King Indian like and usefulness of engines

[corres]

[quote="Lyudmil Tsvetkov"]

[quote="corres"]old tests prove that the power of an engine is mainly determined by the speed of that engine.
[/quote]

at least try to follow latest top engine developments, before commenting.
[/quote]

I think that vivid phantasy does not substitute the concrete knowledge.
One who argues somebody about facts from school-books that people argues the authors of the school-books about their knowledge.
So once again: Old tests prove that the power of an engine MAINLY determined by the speed of that engine. Point.

[Evert]

One who argues somebody about facts from school-books that people argues the authors of the school-books about their knowledge.

I have no idea what you're trying to say here.

So once again: Old tests prove that the power of an engine MAINLY determined by the speed of that engine. Point.

That was probably true of programs tested at that time. To say that the main advantage of Stockfish over micro-Max is speed is bollocks.

Of course you need to define what you mean by "speed": time-to-reported-depth (rather ill-defined given the unequal depth along branches of the search tree), speed of movegen/makemove (meaningless), nodes per second (somewhat arbitrary).

[corres]

I have pointed to school-books and their authors.
Definition of speed arises from them.
Please, debate with them.

[Evert]

I have pointed to school-books and their authors.
Definition of speed arises from them.
Please, debate with them.

You have to do better than vague claims to back up your point.
I know a guy who once spoke to a guy who said something, but that's hardly relevant information, is it?

As it is, you don't even say who supposedly said that, where they said it, when and in what context.

[corres]

[quote="Evert"][quote="corres"]I have pointed to school-books and their authors.
Definition of speed arises from them.
Please, debate with them.[/quote]
You have to do better than vague claims to back up your point.
I know a guy who once spoke to a guy who said something, but that's hardly relevant information, is it?

As it is, you don't even say [i]who[/i] supposedly said that, [i]where[/i] they said it, [i]when[/i] and in what context.[/quote]

I have not written any articles or school-books about chess programs and their behavior. I have quoted only a well known fact about the connection among power, speed and logic of evaluation. As I know you made chess programs so I suppose that you know well the books and articles about engines. Maybe I am mistaken. But if you do not want to agree with the sentence cited by me please debate with the original authors.

[Evert]

I have not written any articles or school-books about chess programs and their behavior. I have quoted only a well known fact about the connection among power, speed and logic of evaluation.

But what exactly is the fact that you quote?
Do you mean that a better evaluation does not prevail against an increase in search depth of 1 ply? I think that is a result from the 70s or 80s, and it was probably true then. What it tells us is that a basic alpha-beta search with a simple evaluation that out-searches a basic alpha-beta search with a complicated evaluation does better.

Modern chess programs don't have a basic alpha-beta search though, and they pretty much all have a lean and simple evaluation function.

You can make Stockfish (say) "faster" (in terms of nodes/s) by stripping the evaluation down, but it will play worse. You can make an engine reach great search depth very quickly by aggressive forward pruning. It will also play worse.
What is needed is a combination of good (or rather: good enough) evaluation, good move ordering and aggressive forward pruning/reductions. It's the combination of these things that makes a program like Stockfish so strong. Pruning decisions are based on the evaluation, so a good-enough evaluation is needed to cut down the search tree to something manageable.

What is obviously true is that if you limit a program to a certain depth (say N) and play it against itself with a larger depth (say N+1), the former is going to be annihilated. I think someone here did a test of this a few years ago.

As I know you made chess programs so I suppose that you know well the books and articles about engines.

I don't. The only book I ever read on the subject is the one in Dutch by Jaap van den Herik, and I think the only papers I've read on the subject were Bob Hyatt's bitboard papers.
Actually, that's not quite true - I've read others on different search algorithms, but only a selection. I'm by no means an expert.

But if you do not want to agree with the sentence cited by me please debate with the original authors.

My point was, you're not making it very clear what you're citing: you don't name the authors, or the book, and what you claim ("speed is most important") is ambiguous.

