Playing Ron Nelson Programs, Today

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
fern
Posts: 8755
Joined: Sun Feb 26, 2006 3:07 pm

Playing Ron Nelson Programs, Today

Post by fern » Sun Apr 08, 2007 10:43 pm

I mean, playing machines like Excalibur Grandmaster, Alexandra, etc... All of them has been programmed by Ron Nelson, to date certainly worthy of a place in the Hall of Fame not only because he was the first or one of the first to produce these contrivances, but because he is the only one coming from the 70's still producing this kind of stuff.
Incredible endurance and determination.
Now, what happens when you play his current creations?
First, you face programs that in positional terms, opening terms, understanding of the game, etc, are no better or scarcely better than 70's-80's machines. That means you face an opponent that get out of the opening always inferior. sometimes to an almost laughable level. It means the machine will play preposterous moves, completely alien to what is going on in the board. And so ity happens that you will be fairly superior very soon.
So you get at the same time irritated because of losing time playing such a silly player, relaxed because of his idiotic moves, sure you will win just pushing the pieces.
BUT...
It happens also that the Nelson program, although positionally a moron, has some tactical improvements in his code. Nelson no doubt has learned something from his colleagues, has added something along the years And has at his disposal far faster processors.
The output?
The game you considered already won get complicated. You play sub-optimal or straight silly moves, fall in 2 moves traps and at last you feel as playing a complete patzer but at the same time one gifted with great calculation power. You play badly, you does not win, you draw and even sometimes you lose because you end playing at 1/3 of your real.
A really unpleasant experience and a real chance to learn how to win patzers.

Draw with GraNDMASTER regards
Fernando

Larry
Posts: 840
Joined: Sun Mar 12, 2006 2:59 am
Location: Sydney

Re: Playing Ron Nelson Programs, Today

Post by Larry » Sun Apr 08, 2007 11:27 pm

The first chess computer I ever played was a Fidelity '8' level, and
it was a Ron Nelson program, although I did'nt know or care who
programmed the thing at the time. It was a "chess machine" to me,
and an immediate fascination. I currently own several machines
programmed by Nelson, but the Excalibur Grandmaster is not one
of them. I'm not at all attracted to the Grandmaster, seems like a
70's computer to me. No lights, you have to look at the display to
see what it's doing.
Nothing against Nelson of course, he's in business and he's
catering to what the market wants, ie, chess comps in the 1800-
1900 range.
"It happens also that the Nelson program, although positionally a moron, has some tactical improvements in his code"


I suspect the Grandmaster program is just a
jazzed up Fidelity '8' level, tweaked but not rewritten, and running
on much faster hardware. Frankly I did'nt even know it was a
Nelson program until you said, and I checked on Wiki. Even the
simplest program running fast enough will be a challenge for a
lot of human players.
I have to admit, even playing against one of Nelson's "morons"
I take my eye off the ball and lose, especially if I'm in a
conversation with someone and dividing my attention.

Larry

JonP01

Re: Playing Ron Nelson Programs, Today

Post by JonP01 » Mon Apr 09, 2007 12:42 am

I don't know that the original post is a completely accurate account of the latest Nelson programs - certainly my experience and those of others is quite different to that. Infact, bugs and hardware issues apart (see my seperate thread about the extremely irritating problems with my DTTC for example), the actual chess playing capabilities of his machines are vastly superior to those of the 1980s. There is no comparison at all in my experience.

As far as the Grandmaster program is concerned, it is actually quite a competent program. It is not simply a rehash of an older program but a completely new (for it's time) program written for the H8 processor. The only commonality between it and the older Fidelity 4K programs is the author. In a database of 166 x 40 moves in 2 games I possess, Grandmaster has infact acquitted itself quite well. Here are the results it has achieved against other dedicated machines:


