Houdini 1.5a - Strelka 5.5 Blitz 1+2 X 360 Games

[sainzlei]

Both engine use 1 core , ponder on , Deep Fritz 8 GUI , Time control : 1'+2"

[Dr.Wael Deeb]

Hello beauty

Bye-bye Robert

[rodolfoleoni]

Hello beauty

Bye-bye Robert

What would it be without the 29.000 hard-coded positions into Strelka?

[geots]

Hello beauty

Bye-bye Robert

What would it be without the 29.000 hard-coded positions into Strelka?

With all due respect, I have no idea- but I would not think the coded positions would be a problem- unless it is illegal for him to put whatever he put in the engine. I haven't heard anyone mentioning the word "illegal" in reference to the positions.

The one thing that is a first for me- and understand I like Richard and Critter- is one engine author going into another's engine and removing this or that and then saying now you can use the new one. I am not quite sure he had that right.


Anyway- the best to you,

george

[mar]

On the first thought, having hardcoded positions sounds like cheating. On the other hand I agree with you George - it's part of the engine. You can say it's a sort of hardcoded persistent hash. Some engines have that too for analysis purposes. Some do book learning to avoid certain lines. Some have finetuned opening books (Just ask Sedat - he's an expert on testing opening books and as we know a good book can add quite some elo). I see nothing wrong with that. We do the same in endgames with EGTBs and/or precalculated tables for some trivial endings. So whether a table or program logic it doesn't realy matter, just a different form of knowledge. So why not do the same in the opening? I think Jury simply came up with a clever way to make Strelka stronger. In fact I find that idea quite amusing and interesting

[rodolfoleoni]

Hello beauty

Bye-bye Robert

What would it be without the 29.000 hard-coded positions into Strelka?

With all due respect, I have no idea- but I would not think the coded positions would be a problem- unless it is illegal for him to put whatever he put in the engine. I haven't heard anyone mentioning the word "illegal" in reference to the positions.

The one thing that is a first for me- and understand I like Richard and Critter- is one engine author going into another's engine and removing this or that and then saying now you can use the new one. I am not quite sure he had that right.


Anyway- the best to you,

george

Hi George,

first, I want to thank you for your very interesting tests.

With all the confusing things which are happening in computer chess it's not easy to define "legal" or "illegal". The discussion would go too far.

I was just wandering how much those positions could have affected the match. If they could have determined the difference, a new fashion could start in chess programming. I consider it not so much different from position learning, except for this one can be turned off (as it is in any rating list).

We can say, we test engines for their capabilities in search and eval, or search + eval + embedded positions. Legal, of course, but, maybe, unfair.

All the best,

[Uri Blass]

On the first thought, having hardcoded positions sounds like cheating. On the other hand I agree with you George - it's part of the engine. You can say it's a sort of hardcoded persistent hash. Some engines have that too for analysis purposes. Some do book learning to avoid certain lines. Some have finetuned opening books (Just ask Sedat - he's an expert on testing opening books and as we know a good book can add quite some elo). I see nothing wrong with that. We do the same in endgames with EGTBs and/or precalculated tables for some trivial endings. So whether a table or program logic it doesn't realy matter, just a different form of knowledge. So why not do the same in the opening? I think Jury simply came up with a clever way to make Strelka stronger. In fact I find that idea quite amusing and interesting

Other programs also can use opening book so it is not the case and I find nothing clever in secret positions that strelka has knowledge about them.

If testers want to test engines with their opening book then there is no problem with some hidden opening book but if this is not the target of testing then it is cheating and it is possible to avoid the problem by testing from positions after some random moves when there is no practical chance to have significant part of the opening positions by some hidden opening book.

[mar]

Other programs also can use opening book so it is not the case and I find nothing clever in secret positions that strelka has knowledge about them.

If testers want to test engines with their opening book then there is no problem with some hidden opening book but if this is not the target of testing then it is cheating and it is possible to avoid the problem by testing from positions after some random moves when there is no practical chance to have significant part of the opening positions by some hidden opening book.

My point is that adding knowledge to a chess program should not be considered cheating.
If you know that move x wins in position y, you certainly want to play it.
In the case of Strelka, the question is whether that knowledge makes it really play better or whether it scores better in test suites.

[gerold]

Other programs also can use opening book so it is not the case and I find nothing clever in secret positions that strelka has knowledge about them.

If testers want to test engines with their opening book then there is no problem with some hidden opening book but if this is not the target of testing then it is cheating and it is possible to avoid the problem by testing from positions after some random moves when there is no practical chance to have significant part of the opening positions by some hidden opening book.

My point is that adding knowledge to a chess program should not be considered cheating.
If you know that move x wins in position y, you certainly want to play it.
In the case of Strelka, the question is whether that knowledge makes it really play better or whether it scores better in test suites.

+1

[Dr.Wael Deeb]

Other programs also can use opening book so it is not the case and I find nothing clever in secret positions that strelka has knowledge about them.

If testers want to test engines with their opening book then there is no problem with some hidden opening book but if this is not the target of testing then it is cheating and it is possible to avoid the problem by testing from positions after some random moves when there is no practical chance to have significant part of the opening positions by some hidden opening book.

My point is that adding knowledge to a chess program should not be considered cheating.
If you know that move x wins in position y, you certainly want to play it.
In the case of Strelka, the question is whether that knowledge makes it really play better or whether it scores better in test suites.

Absolutely valid point of view