Author: Rolf Tueschen
Date: 01:56:28 01/24/04
Go up one level in this thread
On January 23, 2004 at 21:27:43, Dann Corbit wrote: >On January 23, 2004 at 21:08:27, Bob Durrett wrote: >>I don't wish to muddy the waters too much but the fact is that chess-playing >>programs or machines do not enter tournaments with zero information known about >>them. Just as in human tournaments, prior knowledge known prior to any games >>being played in the tournament can be very significant. >> >>Consider a trivial example: Suppose a top GM is to play a chess match against a >>true chess beginner. It is known apriori that the top GM is a whiz at chess and >>the beginner is a washout. >> >>Will it take thirty games to determine who is better? No, it will take ZERO >>games. > >You are wrong. It will take 30 games before we know anything about the unknown >player. ROFL. Dann, you are taken away by your wrong science, believe me. Bob D. spoke of a really new and untrained newcomer, understood? So even a Fischer or Tal after learning the moves of the pieces played like beginners. The example was NOT meant to be "unknown in public tournaments". Of course then a genial player could become a real threat if he has played thousands of games against his uncle. See Morphy for example. It is very funny how we are misleaden if we rigidly follow our "science" and in reality something odd happens. The joke here is that we two discussed the SSDF and there it is NOT the case that a _weak_ newcomer meets a strong old player, but it goes the other way round. In the SSDF they match a very old program with a newcomer, out of the newest generation of progs, and we know in advance the result, because the old prog - most of the time also on weaker hardware [NOT equal hardware!] - has no chance at all. Here at last you should understand that comparisons with human chess are totally useless. In human chess the old experienced player will always be superior to a newcomer [this is what Bob D. said, but what you cannot believe because you are coming from computerchess methinks, but it is trivial for a chessplayer, I mean a real player who already played in human tournaments and clubs], and if I say newcomer, then I mean an untrained newcomer who had just learned the rules of the game. In that case, we dont need 30 games to predict the result but we need zero games to predict who is stronger! And for the SSDF we can say that without any games at all we can predict that these matches between outdated progs on older hardware and the newest generation, although these new entities are completely new [and unexperienced what SSDF is concerned, hehe], we dont need 30 games to predict who is stronger. It is always the new prog. And this is why it's complete nonsense how this is practiced in the SSDF. Now you said that these games against older progs on older hardware were extremely important. I know. Because on these older progs the SSDF based its validity. Because these older progs have played themselves against even older progs and this way the chain goes back to the once held competition between long time forgotten progs against Swedish masters. Hehe. Dann, this is all nonsense, to be true with. It's statistically complete nonsense, it sounds as if we were doing medical homoeopathy without one molecule left in the Oceans of computerchess. But we are still pretending that their is a linkage to the past. And it's true, we human beings have deep roots in our past but these machine progs are completely determined what this is concerned. We KNOW for certain what a new program will do with the old, out-fashioned warrior. Putty! And we can prove it with different depths and speed. So, my argument against SSDF runs like this. Because they are leaving the pool of always new generations and hypostate a linkage between these new progs and their predecessors and make unrealistic and predictable matches between old and new the results are worthless. Because they show, what we already knew _without_ a single game. Bob D. tried to explain that. Now what is happening. We are still talking about chess. And chess itself causes that the outcome depends on the chance of the chess position. Sometimes you are lucky, we chessplayers say. Now what you get with these short matches is something about luck that spoils the naked data that is completely predictable otherwise. And this ways SSDF gets a ranking list that looks like the new progs itself had important differences. But if you examine the deviations of the mere numbers the SSDF itself shows that the hypostated differences are within the limits of normal expectance. The normal variations are the reason for the impression of always new dufferences of basically equal entities in a pool of a new generation. Because hardware is the main factor of today's computerchess. You must be an experienced scientist to understand the hoax of statistical results. But you and everybody else once learned that to keep up a good stats you absolutely MUST keep your measured variable under control, better you must control all the remaining variables you dont measure, because otherwise you dont know what influenced what. This is the base of all SSDF activities. They compare apples with beans. And they publish honestly here are beans and there we have bananas. And beans rank on the top ranks actually. But the mind of human beings is lazy and it reads: Ahem, SSDF published its newest results, oh, surprise, this time Fritz lost position one aggainst Shredder, and some predict, next time Ruffian will be new number one. How interesting. > Consider this: >At one time, Kasparov, Fischer, and Tal had an Elo below 2000 and were >completely unknown. They came out of the woodwork and started blasting the >bejabbers out of people. Just because we know someone is talented, does not >mean we can use that data to extrapolate the level of talent of an unknown >entitiy. All correct. For human chess. > >>The number of games required depends on the prior knowledge about the >>contestants. > >There is no connection at all. This is again, as I said, only caused by lack of experience, that this is thought as a correct conclusion. In real the statement by Bob D. is trivially correct. But this is a question, which is more a pedagogic one. How to explain it to a real knowie who has learned his stuff and thinks he knows it all. Breaking the wall is very difficult. >However, we will gather more and more >information about the strength of the unknown opponent as more games are played. > He could be weaker, stronger, or the same as the great player. Imagine someone >who does not play humans but has played against computers for 5 years. He might >be a very good player that nobody has heard of. Of course, it is not likely >that a player will be better than Kasparov or Anand. But until the games are >played, we won't know. And 3 games against Kasparov will tell us very little. >Even if Kasparov loses all 3 games. For all he will lose. I bet. ;)) > >>I hope this is not too distressful for anybody. : ) > >Bad science. Using your intuition to do science is a very bad idea. Yes. But it is even worse to forget about your own thinking, doing science. Sorry, I didn't mean anyone here arounf personally. ;) > It is good >to form theories using intuition. But it is bad to assert the truth of your >feelings without testing. This is true, but the testing must follow iron rules of science. As I said, the most important is the control of the variables. If you dont you dont even know what you are "measuring". And your results are - yes - they are worthless. Rolf
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.