(9) Grandmaster : 166 (+ 72,= 30,- 64), 52.4 %

Designer Display 2100 : 10 (+ 2,= 4,- 4), 40.0 %
Mephisto MM IV : 10 (+ 4,= 1,- 5), 45.0 %
Super Constellation : 4 (+ 1,= 1,- 2), 37.5 %
Elegance 3MHz : 10 (+ 4,= 3,- 3), 55.0 %
Europa A : 10 (+ 5,= 3,- 2), 65.0 %
Simultano : 16 (+ 7,= 2,- 7), 50.0 %
Advanced Star Chess : 10 (+ 8,= 2,- 0), 90.0 %
Express 16K : 10 (+ 10,= 0,- 0), 100.0 %
Miami : 10 (+ 1,= 1,- 8), 15.0 %
Jade : 10 (+ 2,= 2,- 6), 30.0 %
Turbostar Kasparov : 10 (+ 6,= 1,- 3), 65.0 %
Conchess 2MHz : 10 (+ 5,= 0,- 5), 50.0 %
Cosmos : 10 (+ 1,= 3,- 6), 25.0 %
Primo : 10 (+ 7,= 2,- 1), 80.0 %
Chess 2001 : 10 (+ 6,= 1,- 3), 65.0 %
Amber : 16 (+ 3,= 4,- 9), 31.2 %

As you can see, there are some impressive results there - a 40% score against the Designer 2100, a 55% score against Fidelity Elegance, a 45% score against Ed Schroder's MMIV and a 65% result against the Kasparov Turbostar represent very impressive results. The above result table was good enough to achieve a 1779 rating for Grandmaster in this database of games. Remember these opponents were highly competant machines back in their day. They were certainly whole classes and then some apart from the Nelson machines of the era - i.e Sensory 8, Gambit, Poppy, etc. To give some perspective, on that same database Designer Display 2100 rates at 1846 after 152 games, Fidelity Elegance rates at 1776 after 151 games, Super Constellation at 1767 after 68 games, the Turbostar 1769 after 144 games (and MMIV maintains it's 1904 rating from the 1996 SSDF rating list). Unlike some other ratings you find on the web which are merely estimates, these are ratings achieved under formal 40 in 2 tournament conditions and close to 1500 games played thus far. That 1779 rating for Grandmaster is therefore both realistic and stable. Compare that to any Nelson program pre-Excalibur days and there is a rating improvement easily over 500 points. And that improvement over the old Fidelity machines he authored has alot to do with Nelson finally being given decent hardware at Excalibur as opposed to his days at Fidelity.

Look at the specs of the old Nelson machines back in the days at Fidelity. Take the Gambit for instance. The program was only 4K in size. It is impossible to produce a decent chess program with only 4K. The Kittenger 4K programs or the Mephisto 4K programs can be defeat at odds with barely any thought at all. They play no better than experienced beginners. The best 4K progams around in my opinion are ones such as that found in the original Scysis Sensor Chess and the Scisys Companion II (now apparently found in modified form in some Lexibook models). But even they are all still very weak. In my opinion a lot of the reputation Nelson has in relation to mediocrity has far more to do with never having been given decent hardware to work with back in the 1980s. His machines were always going to look bad in comparison to the Spracklens who were given 4 times the memory at the bare minimum and 6502 processors at the bare minimum. As Australian racing icon Dick Johnson once said, you can't win Wimbledon with a pingpong bat!

In another old versus new comparison, I have done a lot of testing of the Deluxe Talking Touch Chess (DTTC). This machine is of course by Nelson as well. Now this particular machine has a 4 Mhz 6502 processor (so it is only marginally faster than the CPU used in the old Fidelity Elegance, for example. The DTTC has what I believe to be a 27K program. In my testing thus far, this machine appears to play around the high 1600s mark - perhaps low 1700s. So it is not that far behind some very competant mainstream machines of the 1980s that used similar hardware. It is certainly a little stronger than my Mephisto Europa for example. Sure, it is certainly not as strong as a program by top authors given that same hardware, but it is a decent effort nonetheless.

And turning even to his most basic programs today (i.e those found in Chess Station, King Arthur, which run on lesser hardware than the DTTC). In some off hand testing against the Novag Carnelian II (a 16K Kittenger program using an 8 Mhz 6301Y), Carnelian II was incapable of winning a single game. At best it achieved draws. And I note that this 16K Kittinger program rates at 1582 on the database referred to earlier - but only after a small sample of 19 games thus far.

So it is likely the latest Nelson programs play somewhere in the region of the low 1500s to the high 1700s depending on the program and the hardware. Do I think they are poor positional players? I don't realy think they are any better or worse than any other program out there of comparable playing strength.

Now all of the above is not to say that I am at all happy with Excalibur or these machines in other respects. I have noticed many bugs with these programs, although I have it on excellent authority that the Grandmaster has far less software issues than the other machines in the range. But typical problems I have come across include allowing illegal castling, poor use of time allocation (either far too fast or too slow), inbuilt tournament clocks that go haywire after the first time control is reached, not thinking on their opponent's time and display corruption issues where a rolling matrix display is used. Still, the actual chess capabilites of these latest programs cannot really be compared to the older programs of the 1980s. Unfortunately, though, in the end it is the annoying bugs in the Excalibur machines and the non-existent Excalibur customer service that is keeping me from future Excalibur purchases. A pity, because I think given the right circumstances there is no reason why these machines can't be as free from issues as the most solid offerings from other companies in the 1980s.

In closing, I have attached some games that show the latest Nelson programs in their best light. Firstly, two games of the DTTC against Gameboy Chessmaster Advance (which could be headed towards a high 1800s rating - but it only has 13 recorded games to it's credit thus far and it's current performance rating of 2072 is rediculously overstated due to the small game sample). Secondly, some victories by Excalibur GM over some well known machines. All these games are 40 moves in 2 hours:

[Event "?"]
[Site "?"]
[Date "2007.04.02"]
[Round "?"]
[White "Chessmaster GBA"]
[Black "Deluxe Talking Chess"]
[Result "1/2-1/2"]
[PlyCount "149"]

1. d4 d5 2. c4 dxc4 3. Nf3 Nf6 4. e3 e6 5. Bxc4 c5 6. O-O a6 7. b3 Nc6 8. Bb2
b5 9. Be2 Bb7 10. dxc5 Bxc5 11. a4 bxa4 12. Qxd8+ Rxd8 13. Rc1 Bb6 14. Rxa4 Na5
15. b4 Nb3 16. Rc3 Bxf3 17. Bxf3 Nd2 18. Nxd2 Rxd2 19. Rb3 O-O 20. Rxa6 Rfd8
21. g3 Nd7 22. Bc3 Rc2 23. Kf1 Rc8 24. Raa3 h6 25. Be4 Rc1+ 26. Ke2 Nf6 27. Bb7
Rb8 28. Bd2 Rc2 29. Ra8 Rxa8 30. Bxa8 Ng4 31. Be4 Rc8 32. h3 Nf6 33. Bf3 g6 34.
b5 Rc2 35. Ke1 Kf8 36. Rb1 Ke8 37. Ra1 Kd7 38. Ra6 Kc7 39. Ra8 Nd7 40. Bc6 Ne5
41. Kd1 Rc5 42. Be8 Kb7 43. Ra4 Rc8 44. f4 Rxe8 45. fxe5 Bc7 46. Bc3 Rd8+ 47.
Ke2 Rd5 48. Rb4 Ba5 49. Rb3 Bxc3 50. Rxc3 Rxe5 51. Rd3 Kc7 52. Rb3 Kb6 53. Rb4
Rxb5 54. Rxb5+ Kxb5 55. Kd3 f5 56. Kd4 Kc6 57. Ke5 Kd7 58. Kf6 g5 59. Kg6 Ke7
60. Kxh6 Kf6 61. Kh5 f4 62. exf4 gxf4 63. gxf4 Kf5 64. h4 Kxf4 65. Kg6 e5 66.
h5 e4 67. h6 e3 68. h7 e2 69. h8=Q e1=Q 70. Qf6+ Kg3 71. Qg5+ Kf3 72. Qf5+ Kg3
73. Kf6 Qf2 74. Ke6 Qxf5+ 75. Kxf5 1/2-1/2



[Event "?"]
[Site "?"]
[Date "2007.04.05"]
[Round "?"]
[White "Chessmaster GBA"]
[Black "Deluxe Talking Chess"]
[Result "1/2-1/2"]
[ECO "A32"]
[PlyCount "63"]
[TimeControl "900"]

1. d4 Nf6 2. c4 c5 3. Nf3 cxd4 4. Nxd4 e6 5. Nc3 Bb4 6.
Ndb5 d5 7. cxd5 exd5 8. a3 Bc5 9. Nxd5 Nxd5 10. Qxd5 Qa5+ 11. b4 Qxb5 12. bxc5
Nd7 13. Qd4 f6 14. e4 Qxc5 15. Be3 Qxd4 16. Bxd4 b6 17. Bd3 Ne5 18. Bb5+ Bd7
19. Bxd7+ Kxd7 20. O-O Rhc8 21. Rfd1 Ke7 22. f4 Nc4 23. Rd3 Rab8 24. Rad1 Rb7
25. Rh3 h6 26. Rhd3 Rbb8 27. Kf2 Rb7 28. Kg3 Rbb8 29. Bc3 Rb7 30. Bd4 Rbb8 31.
Bc3 Rb7 32. Bd4 1/2-1/2


[Event "All 40/2"]
[Site ""]
[Date "2005.??.??"]
[Round "9"]
[White "Grandmaster"]
[Black "Saitek Simultano"]
[Result "1-0"]
[ECO "D20"]
[PlyCount "61"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. d4 d5 2. c4 dxc4 3. e3 e5 4. Bxc4 exd4 5. exd4 Nf6 6. Nf3 Bb4+ 7. Nc3 O-O 8.
O-O Bg4 9. a3 Bxc3 10. bxc3 c5 11. dxc5 Qxd1 12. Rxd1 Ne4 13. Bf4 Nxc3 14. Rdc1
Bxf3 15. gxf3 Na4 16. Bb5 Nb2 17. Rab1 a6 18. Bf1 Na4 19. Rb4 Nxc5 20. Rxc5 b5
21. Rc7 Rd8 22. Re4 h6 23. Ree7 g5 24. Be5 b4 25. Rxf7 bxa3 26. Bc4 Rd1+ 27.
Kg2 Rg1+ 28. Kxg1 Nd7 29. Rf6+ Kh7 30. Rxd7+ Kh8 31. Rxh6# 1-0


[Event "All 40/2"]
[Site ""]
[Date "2004.??.??"]
[Round "6"]
[White "Grandmaster"]
[Black "Novag Primo"]
[Result "1-0"]
[ECO "B22"]
[PlyCount "49"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. e4 c5 2. c3 d5 3. exd5 Qxd5 4. d4 e6 5. Nf3 Nc6 6. c4 Qe4+ 7. Be3 cxd4 8.
Nxd4 Bb4+ 9. Nc3 Nf6 10. Nxc6 Bxc3+ 11. bxc3 bxc6 12. Qd6 Nd7 13. Rd1 c5 14.
Qg3 g6 15. Be2 Ne5 16. Bxc5 Bb7 17. Qg5 Nc6 18. Qf6 Rg8 19. Rd7 Qb1+ 20. Bd1
Qe4+ 21. Kd2 Kxd7 22. Qxf7+ Ne7 23. Qxe7+ Kc8 24. Bd6 Qxc4 25. Bb3 1-0


[Event "All 40/2"]
[Site ""]
[Date "2004.??.??"]
[Round "10"]
[White "Grandmaster"]
[Black "Fidelity Elegance"]
[Result "1-0"]
[ECO "B00"]
[PlyCount "101"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. e4 Nc6 2. Nf3 d5 3. exd5 Qxd5 4. Nc3 Qa5 5. Bb5 Nf6 6. d4 Ne4 7. Qd3 Nxc3 8.
bxc3 Bf5 9. Qxf5 Qxc3+ 10. Ke2 e6 11. Qe4 O-O-O 12. Bxc6 bxc6 13. Bd2 Qc4+ 14.
Qd3 Qa4 15. c4 c5 16. Qb3 Qxb3 17. axb3 cxd4 18. Ne5 Bc5 19. Nxf7 d3+ 20. Kd1
Bxf2 21. Bc3 Rhg8 22. Nxd8 Rxd8 23. Rf1 Bd4 24. Kd2 Bxc3+ 25. Kxc3 e5 26. Rad1
e4 27. Rfe1 Re8 28. Rxd3 Re5 29. Kd4 Rh5 30. Rh3 c5+ 31. Kxe4 Rg5 32. Rg3 Rxg3
33. hxg3 h6 34. Kd5 h5 35. Re8+ Kd7 36. Rg8 g6 37. Rxg6 a5 38. Kxc5 Kc7 39.
Rg7+ Kd8 40. Kb5 Ke8 41. Kxa5 Kf8 42. c5 Kxg7 43. c6 Kf6 44. c7 Ke7 45. c8=Q
Kf7 46. Qf5+ Ke7 47. b4 Kd6 48. b5 h4 49. gxh4 Ke7 50. h5 Ke8 51. Qh7 1-0

