about IPP's evaluation and search

[liuzy]

We all know that IPP's evaluation function is much better than others.
What's the major difference between IPP's evaluation and others.

[wgarvin]

Isn't there a thread about this already?

[Gian-Carlo Pascutto]

We all know that IPP's evaluation function is much better than others.

Really?

[Eelco de Groot]

We all know that [Deleted]'s evaluation function is much better than others.
What's the major difference between IPP's evaluation and others.

Chris Whittington explained more than a decade past the "secret" of Rybka which is what we are discussing: the "Search Gap © Chris Whittington". In spite of all the progress, it still holds true I think. There was no Rybka yet but Chris was talking about Bruce Moreland's program Ferret as an example of a fast searching program, and not so much has changed in comparison to Rybka. Even the four nominal plies that Rybka searches deeper and does not want to show were predicted by Chris No new paradigms Unfortunately the original post was retracted because Chris claimed copyright on it and he had a conflict with other CCC members, he did not want his intellectual property discussed or quoted when he himself could not do so in the forum. Ed posted it on his website and even offered to frame it in gold but also there Ed had to remove it.

I do not know for sure if this is the original, if it is not it still is a discussion of the same thing, I think this was posted by Chris Whittington in response to Bruce Moreland more than ten years ago and hopefully Thorsten can keep the text on his website. I quote a middle and the concluding part of it:

Take a scenario. Your program now, Ferret, against your program 4 years ago. Or even your program now against your own program on slow hardware. Result inevitable ? Probably. Game style and type ? Probably predictable like so:

Ferret(fast) will have 1,2,3,4 nominal plies on Ferret(slow). Game style and type will be strongly dependant on the nominal ply gap.

a) High gap. Ferret(slow) will likely go down into rapid material collapse. Ferret(fast) may even have some flashy pyrotechnics to demonstrate it. A naive reviewer could call ferret(fast) a spectacular attacking program. He could call ferret(slow) a stupid bean-counter, typical computer.

b) Medium gap. Ferret(fast) will slowly grind ferret(slow) down. Ferret slow will keep finding at its higher iterations, possible loss of material. It
will go panic time, find a way to avoid material loss by giving double pawns instead, or whatever. A naive reviewer will call Ferret(fast) a great
positional player. He'll call Ferret(slow) dumb, accuse it of not having simple knowledge like double pawns, or whatever.

c) Small gap. Probably you'll get reasonable games. The reviewer can't tell much, so, if he'll likely start making things up. Human style, or plays more interesting, or some other nonsense that says nothing.

It's the search gap. Gettit ? Out of this search gap comes all the naive speculation and nonsense that gets written. The program has every style and no style, it has no consistency to play against, only materialism, you can't learn from it, tomorrow it will be different (found another mine in the search gap), only the difference is just a relection of - whoops, trod on another mine. What can you do with such a program ? Use the take-back key and try again ? - and imagine this helps you improve or learn ?

Now, I claim this search gap has no meaning or understanding possibilities for a human. That a human can't relate his heuristics to it. That you can't extract the knowledge out of it and represent it to a human. That you can't even extract the knowledge out of it and represent it to yourself. You can't get heuristics from it. So I call it counting beans - useless for us humans.

Now, take a knowledge program, you can play it and see the play style. You can try and work out what it does and why. There'll be a reason, based on human chess heuristics. The game has plan, and flow, and doesn't consist of hidden minefields. It won't grind you down by search, it will try speculative ideas which it might, or might not, be able to get to work. You can see the speculative ideas, and try them yourself. I think you can, as a human, relate to this type of program. If you know the programmer, maybe you can see patterns into the program that come from him, and so on. I think these types of programs are infused with some force, in so far as any chunk of silicon can be.

I hate materialists.

Chris Whittington

[zamar]

IPP's strength is in complicated and fine tuned search tree pruning techniques. There is nothing special in evaluation function.

[mclane]

it's on my web site for long:

http://www.thorstenczub.de/ihatematerialists.html

and not to forget chris' messianic text:
http://www.thorstenczub.de/complcss2.html

[mcostalba]

it's on my web site for long:

http://www.thorstenczub.de/ihatematerialists.html

I don't know this guy, Chris Whittington, but I have the impression that he was a prominent chess engine programmer in the last century. What he says about materialism and so on, looks a bit funny now days (no joke absolutely, I have great respect for his ideas).

The main point that now days seems to be obsoleted out is the anthropocentric vision of a chess engine:

"The program has every style and no style, it has no consistency to play against, only materialism, you can't learn from it, tomorrow it will be different (found another mine in the search gap), only the difference is just a relection of - whoops, trod on another mine. What can you do with such a program ? Use the take-back key and try again ? - and imagine this helps you improve or learn ? "

and

Now, take a knowledge program, you can play it and see the play style. You can try and work out what it does and why. There'll be a reason, based on human chess heuristics. The game has plan, and flow...

I think that computers are computers and humans are humans. Applying humans metrics to a chess engine is deeply misleading. The way a chess engine "plays" (but it should be more correct to say "returns") a best move has nothing to do with how a chess player chooses the move to play.

Trying to find a way to teach to a software what to do for playing chess human like is not recognizing that the strength in a computer lays elsewhere. The strong results you get IMHO when you teach a chess engine to play chess computer like !

Because you have to foster what they are good at and avoid what they are weak at.

I can understand his opinion was heavily influenced by a period in which GM could beat a chess engine so that was "natural" to be persuaded that a better engine would be one that could play like a GM.

Nowdays that this human-computer competition has been obsoleted out it is easier for us to recognize that a chess engine plays better when you are able to find ways and algorithms that more easily adapt to the way the computer operates and allow the engine to operate at its best.

These algorthims and techniques "computer friendly" are typical absolutely "human unfriendly". For instance computers go nuts when you give them a table full of silly and stupid numbers to lookup or when you give them as few and most simple ways (also if "dumb" from a human perspective) to generate moves, to search, to pick the next move and so on.

[mclane]

chris whittington was indeed a commercial programmer from the last century. he lived on an island, was a carpenter, and all kind of different jobs (even politician) and ended as a farmer, having many children, many animals and and and.

he also made many many versions of his chess program. so many that i have forgotten all the different names of these versions.

chess player 2150, chess simulator, chess player 2175, complete chess system, chess system tal, and many many clones of his own chess program sold all over the world.

the best text by him is IMO
http://www.thorstenczub.de/complcss2.html




https://chessprogramming.wikispaces.com ... hittington

[Uri Blass]

IPP's strength is in complicated and fine tuned search tree pruning techniques. There is nothing special in evaluation function.

You may be right or wrong but how do you know it?
It is possible that only the search is the reason that IPP is stronger than stockfish and it is possible that both search and evaluation are the reason that IPP is stronger than stockfish.

Did you test and find that replacing stockfish's evaluation by IPP's evaluation does not make stockfish significantly stronger?

Uri

[mcostalba]

Did you test and find that replacing stockfish's evaluation by IPP's evaluation does not make stockfish significantly stronger?

Copy 'n paste does not work. Thank you.


Does not work with evaluation, does not work with pruning, does not work with search

Copy 'n paste _definitely_ does not work.