See, when you say "speed" I tend to think "nodes per second", or even "efficiency of move generation". Which is pretty much irrelevant (within reason). What I now suspect you meant is "time to reach depth X", which is something different entirely.

EDIT: so I guess I'm not so much saying that I disagree with what you say, but that what you say isn't very clear and can't really be discussed very well because of that. Also, "it's in a book", or "it's in a peer-reviewed paper" sadly doesn't mean it's right...

[Dann Corbit]

I have pointed to school-books and their authors.
Definition of speed arises from them.
Please, debate with them.

If you take Stockfish, and simplify the eval to only count wood, it will run faster. And be 1000 Elo weaker.

[corres]

[quote="Dann Corbit"][quote="corres"]I have pointed to school-books and their authors.
Definition of speed arises from them.
Please, debate with them.[/quote]
If you take Stockfish, and simplify the eval to only count wood, it will run faster. And be 1000 Elo weaker.[/quote]

Please, read my primary note:
"I think that developers of Stockfish, Houdini,and Komodo have a success in balancing between speed and knowledge."
Naturally a well balanced engine like Stockfish will degrade on an effect mentioned by you.

[corres]

Thanks for detailed answer.
Before my reaction I should express that in my opinion the main issue are my English is not so fluent than yours and you have not read too many books and articles about your hobby.

[quote="Evert"]
But what exactly [i]is[/i] the fact that you quote?
Do you mean that a better evaluation does not prevail against an increase in search depth of 1 ply? I think that is a result from the 70s or 80s, and it was probably true then. What it tells us is that a basic alpha-beta search with a simple evaluation that out-searches a basic alpha-beta search with a complicated evaluation does better.
Modern chess programs don't have a basic alpha-beta search though, and they pretty much all have a lean and simple evaluation function.
You can make Stockfish (say) "faster" (in terms of nodes/s) by stripping the evaluation down, but it will play worse. You can make an engine reach great search depth very quickly by aggressive forward pruning. It will also play worse.
What is needed is a combination of good (or rather: good enough) evaluation, good move ordering and aggressive forward pruning/reductions. It's the combination of these things that makes a program like Stockfish so strong. Pruning decisions are based on the evaluation, so a good-enough evaluation is needed to cut down the search tree to something manageable.
[/quote]

If you think that behavior of modern engine deviates from the old ones,
please make a lot of tests to prove it and write an article on your result.
I shall read it.

[quote="Evert"]
What [i]is[/i] obviously true is that if you limit a program to a certain depth (say N) and play it against itself with a larger depth (say N+1), the former is going to be annihilated. I think someone here did a test of this a few years ago.
[/quote]

Who is "someone" ? You expect from me to cite precise and you...
It is obvious that difference in value of evaluation between depth = 1 ply and depth = 2 ply may be more higher than let's say between depth = 30 and depth = 31. In the later case the annihilation is rather possible.

[quote="Evert"]
I don't. The only book I ever read on the subject is the one in Dutch by Jaap van den Herik, and I think the only papers I've read on the subject were Bob Hyatt's bitboard papers.
Actually, that's not quite true - I've read others on different search algorithms, but only a selection. I'm by no means an expert.
[/quote]

Thanks for your sincerity.

[quote="Evert"]
My point was, you're not making it very clear what you're citing: you don't name the authors, or the book, and what you claim ("speed is most important") is ambiguous.
See, when you say "speed" I tend to think "nodes per second", or even "efficiency of move generation". Which is pretty much irrelevant (within reason). What I now suspect you meant is "time to reach depth X", which is something different entirely.
EDIT: so I guess I'm not so much saying that I disagree with what you say, but that what you say isn't very clear and can't really be discussed very well because of that. Also, "it's in a book", or "it's in a peer-reviewed paper" sadly doesn't mean it's right...
[/quote]

I think this is not the forum of science where is expected the precise quoting. Because of the common themas it is supposed to know the basic letters.

[Guenther]

...

No one can read your posts anyway until you finally enable 'BBCode'
for readable quotes in your replies, as you were already told a while ago.

I doubt though that it will help much for the semantic part.