Hi Geno,GenoM wrote:I don't argue about that computers are stronger than people.Don wrote:I am one of the worlds leading authorities in computer chess and I am also a tournament chess player. I have expertise in both fields and due to this I am far more qualified that even Grandmasters to make such a statement.GenoM wrote:I missed his name in the list, my mistake. But nevertheless it'd be interesting to hear his opinion too.Don wrote:Larry Kaufman is a Grandmaster and you are probably missing others too. Although we work on Komodo together Larry has most of the say on evaluation issues.GenoM wrote:FIDE master.Houdini wrote:What minimum rating do you have in mind?GenoM wrote:Ask programmers that are strong chess players.
AFAIK, there is 4 chess programmers that fit in: you, Vasik, Vincent and Miguel.
Interesting to hear your opinion on the matter.
Regards,
Geno
There is some issues in defining the topic. What is 'positional' sacrifice, eg.
There are some bold statements from your part.
For now I'm stopping at the last one -- about chess engines that sometimes play better positional moves than Super GM's.It's unclear to me how did you reach to such a conclusion if you're not a reasonably strong chess player. Is there some proof for this or it is based on your belief? Please, explain.Don wrote:In most cases the combination of excellent evaluation and a highly selective search that chases down as many ambiguities as possible gives you a program that plays positional moves better than Grandmasters.
Regards,
Geno
I just posted something about the beliefs of Grandmasters a few decades ago, and they were MUCH stronger than I am at chess but I was right and they were wrong because they simply didn't understand how or why they were good at chess. It's like a bird that knows how to fly but cannot explain HOW he does it. It was their belief that no chess program could break the master barrier because it required skills that only they could possess, not computers. Being a strong chess player did not qualify them to understanding much about how chess skill works. In fact being a strong chess player may even have been a barrier for them as it obviously stripped them of their objectivity.
Having said all of that, I think that what I am saying at this point is pretty obvious to most people anyway. Good players tend to get beaten by computer even before their first tactical error. Have you not noticed that the top computers are giving grandmaster odds now and still beating them?
There is a saying that chess is 99% tactics and it explains why programs are stronger, don't you think?
And there is 1% strategy in the chess game where people are better, but unfortunately that's irrelevant regarding to chess strength.
With all respect to you as "one of the worlds leading authorities in computer chess",
regards,
Geno
The saying that chess is 99% tactics is just a saying, it's doesn't have any meaning except to express that tactics is really important in chess. You could also say that chess is 90% tactics and it would be no less true or false since this saying is ill-defined anyway.
But you know that chess is not 99% tactics if you play chess at all. What chess player doesn't understand the importance of pawns, what a weak pawn is, what the consequences of double pawns, both good and bad is? Entire books are written in trying to help players appreciate the importance of positional play and we have grand-masters noted for their positional play and others noted especially for their attacking play. They don't get this recognition because they are slightly better at 1% of the game. So really it's quite clear that even though tactics dominate chess, positional understanding is an extremely important part of the game.
When people talk about this stuff they greatly simplify things in their minds. They talk about tactics as if it is completely separate from positional play and one has nothing to do with the other, but that is how the brain works - it's useful to choose simplistic models but not to take it too far.
The reality is that if you look deeper to improve your tactics, your positional play also benefits enormously too. Here is a simplistic example to make my point:
Let's say that you have a program that does not have any positional evaluation, just a table of piece values. Even given a 30 ply search it is going to play horrible chess. However, there will be many cases where it will play fine positional moves, for example a clever move to avoid getting a weak pawn. It might do this simply because it's the only move that doesn't eventually lose the pawn in 30 ply. In a very real sense, ANY KNOWLEDGE that you place in the evaluation function will be used to simulate unimplemented positional concepts. Even the purely tactical concept of the value of a piece will be "bootstrapped" into at least a rudimentary simulation of positional understanding.
Of course the better evaluation you have, the more powerful this "bootstrapping" will be. If you already understand the value of a weak pawn you won't need a 30 ply search to avoid getting one, but you might use that knowledge to avoiding the threat of getting one, a threat that may distract you from something else more important.
So what programmers try to do is to fill in the big missing gaps in positional knowledge so that the computer can find the same things with less depth and use that knowledge to leverage other concessions from the opponent. The combination is lethal, it's really difficult to outplay a great chess program positionally.