Author: Thorsten Czub
Date: 18:39:18 12/26/97
Go up one level in this thread
>How did you get this information? It seems very likely to be accurate >to me. It is an open secret in germany/austria for many years (as I know). But I have always registered that many people do not know or consider. You can easily find out by some special test positions that genius is NOT finding when computing on it. Shift the color to move ONE ply and genius sees the problem in nearly a second. So: from the root you can let it calculate 11 plies or even more, it will not find the key move although it reached depths of the plot. But it pruned away the key move with the 1st ply pruning. if you take back one ply or play forward one ply it sees the point in an instant. WHY ? because the important moves (threads and tactics) were now seen with the brute-force components of the search (plies 2,4,6,8...). I have tried to teach this in rgcc by giving positions where any genius user can find out himself. But the echo from this was very low. So I guess for some reason nobody is interested WHY genius always plays that boring and how it works. Chris always listens. The last person I tried to explain about was Stefan (Meyer-Kahlen). But this way of Richard is from my point of view overtaken by null-move ! If a normal symmetrical search is as fast as HIS approach, why use HIS way ? I discussed with Stefan that I believe that Richard shiftet the search AFTER he did it 1,3,5,7 asymmtrical way. I think he found a way to prove the results of the ONE method with a shifted search. I guess this method is a general working method to prune without making a blinder mistake, but the price is that you will get this INACTIVE playing-style. You will SEE any thread against you, but you sometimes do NOT see the best move of your own. Therefore your playing style is somehow passively. Since null-move is also an effective way richard lost lookahead. Therefore the other programs overtook him. He knew this. I always asked him to sent me a version of Genius that was symmetrical. But he was never able or willing to do so. I think by having on normal genius and another symmetrical one, you could crack the secret of genius completely. Therefore I think he was not "interested" in opening his cards by sending me a symmetrical version. Or - a second possibility - he is not able to write a symmetrical genius because this would not work, because it would be a contradiction. > Larry Kaufman is the master at figuring out which algorthms >a program uses by playing with it and perfoming special experiments >with problems and such. May be. don't underestimate german engineering :-) > We spent a lot of time trying to figure >out what Richard does, I think a lot of people have. One thing >that was clear though is that the evaluation is very good too. But >we were more interested in the search. I think the search idea is - as I said - an open secret for years. I am sure if you take a few programmers into one hotel room and a pc with some genius versions and a few testers who have their known positions they can crack genius search (so that anybody can retry it himself) after 8 hours study with some beer or red-wine :-) But the programmers should be compatible with each other. > >Other observations? His Pv's were almost always odd length. I had >the idea he only did evaluation on a given parity, for instance only >after the opponents move. In Paris, as I said before, we found out a special matrix of these main-lines by experimenting with selectiveness and limited-search-depth. Maybe Dan Wolf or the others from the hotel room remember this matrix. It was - when I remember it right (ply * 2)+1 or something like this. No - it was 2 to the power of PLY so (2 to the power 0 )+1= 2 plies main line (2 to the power 1 )+1= 3 plies main line (2 to the power 2 )+1= 5 plies main line etc. Dan wrote it down. I remember it only in weak images... But you can believe me, the concept of richards search should be easily cracked with some good people sitting together in good atmosphere. > >What else can you tell me so we can figure out his search algorithm?! I could give examples and - as I said - maybe dan remembers the scheme. >I have had lots of conversations with Richard but he never tells you >anything very important about his program. :-) Mr. Richard Lang who has lost the power to speak !! >Do you blame him? No ! I blame him that I tried 3 times to work with him and all feedback I got was nothing after a few email exchanges. Maybe the main problem was always my wish to get a genius with a symmetrical search. This was not possible for him to manage. > He >leaks small details occasionally but just enough to confuse you! But >he's actually a very nice person. I think the same. Nice guy. Without him computer chess would have been very very behind the evolution we have today ! Richard was the main engine for a couple of years and his unique search idea gave him some pretty 150 ELO points advantage. But - these times have gone. No HE is 80 ELO's behind them. And I doubt if he can reach the others. They TALK to each other and exchange ideas. HE is a secret and has lost the power to speak. How can you survive without talking with other strong programmers ?? We will see. > > >-- Don
This page took 0.01 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.