[Event "All 40/2"]
[Site ""]
[Date "2003.??.??"]
[Round "4"]
[White "Mephisto MM IV"]
[Black "Grandmaster"]
[Result "0-1"]
[ECO "B17"]
[PlyCount "82"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. e4 c6 2. d4 d5 3. Nc3 dxe4 4. Nxe4 Nd7 5. Bc4 Nb6 6. Bb3 Nf6 7. Nxf6+ exf6
8. Ne2 Bd6 9. c4 Bb4+ 10. Bd2 Bxd2+ 11. Qxd2 O-O 12. O-O Re8 13. Rfe1 Bg4 14.
f3 Bf5 15. Ng3 Be6 16. Qd3 a5 17. Rac1 a4 18. Bc2 g6 19. b3 axb3 20. Bxb3 Nd7
21. Rcd1 Qa5 22. d5 Qb6+ 23. Qd4 cxd5 24. cxd5 Qxd4+ 25. Rxd4 Bf5 26. Red1 Nc5
27. Nxf5 gxf5 28. Bc2 Re2 29. Bb1 Kg7 30. Kf1 Re5 31. Rc4 Na4 32. Rc7 Nb6 33.
Rxb7 Nxd5 34. Rd3 Rae8 35. Rdb3 Re1+ 36. Kf2 R8e2+ 37. Kg3 Rg1 38. Kh4 Rexg2
39. Rxf7+ Kxf7 40. Rb7+ Kg6 41. Bxf5+ Kxf5 0-1

[Event "All 40/2"]
[Site ""]
[Date "2003.??.??"]
[Round "4"]
[White "Turbostar Kasparov"]
[Black "Grandmaster"]
[Result "0-1"]
[ECO "E42"]
[PlyCount "90"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. e3 c5 5. Nge2 cxd4 6. exd4 d5 7. a3 Be7 8.
cxd5 exd5 9. Nf4 O-O 10. Qb3 Nc6 11. Be3 Na5 12. Qa4 Be6 13. Bd3 Rc8 14. O-O
Bd6 15. Nxe6 fxe6 16. Nb5 Ng4 17. h3 Bh2+ 18. Kh1 Nxe3 19. fxe3 Rxf1+ 20. Rxf1
Bb8 21. e4 dxe4 22. Bxe4 a6 23. Nc3 Qd6 24. Kg1 Nc4 25. Bxb7 Qxd4+ 26. Kh1 Qe5
27. g3 Qxg3 28. Qc2 Qxh3+ 29. Kg1 Ba7+ 30. Rf2 Qg3+ 31. Bg2 Nxb2 32. Qxb2 Rxc3
33. Qd2 Bxf2+ 34. Qxf2 Rc1+ 35. Qf1 Qxa3 36. Bh3 Qe3+ 37. Kh2 Rxf1 38. Bxf1 a5
39. Kg2 a4 40. Kh2 Qf2+ 41. Kh3 Qxf1+ 42. Kg4 Qf2 43. Kg5 Qg3+ 44. Kh5 g6+ 45.
Kh6 Qf4# 0-1

JonP01

Re: Playing Ron Nelson Programs, Today

Post by JonP01 » Mon Apr 09, 2007 1:13 am

And of course, in trying to maintain a sense of balance, here are some games from the Excalibur "bloopers" reel ;) Bear in mind, however that these horrific games all counted towards the ratings as described in the thread above. I have always been critical of people who claim player "x" is stronger than player "y" because they found or quote a series of games where player "x" won. They rather conveniently forget about losses and draws. In the end, an accurate rating achieved over a large sample of games is the only indicator of comparative chess strength. So here is part of the reason why the Nelson programs are not High Expert rated!



[Event "?"]
[Site "?"]
[Date "2007.04.02"]
[Round "?"]
[White "Chessmaster GBA"]
[Black "Deluxe Talking Chess"]
[Result "1-0"]
[PlyCount "43"]

1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nc6 5. Nc3 Nge7 6. Ndb5 d5 7. Bf4 e5 8.
exd5 exf4 9. d6 Nf5 10. Nc7+ Kd7 11. Qd5 Nfd4 12. Qxf7+ Kxd6 13. O-O-O Rb8 14.
N7b5+ Kc5 15. Na4+ Kb4 16. Nxd4 Be6 17. Qxe6 Qb6 18. Qb3+ Ka5 19. Nxb6 Bb4 20.
Nc4+ Ka6 21. Qa4+ Ba5 22. Qb5# 1-0



[Event "All 40/2"]
[Site ""]
[Date "2004.??.??"]
[Round "5"]
[White "Conchess 2MHz"]
[Black "Grandmaster"]
[Result "1-0"]
[ECO "B13"]
[PlyCount "65"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. e4 c6 2. d4 d5 3. exd5 cxd5 4. Bd3 Nc6 5. c3 Nf6 6. Bf4 Bg4 7. Ne2 Qb6 8.
Qc2 Rc8 9. O-O e5 10. dxe5 Nb4 11. Qd2 Nxd3 12. Qxd3 Qxb2 13. Nd2 Bxe2 14. Qxe2
Ne4 15. Rfb1 Qxc3 16. Nxe4 Qc4 17. Qg4 Rb8 18. Nd2 Qd4 19. e6 Rd8 20. exf7+
Kxf7 21. Rxb7+ Be7 22. Rxe7+ Kf8 23. Nb3 Qf6 24. Bg5 h5 25. Qg3 Qg6 26. Rxa7
Re8 27. Be7+ Rxe7 28. Qxg6 Rxa7 29. Rc1 Rg8 30. Qe6 g6 31. Rc8+ Kg7 32. Rxg8+
Kh7 33. Qxg6# 1-0



[Event "All 40/2"]
[Site ""]
[Date "2003.??.??"]
[Round "4"]
[White "Mephisto Miami"]
[Black "Grandmaster"]
[Result "1-0"]
[ECO "E60"]
[PlyCount "51"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. d4 Nf6 2. c4 g6 3. Nf3 e6 4. Nc3 Bb4 5. Qb3 Na6 6. Bg5 h6 7. Bxf6 Qxf6 8.
O-O-O O-O 9. e3 Bxc3 10. Qxc3 b6 11. Be2 Bb7 12. h4 Qf5 13. Rh3 c5 14. h5 cxd4
15. Rxd4 e5 16. Rd6 g5 17. Rxh6 g4 18. Rg3 Kg7 19. Nxe5 Nc5 20. Bxg4 Qxf2 21.
Bf5+ Kxh6 22. Ng4+ Kxh5 23. Nf6+ Kh6 24. Rh3+ Qh4 25. Rxh4+ Kg5 26. Rh5# 1-0



[Event "All 40/2"]
[Site "Blue Mountains"]
[Date "2007.??.??"]
[Round "16"]
[White "Novag Amber"]
[Black "Grandmaster"]
[Result "1-0"]
[ECO "D20"]
[PlyCount "51"]
[EventDate "2005.??.??"]
[EventType "tourn"]
[EventCountry "AUS"]

1. d4 d5 2. c4 dxc4 3. e4 e5 4. Nf3 exd4 5. Bxc4 Bb4+ 6. Bd2 Bxd2+ 7. Nbxd2 Nc6
8. O-O Nf6 9. e5 Ng4 10. h3 Ngxe5 11. Nxe5 Nxe5 12. Re1 f6 13. f4 Kf8 14. fxe5
d3 15. Qh5 Qe8 16. Qxe8+ Kxe8 17. exf6+ Kf8 18. Re7 g5 19. Rae1 Bxh3 20. Rf7+
Kg8 21. Ree7 Be6 22. Rg7+ Kf8 23. Bxe6 h5 24. Rgf7+ Kg8 25. Rh7+ Kf8 26. Rxh8#
1-0

Steve B

Re: Playing Ron Nelson Programs, Today

Post by Steve B » Mon Apr 09, 2007 2:02 pm

hi Jon

well there is no question that the program in the GM (and Mirage) are not the same as the SC8.neither are the programs in Alexandra/DTTC etc et the same..Nelson has been improving his engines consistently throughout the passage of time

something puzzles me about all of the GM's results though
having owned two Gm's since first released and after having played literally 100's of games against them(not other computers ..but me myself playing) one quickly notices that the GM has a very LIMITED book response when left on its own to decide what opening to play
the only way to get into its bigger book if for the player to INTERCEDE and force it to play another opening move

so i wonder..do all of these match results come from the GM deciding for itself the opening(which i doubt) or did Cameron Intercede and force it to play a different line and if so does he makes choices in the line itself or only the first few moves?

i notice a wide disparity in the rating for the GM
Cameron and SS have it rated in the mid-high 1700's while a few other sites(.INFO and Kurt) show it at mid 1800's

i wonder if the disparity is a result of just letting the GM choose its own opening(in which case there have to be multiple examples of the same exact games being played) or having operator intervention to steer the GM into a more varied opening response(which i consider a necessary evil for a correct rating)

