qxc4 O-O 14. Kb1 Qb4 15. Qb3
Rfd8 16. g3 Ng4 17. Rxd8+ Rxd8 18. Ka1 Nxe3 19. fxe3 Qxb3 20. cxb3 Rd3 21. Re1
Bb4 22. Rd1 Rxe3 23. a3 Bf8 24. Rd7 b5 25. Ra7 Re1+ 26. Ka2 c5 27. Ra8 Rh1 28.
Nd5 f6 29. Nc7 Kf7 30. Rxa6 b4 31. h4 bxa3 32. bxa3 Rh2+ 33. Ka1 Rg2 34. Nd5
Rxg3 35. a4 Rxb3 36. Ra7+ Kg8 37. Ka2 Rh3 38. a5 Rxh4 39. a6 Rxe4 40. Kb3 Rd4
41. Rd7 e4 42. a7 c4+ 43. Kc2 Rd3 44. a8=Q e3 45. Ne7+ Kf7 46. Nd5+ Kg8 47.
Ne7+ Kf7 48. Nd5+ Kg8 49. Ne7+ {CXG Sphinx Legend announces draw by repetition.
} 1/2-1/2

It was quite interesting until the ending where Gavon JFresh v. 0.1a did not have a clue, playing draw by repetition in an easily won Endgame. But CXG Sphinx Legend was also not much better in the endgame.

[fen]5bk1/3R2pp/P4p2/2pNp3/3r4/1K6/8/8 w - - 0 41[/fen]

In the above position with CXG Legend to move, the move 41. ... c4+ would have won the game for Legend. 41. ... e4? was just wrong.

Well a lucky draw in the end for CXG Legend, but only because JFresh does not have a clue in endgames.

Best regards
Nick
Carl Bicknell
Member
Posts: 252
Joined: Mon May 04, 2009 10:06 pm

Post by Carl Bicknell »

Your Tasc CM King 2.54 - how are you running that? As an internal card in an old PC?

Also what is it running at 16 MHz or 30 MHz?

I'd love to see one of these in action!
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Busy testing the Spinx Comet. Disappointing is that in game 3 it misses 22. Db3+ (mate in two). The first level it sees it is A8 (60 secs average) ...
Still I like the program. Could it be possible to port it to a much faster processor? Now it runs at 1.79 MHz.
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Carl Bicknell wrote:Your Tasc CM King 2.54 - how are you running that? As an internal card in an old PC?

Also what is it running at 16 MHz or 30 MHz?

I'd love to see one of these in action!
Hi Carl,

The CM512KB cards run at 30 MHz but supposedly you have to half the tact therefore it processes at 15 mhz.

The 1024KB cards are 14 and 15 Mhz, these are actually slightly less strong as the 512KB card. There is also a 128KB card and finally another 512kB card with 32MHz were the processor has a gold label. This one would technically be the strongest but hard to find. These are also all internal cards.

I also have a couple of external CM's but they don't function too well, the internal cards are much better.

Overall though there is really not too much difference in the playing strength of any of these cards. They are all strong and they all accept every available King or Schroeder program that was made for the CM.

I also have an external Amiga Chessmachine card which works but I cannot use it at the moment because the original disk is corrupt so I am searching for a copy of the 3.1/2 Amiga install disk. That one is also 1024 KB.

The one I am using for this test is the 512KB at 30 MHz (15).

I use all of the above cards on different old DOS computers that I have so that I can play them against each other as well.

You need ISA slots in your computer to run the internal cards which obviously modern computers no longer have.

Best regards
Nick
User avatar
paulwise3
Senior Member
Posts: 1505
Joined: Tue Jan 06, 2015 10:56 am
Location: Eindhoven, Netherlands

Post by paulwise3 »

Hi Nick,
I just finished the 5 testgames for the Sphinx Comet.
Resulting elo's game 1..5 divided by 5 gives
(2020+1695+1655+1716+1620)/5 = 1741,2

From several moves it became clear that it is too slow to compute deep variants, and suffers from horizon effect. I guess it is about 3 to 4 ply. Maybe even less, as it missed a mate in 2. But then still the results are very good, given an official elo below 1500.
I will send you the details tomorrow.

Regards, Paul
Brian B
Member
Posts: 74
Joined: Mon Jun 09, 2014 10:37 pm

Post by Brian B »

First off, I really enjoy reading about these tests and the relative performance of these computers against each other and the occasional human. It is an incredible amount of work and I do appreciate the effort.

I was wondering if there should be some way to weight certain moves at a higher level? Granted, if a computer makes a bonehead move, it will pay the price with a very low rating. It seem to me that a rating of zero wouldn't penalize a computer enough if this move came at a critical point in the game. How many games are decided not by great moves, but by one costly mistake? These results for these computers are relative to each other, yet other than having all these computers play games against other computers, I wish there was a way to identify a key move and grade accordingly.

Regards,
Brian B
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Brian B wrote:First off, I really enjoy reading about these tests and the relative performance of these computers against each other and the occasional human. It is an incredible amount of work and I do appreciate the effort.

I was wondering if there should be some way to weight certain moves at a higher level? Granted, if a computer makes a bonehead move, it will pay the price with a very low rating. It seem to me that a rating of zero wouldn't penalize a computer enough if this move came at a critical point in the game. How many games are decided not by great moves, but by one costly mistake? These results for these computers are relative to each other, yet other than having all these computers play games against other computers, I wish there was a way to identify a key move and grade accordingly.

Regards,
Brian B
Hi Brian,

Each move is weighted from very bad to very good based on a consistent evaluation formula not changed by a human. The problem with a human adding extra weight beyond a consistent formula for certain moves is that now you are being biased again because a human's opinion differs and does not treat everything equal. Consistent formulas however do.

My rating really does not care who is being evaluated and my opinion on what is good or bad is irrelevant, since my opinion is not a factor in the evaluation.

Best regards
Nick
Brian B
Member
Posts: 74
Joined: Mon Jun 09, 2014 10:37 pm

Post by Brian B »

spacious_mind wrote:
Brian B wrote:First off, I really enjoy reading about these tests and the relative performance of these computers against each other and the occasional human. It is an incredible amount of work and I do appreciate the effort.

I was wondering if there should be some way to weight certain moves at a higher level? Granted, if a computer makes a bonehead move, it will pay the price with a very low rating. It seem to me that a rating of zero wouldn't penalize a computer enough if this move came at a critical point in the game. How many games are decided not by great moves, but by one costly mistake? These results for these computers are relative to each other, yet other than having all these computers play games against other computers, I wish there was a way to identify a key move and grade accordingly.

Regards,
Brian B
Hi Brian,

Each move is weighted from very bad to very good based on a consistent evaluation formula not changed by a human. The problem with a human adding extra weight beyond a consistent formula for certain moves is that now you are being biased again because a human's opinion differs and does not treat everything equal. Consistent formulas however do.

My rating really does not care who is being evaluated and my opinion on what is good or bad is irrelevant, since my opinion is not a factor in the evaluation.

Best regards
Hi Nick, thanks for your thoughts.

I wasn't thinking of a human evaluation, I was thinking of a computer evaluation of the resulting position after a given move. So, not an evaluation of the move, but of the position after the move. Not sure how it could be done, but if there were a way to rate the actual position after a move as either winning or losing it would be interesting. This is being done now to some extent by rating the move itself, yet I don't see a true order of magnitude. How bad is a really bad move? I understand that this idea isn't very practical, just food for thought.

Regards,
Brian B
Post Reply