question to prof Hyatt re kaufman and best positional moves

[duncan]

To Prof Hyatt,

could you comment on this post by kaufman. ?

also do you have a rough estimate of how many elo points todays top program's valuation is worth.

My apologies if you have answerd this all before, as I have missed it.

http://rybkaforum.net/cgi-bin/rybkaforu ... l?tid=6547

If you pick an engine that has only material values and no knowledge (I'm sure some exist), that would be a pure test of your idea that a human could compete with Rybka 3 if he never made tactical errors (within the depth a computer could search). I am sure that the result would be an overwhelming victory for Rybka 3. Rybka 3 would win not because of superior chess knowledge/understanding (compared to strong GM) but because it is implemented at the end of a deep search.

[LarsA]

To Prof Hyatt,

could you comment on this post by kaufman. ?

also do you have a rough estimate of how many elo points todays top program's valuation is worth.

My apologies if you have answerd this all before, as I have missed it.

http://rybkaforum.net/cgi-bin/rybkaforu ... l?tid=6547

If you pick an engine that has only material values and no knowledge (I'm sure some exist), that would be a pure test of your idea that a human could compete with Rybka 3 if he never made tactical errors (within the depth a computer could search). I am sure that the result would be an overwhelming victory for Rybka 3. Rybka 3 would win not because of superior chess knowledge/understanding (compared to strong GM) but because it is implemented at the end of a deep search.

To understand Kaufman's post, the question answered is needed:

Code: Select all

What do you think about the following match idea:

Rybka 3 plays at normal conditions against a strong human, an this human is assisted by an engine, which does not give positional evaluations (or very poor positional evaluations), but it gives strong tactical values, combinations, and the mainline of that combination.

Maybe this could be the R2-winfinder,
or it could be a Toga with castrated evaluation function,
or something else.

Would we have a match then, where the positional abilities of the human and Rybka will give the deciding fact?
What result would you expect?

Quap 

[bob]

Here is my opinion. If (really big if) you can design an interface that does not cost the human time, then the human would win hands down. If he makes no tactical errors that the opponent could see, and if he can use a real GM evaluation to choose between moves that are equal tactically, then the computer would be helpless.

But the interface is the issue and does not exist today... Correspondence would be interesting, where the human could choose the objectively best move from a list of moves known to be completely safe tactically. Then the interface would not be such an issue.

[duncan]

how much extra time would you need to give the human to compensate, because you do not have such an interface ?

[duncan]

http://rybkaforum.net/cgi-bin/rybkaforu ... ;pid=97067

Meaning no disrespect to Prof. Hyatt, unless I am misunderstanding him somehow I think his answer is totally wrong. He seems to imply that if a Grandmaster is given an engine that knows only material values, and some reasonable time handicap to allow for engine use, then the GM would be strongly favored against Rybka 3. If we can agree that double time is enough to make up for interface inefficiency, and if we agree to stipulate a 2600 FIDE rating for the GM (+ or - 25 Elo), I would be willing to wager $5,000 of my own money on such a match, say of four games at 40/2 vs. 40/1 (I'm betting on Rybka 3 of course). We would jointly fund the match. Any interest?

[Uri Blass]

Here is my opinion. If (really big if) you can design an interface that does not cost the human time, then the human would win hands down. If he makes no tactical errors that the opponent could see, and if he can use a real GM evaluation to choose between moves that are equal tactically, then the computer would be helpless.

But the interface is the issue and does not exist today... Correspondence would be interesting, where the human could choose the objectively best move from a list of moves known to be completely safe tactically. Then the interface would not be such an issue.

Here is Larry kaufman's reply from the rybka forum:

http://rybkaforum.net/cgi-bin/rybkaforu ... 7#pid97067

Uri

[Uri Blass]

We need to define exact conditions that Bob Hyatt expect the GM to win to be able to test if he is right or wrong.

The main problem with only material engine is that Bob may think that it is weaker tactically than rybka and not only positionally.

Uri

[duncan]

I think rudolf huber's sos has a material only version. if both rybka and sos were limited to 16 ply , would that be fair, or are ply defined too differently.

what happens if mr kaufman compiles a version with all the evaluation code removed, would that be a reasonable test.

or would a limited form of takebacks with a referee be needed in case it is impossible to make them tactically equal.

[bob]

how much extra time would you need to give the human to compensate, because you do not have such an interface ?

Hard question to answer. Dave Slate played a game vs IM David Levy at an ACM event years ago. Dave used chess 4.x as an assistant, and had a human helper to operate the program and have it search whatever he wanted so that he didn't have to "leave the game" to run the program. There was a discussion after the game and his main comment was that the "interface" (even with a helper) was so unnatural that it probably ended up hurting him more than helping.

I suppose that given enough thought, one could design an interface that would provide the information needed, but it is not the way current programs operate.

[bob]

http://rybkaforum.net/cgi-bin/rybkaforu ... ;pid=97067

Meaning no disrespect to Prof. Hyatt, unless I am misunderstanding him somehow I think his answer is totally wrong. He seems to imply that if a Grandmaster is given an engine that knows only material values, and some reasonable time handicap to allow for engine use, then the GM would be strongly favored against Rybka 3. If we can agree that double time is enough to make up for interface inefficiency, and if we agree to stipulate a 2600 FIDE rating for the GM (+ or - 25 Elo), I would be willing to wager $5,000 of my own money on such a match, say of four games at 40/2 vs. 40/1 (I'm betting on Rybka 3 of course). We would jointly fund the match. Any interest?

I'd say you would lose that wager with high probability. But the interface is the first issue. If the GM has an engine that is tactically as strong as Rybka, he will be a "bear" to play against.

But you would have to have some GM friends to understand why this is true. Their positional understanding of the game is simply remarkable, particularly when compared to any program that almost certainly has less than 10,000 lines of C in the evaluation.