Perplexed Regards
Steve

JonP01

Re: Playing Ron Nelson Programs, Today

Post by JonP01 » Mon Apr 09, 2007 2:52 pm

Hi Steve,

Perhaps Cameron will clarify himself, but so far as I know he has always adopted a very strict policy of letting a dedicated unit be 100% responsible for it's own moves. I would thus be extremely surprised if he did anything as regards forcing particular openings. I have not really looked at all of the games in detail yet, however my feeling is that since most of his other machines have wide libraries, the chance of repeat games is minimal, even with such a large sample. For instance, he has a Fidelity Elegance and I have an Excellence. I have a library "problem" with the Excellence, but his Elegance has a lot of variety in the opening. I would have to be careful about repeat games with the Excellence, especially if it were playing the GM, but the chance of repeat games in Elegance versus GM would - I think - be very low. For my own part, I would look up my database for the position at the end of the opening to see if the two contenders had reached that particular position before. In which case I would re-start the game. I imagine Cameron has some method for picking up on potential repeat games.

I know other web sites have GM rated a bit higher, although as you point out, the differences between the SS, Cameron's and my own estimate (before I got the game database) only differed by literally a few points.

I should point out that my estimate was actually based on game results at schachcomputer.info, however I did some filtering of the results to get my estimate of 1785. I noticed that the GM had been pitted against opponents that in the "real world" it would never meet for rating purposes (as you know, rated tournaments are in classes - one of the many reasons being being to preserve the statistical relevance of the ELO rating system). By that I mean playing rated games against opponents far stronger or far weaker. The issue I felt was compounded by relatively small game samples against the opposition far outside the GMs nominal rating band. So you only need a few lucky games against much higher opposition (or unlucky ones against much lower opposition) and it can potentially skew the rating. When I adjusted for this (i.e kept only the games with statistically significant samples and within say 150 points plus or minus), I got a 1785 estimate rating for it.

Additionally, if you carefully peruse the rating list at schachcomputer.info, you would note that a significant number of machines have ratings which have been calculated based upon very small game samples. Whilst this is clearly acknowledged in the rating tables, it is possible these early ratings may still be used to influence the ratings of the other machines. When I arrived at the 1785 estimate for GM, I took my filtered results and then applied the ratings from the 1996 SSDF list to the machines which it had played. I felt this was much more accurate than using the ratings published there. In any event, I must have done OK, since my estimate was only 6 points higher than Cameron's GM rating after 166 independant games.

Whilst I might seem biased, I have total faith in Cameron's rating and therefore it follows I have faith in the SS rating too. One can never be 100% sure how rated computer games are played by other people. I'm not suggesting cheating by any means, but operators could possibly do things such as resign on behalf of machines, adjudicate draws, speed the game up towards the end, press the "move" button on (seemingly) obvious moves due to impatience, force particular openings, etc, etc. Doing any of these things can of course influence the attempt to obtain an accurate rating.

Incidentally, you will like the quaint bug (yet another) I discovered in the DTTC today. You know how the previous version of the program permitted illegal castling? Well, the DTTC version does not allow legal castling!!! :shock: Ron fixed the first castling probelm, but created a new bug in the process. Luckily, the bug only occurs on takeback. If you have done any of the following (i) already castled (ii) moved the king (iii) moved the castling side rook and you then take back (undo) the game to before any of those things happened...then you can no longer castle again later in the game - even though it is legal!

Still, on the bright side I have found a workaround for the corrupt / nonsense crazy clock bug. I have now determined this only occurs if one side exceeds the time limit. That said, DTTC exceeds the time limit in about 75% of it's games (another bug) - which is why I have noted the bug as often as I have. Anyway, the workaround is to reset the time control once the second time control kicks in. However, the way you reset the time control is the key. You have to enter a completely different time control (for example, a move every second) and then set the machine to that. You then need to go into the level function a second time to re-enter the correct tournament time control. If you directly try to re-enter the tournament time control you actually want the first time around, you will still suffer the crazy clock syndrome.

