Movei00_8_403 and Movei00_8_418 in Chess960

[pichy]

Great,but you must reverse the opening books and play even number of games and then calculate the general result

I simply let the two engines select whatever opening they want. Do you expect me to make the calculations for both engines too

Engine Score Mo Gl S-B
1: Movei00_8_421 5.0/9 ········ 101=010=1 16.00
1: Glaurung 4.0/9 010=101=0 ········ 16.00

9 of 30 games played
Name of the tournament: Arena tournament
Site/ Country: Jorge, United States
Level: Blitz 5/1
Hardware: AMD Athlon(tm) Processor 1202 MHz with 512MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: C:\Program Files\Arena\Arena.pgn
Website:
E-Mail Address:

[Uri Blass]

[Dr. Wael I really do NOT understand why every time I post matches using HS Opening Book you have to criticize the Opening lines, but when you were interviewed you recommended to use HS Mainbook.abk under Arena]

As I wrote before,the Arena Main Book is the gold standatd and I do think so indeed,but....
When I see something like 1.e4 c6 2.d4 d5 3.Nd2 g6?! followed by 5....Nh6 I can't hold myself making a statement....
Take a look at whatever a professional opening book or analyse these opening lines and you'll see what I mean....

Note that Uri didn't believe your results against Glaurung because Glaurung was handicapped in the opening phase and you recieved an odd results....
I don't beleive no more in the generic opening books and that's why I am working over special tuned opening books for certin engines

P.S.Soon a super strong chess engine will make it's first official release,stay tuned and don't fall asleep please

This is not what I said.
I believe that movei was lucky against glaurung and I suspected that maybe movei prefer his hardware relative to glaurung.

I did not talk about book and in my tests I do not use opening book
but match from predefined positions.

Note that I also talked about 403 or 418 when I did not believe that it can be at similiar level to glaurung.

I also believe that latest version is weaker than glaurung but it was not tested by me in games except games against previous version.

I can add that I will be happy if I find that it can beat fruit2.0 in my tests and at this point of time I do not expect a very big improvement.

I will be happy with 50 elo relative to version 403.

Uri

[pichy]

[Dr. Wael I really do NOT understand why every time I post matches using HS Opening Book you have to criticize the Opening lines, but when you were interviewed you recommended to use HS Mainbook.abk under Arena]

As I wrote before,the Arena Main Book is the gold standatd and I do think so indeed,but....
When I see something like 1.e4 c6 2.d4 d5 3.Nd2 g6?! followed by 5....Nh6 I can't hold myself making a statement....
Take a look at whatever a professional opening book or analyse these opening lines and you'll see what I mean....

Note that Uri didn't believe your results against Glaurung because Glaurung was handicapped in the opening phase and you recieved an odd results....
I don't beleive no more in the generic opening books and that's why I am working over special tuned opening books for certin engines

P.S.Soon a super strong chess engine will make it's first official release,stay tuned and don't fall asleep please

This is not what I said.
I believe that movei was lucky against glaurung and I suspected that maybe movei prefer his hardware relative to glaurung.

I did not talk about book and in my tests I do not use opening book
but match from predefined positions.

Note that I also talked about 403 or 418 when I did not believe that it can be at similiar level to glaurung.

I also believe that latest version is weaker than glaurung but it was not tested by me in games except games against previous version.

I can add that I will be happy if I find that it can beat fruit2.0 in my tests and at this point of time I do not expect a very big improvement.

I will be happy with 50 elo relative to version 403.

Uri

I guess that now that I am using his own Rybka V2.0_0_0.abk opening Book Dr. Wael will stop looking for bad opening as an excuse or simply continue to say that it has been a lucky result for Movei421.
Engine Score Mo Gl S-B
1: Movei00_8_421 8.0/13 ············· 101=010==111= 40.00
2: Glaurung 5.0/13 010=101==000= ············· 40.00

13 of 30 games played
Name of the tournament: Arena tournament
Site/ Country: Jorge, United States
Level: Blitz 5/1
Hardware: AMD Athlon(tm) Processor 1202 MHz with 512MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: C:\Program Files\Arena\Arena.pgn
Website:
E-Mail Address:

[Dr.Wael Deeb]

Hmmm,strang....again Movei is leading even though Uri said that the gap between the 2 engines is big in favor of Glaurung

[Uri Blass]

Hmmm,strang....again Movei is leading even though Uri said that the gap between the 2 engines is big in favor of Glaurung

Note that I did not say nothing about version 421

I said that the gap is big for glaurung with version 403
I did not test version 421 against glaurung and I also did not test the specific personality that I recommended for Jorge except games against previous version(I am not sure about the version that he used)

