Strelka 5.5 x64 BEATS Houdini 2.0 x64

[Strelkaman]

Атлон х2 - 4000+, 1+0, 128 хэш, без книги и баз, Ф12:

Code:

1 2 3
1 Strelka 5.5 x64 ******** 0Ѕ10Ѕ1ЅЅ 11111Ѕ11 11.5/16
2 Houdini 2.0 x64 1Ѕ01Ѕ0ЅЅ ******** 11101011 10.0/16
3 Deep Rybka 4.1 x64 00000Ѕ00 00010100 ******** 2.5/16

[mar]

Атлон х2 - 4000+, 1+0, 128 хэш, без книги и баз, Ф12:

Code:

1 2 3
1 Strelka 5.5 x64 ******** 0Ѕ10Ѕ1ЅЅ 11111Ѕ11 11.5/16
2 Houdini 2.0 x64 1Ѕ01Ѕ0ЅЅ ******** 11101011 10.0/16
3 Deep Rybka 4.1 x64 00000Ѕ00 00010100 ******** 2.5/16

Hi Dmitri,

16 games is not enough to draw any conclusions. Try to play a few thousand. Also, you didn't post the conditions you used for the tournament.

Martin

[Strelkaman]

Атлон х2 - 4000+, 1+0, 128 хэш, без книги и баз, Ф12:

Code:

1 2 3
1 Strelka 5.5 x64 ******** 0Ѕ10Ѕ1ЅЅ 11111Ѕ11 11.5/16
2 Houdini 2.0 x64 1Ѕ01Ѕ0ЅЅ ******** 11101011 10.0/16
3 Deep Rybka 4.1 x64 00000Ѕ00 00010100 ******** 2.5/16

Hi Dmitri,

16 games is not enough to draw any conclusions. Try to play a few thousand. Also, you didn't post the conditions you used for the tournament.

Martin

Hi,Martin the conditions are on the top in Russian: Athlon x2-4000,time 1+0,128 hash,no books,no TB.I agree that it's too early for conclusions.Thanks

[JuLieN]

[moderation]
Thread moved to the correct sub forum.
[/moderation]

[gleperlier]

Hi,Martin the conditions are on the top in Russian: Athlon x2-4000,time 1+0,128 hash,no books,no TB.I agree that it's too early for conclusions.Thanks

Hi,

You can stop your test. Making 1+0 games are almost useless, then making 1000 games, no books no TB, what do you want ? 2 or 3 games played hundred of times ?


Gab

[Strelkaman]

Hi,Martin the conditions are on the top in Russian: Athlon x2-4000,time 1+0,128 hash,no books,no TB.I agree that it's too early for conclusions.Thanks

Hi,

You can stop your test. Making 1+0 games are almost useless, then making 1000 games, no books no TB, what do you want ? 2 or 3 games played hundred of times ?


Gab

I don't trust opening books in tests.I think you can test both engines with opening books in.You must see the different in each game and not push the button,running the tests and then watch the results.We are missing important messages.A week ago i was testing Strelka 55 w/32 vs Houdini 1.5a w/32 and Strelka did a sacrifice that prooved wrong but he did it.Iam interested in Engine behaviour most of all.There is no loop without opening books at all.In 1-0 time some engines change their strategy in the middlegame.I agree about the 1000 rule.Its too early

[Dr.Wael Deeb]

Hi,Martin the conditions are on the top in Russian: Athlon x2-4000,time 1+0,128 hash,no books,no TB.I agree that it's too early for conclusions.Thanks

Hi,

You can stop your test. Making 1+0 games are almost useless, then making 1000 games, no books no TB, what do you want ? 2 or 3 games played hundred of times ?


Gab

I don't trust opening books in tests.I think you can test both engines with opening books in.You must see the different in each game and not push the button,running the tests and then watch the results.We are missing important messages.A week ago i was testing Strelka 55 w/32 vs Houdini 1.5a w/32 and Strelka did a sacrifice that prooved wrong but he did it.Iam interested in Engine behaviour most of all.There is no loop without opening books at all.In 1-0 time some engines change their strategy in the middlegame.I agree about the 1000 rule.Its too early

But again you'll have a lot of repeated games without using an opening book....
Dr.D

[Strelkaman]

Hi,Martin the conditions are on the top in Russian: Athlon x2-4000,time 1+0,128 hash,no books,no TB.I agree that it's too early for conclusions.Thanks

Hi,

You can stop your test. Making 1+0 games are almost useless, then making 1000 games, no books no TB, what do you want ? 2 or 3 games played hundred of times ?


Gab

I don't trust opening books in tests.I think you can test both engines with opening books in.You must see the different in each game and not push the button,running the tests and then watch the results.We are missing important messages.A week ago i was testing Strelka 55 w/32 vs Houdini 1.5a w/32 and Strelka did a sacrifice that prooved wrong but he did it.Iam interested in Engine behaviour most of all.There is no loop without opening books at all.In 1-0 time some engines change their strategy in the middlegame.I agree about the 1000 rule.Its too early

But again you'll have a lot of repeated games without using an opening book....
Dr.D

If you test engines without opening books and TB the result is more logical.Start an engine match to see it for yourself.There is no loop in openings because an engine reacts differently in judging the moves.There is no learning option afterall

[M ANSARI]

What they are trying to tell you is that without books right out of the opening and with your time controls chances are that plies will be very low and many of the games will be repeated and thus give you a poor idea of how an engine is performing. Much better is to give it starting positions that are played by both engines switching sides. 1 0 is pretty useless to test unless you do thousands of games and you are testing 2 engines with identical time management setups. Actually I think 1 0 is pretty useless testing anything especially when testing different engines ... and if you want a more accurate representation of results test at 1 + 1s, this way you avoid time losses and minimize losses by engines that have their time management setup for more realistic games.

[Strelkaman]

What they are trying to tell you is that without books right out of the opening and with your time controls chances are that plies will be very low and many of the games will be repeated and thus give you a poor idea of how an engine is performing. Much better is to give it starting positions that are played by both engines switching sides. 1 0 is pretty useless to test unless you do thousands of games and you are testing 2 engines with identical time management setups. Actually I think 1 0 is pretty useless testing anything especially when testing different engines ... and if you want a more accurate representation of results test at 1 + 1s, this way you avoid time losses and minimize losses by engines that have their time management setup for more realistic games.

Except Strelka test engines because Strelka uses the best move method.If you test Strelka vs Strelka it doesn't loop.I agree with your 1.1 time management