Beta testing for Nelson Regards

Jonathan

Steve B

Re: Playing Ron Nelson Programs, Today

Post by Steve B » Mon Apr 09, 2007 3:10 pm

hi Jon

IIRC..you will see that the GM will only reply with 2 different moves against D4 and i think only 2 against E4

as White it will only open with 2-3 moves

I could be alittle off with the numbers
i cant verify this for myself now because i dont have either of my GM'S with me to check this

but when you receive yours you will immediately notice the limited book response when left to its own design

I just cant see how literally 100's of games were played without repeat games coming up ..when i say repeat games i also mean a game played against a Miami say..but an exact game against an Amber UNLESS there was operator intervention at the very least..the first move for the GM

so for example..the GM will NEVER open with C4 on its own..However if you FORCE C4 and then if the other computer goes into an English(C5) then the GM will be able to play this line fairly deeply...or it can also transpose into a QGD

when your GM arrives will see what i am referring to

Best
Steve

User avatar
fern
Posts: 8755
Joined: Sun Feb 26, 2006 3:07 pm

Re: Playing Ron Nelson Programs, Today

Post by fern » Mon Apr 09, 2007 3:21 pm

Great post, pal. Certainly you have made extensive research about this. But nevertheless I sill think my point is valid in general. As I said, Nelson did improve his stuff over the years and certainly he does have now better hardware, but, still, he does not produce a really strong program, as he never did in his early years.
This is not to condemn or otherwise Nelson. He probably had -and have- in mind the average player and does not give a shit for winning some extra elo points as PC programmers do.
In any case my experience playing Nelson program was and is the same: you feel interest because of his shallow book -as Blincoe pointed-, his preposterous moves. It does have, of course, tactical acumen enough to bother you precisely when you are relaxing.
My fault: I must develop an attitude to kill even innocent rabbits.

My bst
Fernando

JonP01

Re: Playing Ron Nelson Programs, Today

Post by JonP01 » Mon Apr 09, 2007 3:52 pm

Hi again Steve,


I have just now gone through every single Grandmaster game in the database. There are only two instances of an opening being played twice against the same opponent (with GM having the same colour in both instances). This occured in two games against Designer 2100 and two games against Mephisto Miami. In the case of the Designer, the two games with the same opening code (C55) had different results. One game lasted 68 moves and the other 50 moves (both ended with Designer getting the full point, but I clarify that all games are played to mate or resignation - should the losing computer have a resignation function). In the two games with Miami, the repeated opening was C54 and the results were quite different. One game ended in a draw after 39 moves and another a victory to Miami after 76 moves.

Additonally, I note that there are infact 102 unique ECO opening codes used in all the GM games. I have listed them below. I am quite confident that there are no repeat games in the database although I have not personally cross referenced every single final position. But since almost every opening is either different or played as alternating colours, and the number of moves in all the games are also unique per opponent, I don't really feel there is any need to that. Bar the above two cases cited above, all the openings are different, or where they are the same then GM played them as white then as black. Only Cameron can truly clarify as to whether the openings were "forced" or not. As I stated earlier though, it takes two to tango, and the opening in a chess game is obviously not determined by one player only.

Here is a list of the 102 unique codes that I extracted from the games. I realise that a unique ECO code does not necessarily mean a unique opening, however after sorting the games by opponent, the number of moves in each game is different. The chances therefore of repeat games when the total moves are different each time and the ECO codes are different are exceptionally remote indeed.


A00
A02
A06
A11
A12
A19
A22
A28
A32
A36
A52
A57
A59
A60
A65
A77
B00
B01
B03
B06
B07
B10
B13
B14
B15
B17
B18
B19
B22
B23
B29
B41
B42
B43
B45
B51
B52
B67
B70
B76
B81
B83
B86
B89
B90
B98
B99
C05
C06
C08
C09
C10
C28
C35
C41
C42
C43
C46
C47
C48
C49
C50
C54
C55
C64
C67
C68
C69
C77
C81
C82
C88
C95
C96
D01
D02
D07
D10
D11
D18
D20
D25
D29
D30
D42
D45
D58
D90
E00
E01
E11
E12
E15
E21
E32
E35
E39
E42
E44
E48
E50
E60

Steve B

Re: Playing Ron Nelson Programs, Today

Post by Steve B » Mon Apr 09, 2007 4:06 pm

hi Jon

then what has to be happening is that when Cameron sees a repeat game(checking it against the other games already played in the database) ..then he must be aborting the game and restarting until the other computer varies its opening

there really is no way that the GM is going to play 100 different opening's in 100 different games unless there is some operator intervention (either the GM or the other computer or aborting games )at some point

but it is important to note that i consider this intervention as NECCESSARY to get a good and accurate feel for the Gm's level of play otherwise you are looking at the same 10-15 games over and over

i am not criticizing intervention in this case..merely pointing it out and saying it is necessary

Steve

Post Reply