Note that I tested 421 default against
421 with the following parameters:

pawn_square_table 1000 1010 1010 1010 1010 1010 1010 1000 130 130 130 130 130 130 130 130 115 115 115 115 115 115 115 115 100 100 100 105 105 100 100 100 90 92 95 100 100 95 92 90 90 91 92 85 85 92 91 90 91 90 90 70 70 90 90 91 0 0 0 0 0 0 0 0
candidatefactor 3

I used fixed number of nodes of 400,000 nodes per second and the result was 58-42 for the new version(not enough games to be sure that this personality is better but a good result)

I am not sure if Jorge used this personality or this personality without candidatefactor 3 because the personality that I sent him was without candidatefactor 3 and only later I mentioned in an email that candidatefactor 3 may be better.

The change seems to be logical because I added evaluation of candidates pawns already in 418(the evaluation of candidate pawns is added to the normal evaluation of passed pawns in case that they are also passed pawns)

It means that the average value of pawns got higher so it is logical to reduce the value of pawns as compensation and it is logical that reducing the evaluation of pawns allows to make another increase in the evaluation of candidate pawns.

Of course this is only some intuition to the changes that I made and only games can prove if the changes are really good.

Uri

[Uri Blass]

Things may be dependent on the hardware that is used
and maybe glaurung simply have problems against movei inspite of the fact that movei is weaker when glaurung can perform better against other opponents.

I was surprised to find the following in the CEGT blitz rating list

http://www.husvankempen.de/nunn/40_4_Ra ... liste.html

239 Movei 00.8.403 2583 51 51 136 45.2 % 2616 25.7 %
240 Glaurung 1.2.1 2582 28 28 461 38.8 % 2661 23.0 %
256 Movei 00.08.403 2566 20 20 830 43.4 % 2612 28.3 %

For comparison we have

153 Glaurung 1.1 2659 16 16 1245 48.2 % 2671 28.0 %
172 Glaurung 1.0.2 2645 14 14 1761 55.9 % 2604 24.6 %


Note that in this list Glaurung1.2.1 lost a match against default movei
+ 13 = 9 - 18 so maybe Glaurung simply has problems against movei in normal chess.

Note that I did not test movei against glaurung except chess960 and my conclusion that glaurung1.2.1 is significantly stronger than movei was also based on chess960 games of movei against glaurung.


http://www.husvankempen.de/nunn/40_4_Ra ... n/240.html

In the 40/20 rating list we have a different picture
http://www.husvankempen.de/nunn/40_40%2 ... liste.html

111 Glaurung 2-Epsilon/2 2689 60 60 85 40.6 % 2755 36.5 %
120 Glaurung 1.2.1 1CPU 2671 19 19 917 50.8 % 2665 31.4 %
131 Glaurung 1.0.2 2658 15 15 1424 48.8 % 2666 27.9 %
157 Fruit 2.0 2611 16 16 1311 47.3 % 2630 30.8 %
196 Movei 0.08.403 2557 22 22 671 50.3 % 2555 32.6 %

403 never played against glaurung but you can see a gap of more than 100 elo for Glaurung relative to movei

CCRL 40/4 shows the following

Glaurung 1.2.1 32-bit 2757 +9 −9 40.8% +65.8 29.6% 4661
59.1%
15
30 Movei 0.08.418 2657 +26 −27 48.0% +12.2 31.5% 480
50.1%
Movei 0.08.403 2641 +22 −22 54.6% −35.3 28.8% 732

Again more than 100 elo for glaurung.

You can see a similiar picture in the 40/40 rating list

Uri

[pichy]

Hmmm,strang....again Movei is leading even though Uri said that the gap between the 2 engines is big in favor of Glaurung

Note that I did not say nothing about version 421

I said that the gap is big for glaurung with version 403
I did not test version 421 against glaurung and I also did not test the specific personality that I recommended for Jorge except games against previous version(I am not sure about the version that he used)

Note that I tested 421 default against
421 with the following parameters:

pawn_square_table 1000 1010 1010 1010 1010 1010 1010 1000 130 130 130 130 130 130 130 130 115 115 115 115 115 115 115 115 100 100 100 105 105 100 100 100 90 92 95 100 100 95 92 90 90 91 92 85 85 92 91 90 91 90 90 70 70 90 90 91 0 0 0 0 0 0 0 0
candidatefactor 3

I used fixed number of nodes of 400,000 nodes per second and the result was 58-42 for the new version(not enough games to be sure that this personality is better but a good result)

I am not sure if Jorge used this personality or this personality without candidatefactor 3 because the personality that I sent him was without candidatefactor 3 and only later I mentioned in an email that candidatefactor 3 may be better.

