Author: Robert Hyatt
Date: 16:01:25 04/26/00
Go up one level in this thread
On April 26, 2000 at 18:28:15, Jeroen van Dorp wrote: >>No, but if he believes that Crafty does well _only_ because it has superior >>hardware, then that same superior hardware should make _any_ program do well, >>unless the semantics of the English language have changed more than I >>realize. > >Bob, please keep the same academic exactness with your reading as with >developing Crafty :) Don't get swept away. I thought it good to reproduce the >whole paragraph from the article, just to show that the KC review is a bad piece >of text crunching, and that it's not as bad as it seems there. > >It didn't help, I realize that now. I'm among the believers. Is that the goal of >this discussion board? > >Whta's gone wrong? "Let's test all programs on the same hardware, because I >don't know what the high ratings on ICC mean *in comparison* to other top >programs on different (slower) hardware" gets *mutilated* to "Crafty is only >strong because it's on superiour hardware, for the rest it's a lousy program". > >I'm not interested in accusing or defending De Gorter, just becoming more and >more curious why there is such a strange reaction from people here to critics on >Crafty. I would be glad instead. It must be my problem, no doubt. I don't mind criticism. Pointed criticism. IE "Crafty doesn't handle kingside attacks well enough and gets ripped apart too easily by XXX". Or Crafty doesn't understand an outside passed pawn well enough and loses too many endgames due to that shortcoming. But making statements not backed up by his own data is a bit over the top. Best rule of thumb about _any_ program: if you can't find something good to say, don't say anything. If that is to be violated, then it should at least be violated by presenting some data, some positions, some results. The general idea has become no matter what Crafty does, it only does it "because". IE someone on ICC says hey, I don't think your quad xeon has a chance against my 1ghz athlon running XXX, care to match me to let me prove it?" I say sure. He gets smashed. He leaves with "yeah, but you have such a big hardware advantage you should win..." as though I initiated the thing. If crafty wins the ICC tournament, it is either good luck for it, bad luck for some favored opponent, or a book problem, or a hardware advantage. If it has a bad result, it is because it is no good, it plays poor positional chess, it can't play endgames, it is too weak tactically. in short, no matter what it does, there is always an excuse for why it did it. And why it really shouldn't count. New ICC record blitz? Aw, it isn't playing any computers, only humans... anybody can do that. Can they? Have they? I don't think it is so easy. It is easy to talk the talk. Harder to walk the walk. I don't think this will change. Every good result is greeted by hoots and jeers. Every bad result is greeted by "I told you so"s. Can you say "lose-lose situation"??? :) I became use to this in the days of Cray Blitz. "You can't hope to compete with people from the top CS universities in the country." "you can't hope to compete with the commercial guys." "You can't hope to compete with the guys that have been winning these tournaments so long." I win the WCCC in 1983, it's a fluke. I win it again in 1986, it is a bigger fluke. Hitech played better chess. the CB guys cheated. there _must_ have been a reason. I sit here, the only person on the planet with two WCCC trophies in my office, and see that kind of response. And people wonder why I get a bit sensitive about reporting that really is not classic 'impartial journalism'. Nobody would take the results de Gorter had, write the conclusions he wrote, and spend more time putting down a freeware program than he did writing about commercial programs, all in an article supposedly a review of commercial chess programs. Not even an honest title, IMHO... IE "How does Crafty stack up against the commercial chess programs?" I see the flaw in the arguments he wrote. Others see them as well. I just don't feel like spending a lot of time on that particular subject, because I _really_ don't spend any time worrying about what he thinks about Crafty. Any more than I spend time worrying about what Kasparov thinks about it. If he or de Gorter thinks it is weak, I am sure they can visit ICC and demonstrate how weak? I have more than enough good comments from a strong group of GM players that are 180 degrees out of phase with his comments. Yes it makes mistakes, but I haven't found a one that says "this thing plays positionally weak" or "this thing has no idea about chess." Because at blitz it demonstrates just the opposite. Not all games are won on tactical brilliancies. It plays some really outstanding endgames as well. You can watch it on ICC if you want. Even if de Gorter says it doesn't... > >As a Dutchman you can guess I'm no expert om English semantics, and reading the >article in Dutch and translating it can be awkward, but I think in this respect >my translation is okay. > >He *does not* say Crafty deserves its high rating because of the superior >hardware. He tells us: I saw Crafty online is very strong. But it is running on >this superfast hardware. It's difficult to say something about relative strenght >if you don't test on the same hardware. He is right. Most programs don't swallow >dual or quad processors; Crafty does, but then you should compare a quad 500 for >Crafty with Fritz on let's say a Celeron 333. > >Well then, how should you compare it with top commercial hardware? Let's all >test them on the same hardware, be it "superiour", " regular" or "obsolete". >That's the base line. Maybe that's wrong. Say so. > >About all his conclusions, I have my own thoughts, but I think they're not >interesting. > >If I were you, I would be happy with every piece of commentary on my program. >Maybe I understand it wrong, but I consider Crafty a development project, >boosting computer chess in general. >The last *is* true, as a lot of top comemrcial programs pay respect in their >programming to Crafty. You know that better than anyone. > I would be perfectly happy if CCC set up a filter that automatically deletes _every_ message with the word "crafty" embedded in it. I would enjoy the discussions a lot more if nobody, including me, used the C-word... >De Gorter says: Nimzo 7.32 is the most balanced playing program, Crafty plays >unbalanced. If I were *you* I could come up with: I don't think he knows a pawn from a band-aid. Maybe "unbalanced" is what I am shooting for. At least if balanced -> slowly getting ground up in a simple but long endgame. Yes I am trying for unbalanced positions. But I don't think that is the definition of unbalanced he is using. I don't think he is strong enough to be able to assess what is balanced and what is not if he can't at least beat the thing. I'm convinced that many of Kasparov's "blunders" (as found by computers) are simply plans so deep the computer has no clue. So the machines, like de Gorter, say "poor play, that loses 1/2 a pawn" when in reality, it might be winning the game... > >1. it's version 16.6, I know that, been there, done that, old stuff; >2. that's strange, show me those games, tell me those settings, not only from >Crafty, but from Nimzo 7.32 as well. Let me validate your results. More proof >please, no vague interpretations; >3. I will show you that superiour hardware does not affect performance that >much. > >Only a few possibilities. Superior hardware does help. Otherwise I have wasted a lot of time. But it is not enough in and of itself. It still takes a decent program, or a GM will eat it up if it is running on the fastest machine on the planet. > > >Maybe I'm wrong and Crafty is just a private love baby, with you not interested >in progress. Maybe you're only looking for compliments, who knows. I'd be perfectly happy with no compliments, no complaints, no results, no mention of it at all, since they generally turn toward something less than pleasant. > >I guessed not. > >I did scientific research myself. I know that discarding critic emotionally is >as bad as accepting results from a few tournaments or some uncontrolable >observations. It's both the end of science. > >Prove me wrong. > >Jeroen ;-}
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.