Uri, I used the personality without candidatefactor 3 for this result against Glaurung 1.2.1. Next match I will use candidatefactor 3

I used the personality without candidatefactor 3


Engine Score Mo Gl S-B
1: Movei00_8_421 22.0/40 101=010==111=010=1101011=10==010=11010=0 396.00
2: Glaurung 18.0/40 010=101==000=101=0010100=01==101=00101=1 396.00

40 games played / Tournament finished
Name of the tournament: Arena tournament
Site/ Country: Jorge, United States
Level: Blitz 5/1
Hardware: AMD Athlon(tm) Processor 1202 MHz with 512MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: C:\Program Files\Arena\Arena.pgn
Website:
E-Mail Address:



The change seems to be logical because I added evaluation of candidates pawns already in 418(the evaluation of candidate pawns is added to the normal evaluation of passed pawns in case that they are also passed pawns)

It means that the average value of pawns got higher so it is logical to reduce the value of pawns as compensation and it is logical that reducing the evaluation of pawns allows to make another increase in the evaluation of candidate pawns.

Of course this is only some intuition to the changes that I made and only games can prove if the changes are really good.

Uri

[pichy]

This is with candidatefactor 3 so far the score is more impressive


Engine Score Mo Gl S-B
1: Movei00_8_421 9.5/12 ············ 11100111=111 23.75
2: Glaurung 2.5/12 00011000=000 ············ 23.75

12 of 40 games played
Name of the tournament: Arena tournament
Site/ Country: Jorge, United States
Level: Blitz 5/1
Hardware: AMD Athlon(tm) Processor 1202 MHz with 512MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: C:\Program Files\Arena\Arena.pgn
Website:
E-Mail Address:

[Uri Blass]

This is with candidatefactor 3 so far the score is more impressive


Engine Score Mo Gl S-B
1: Movei00_8_421 9.5/12 ············ 11100111=111 23.75
2: Glaurung 2.5/12 00011000=000 ············ 23.75

12 of 40 games played
Name of the tournament: Arena tournament
Site/ Country: Jorge, United States
Level: Blitz 5/1
Hardware: AMD Athlon(tm) Processor 1202 MHz with 512MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: C:\Program Files\Arena\Arena.pgn
Website:
E-Mail Address:

My result with fixed number of nodes candidatefactor 3 and changes in pawn value against the default:

1)58-42 400,000 nodes per move.
2)52.5-47.5 200,000 nodes per move.

I do not believe that it is stronger than glaurung and maybe glaurung does not like your hardware or does not like movei.

It may be interesting to test movei against other programs with similiar strength

possible opponents that are supposed to be slightly weaker than glaurung1.2.1 but stronger than movei based on the ccrl list are

1)Slow Chess Blitz WV2.1
2)Alaric 0704
3)Bright-0.1d
4)Delfi5.1

[pichy]

This is with candidatefactor 3 so far the score is more impressive


Engine Score Mo Gl S-B
1: Movei00_8_421 9.5/12 ············ 11100111=111 23.75
2: Glaurung 2.5/12 00011000=000 ············ 23.75

12 of 40 games played
Name of the tournament: Arena tournament
Site/ Country: Jorge, United States
Level: Blitz 5/1
Hardware: AMD Athlon(tm) Processor 1202 MHz with 512MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: C:\Program Files\Arena\Arena.pgn
Website:
E-Mail Address:

My result with fixed number of nodes candidatefactor 3 and changes in pawn value against the default:

1)58-42 400,000 nodes per move.
2)52.5-47.5 200,000 nodes per move.

I do not believe that it is stronger than glaurung and maybe glaurung does not like your hardware or does not like movei.

It may be interesting to test movei against other programs with similiar strength

possible opponents that are supposed to be slightly weaker than glaurung1.2.1 but stronger than movei based on the ccrl list are

1)Slow Chess Blitz WV2.1
2)Alaric 0704
3)Bright-0.1d
4)Delfi5.1

I am just styarted to test Movei421 versus Slow Chess Blitz WV2.1, but this was the final result versus Glaurung 1.2.1.


This is with candidatefactor 3 a very impressive result for Movei421

Engine Score Mo Gl S-B
1: Movei00_8_421 24.5/40 ········································ 11100111=11101101=101=00100==0=10=110111 379.75
2: Glaurung 15.5/40 00011000=00010010=010=11011==1=01=001000 ········································ 379.75

40 games played / Tournament finished
Name of the tournament: Arena tournament
Site/ Country: Jorge, United States
Level: Blitz 5/1
Hardware: AMD Athlon(tm) Processor 1202 MHz with 512MB Memory
Operating system: Microsoft Windows XP Professional Service Pack 2 (Build 2600)
PGN-File: C:\Program Files\Arena\Arena.pgn
Website:
E-Mail Address: