Author: Robert Hyatt
Date: 19:57:31 07/17/00
Go up one level in this thread
On July 17, 2000 at 16:00:30, Jeroen Noomen wrote: >On July 17, 2000 at 14:58:22, blass uri wrote: > >>>In open positions DJ scored 4 out of 6, which is a TPR way above 2800. >>>In closed positions DJ scored 0,5 out of 3, which is rather bad. >> >>I think it scored 2 draws against akopian and adams in closed positions. > >The game vs. Adams was a Ruy Lopez exchange. This is surely not a blocked >position. It was an ending, quite open with White having the better pawn >structure and Black having the two bishops. I absolutely do not count this one >as a blocked position. In a blocked position the situation of the pawns is thus: >They face each other and can't move. Clearly this is not the case in this game. > Careful. I go way further than that in Crafty. IE you have a pawn at h6 and no g pawn. I have a pawn at h3 and no g pawn. I call the h pawn blocked. Or you have pawns at c5, e5 and d6. I have pawns at c4, e4 and d3. I count the d3 pawn as blocked (d6 also). It isn't going anywhere. There are many cases. I know what you _meant_. But it has been amazing how often Crafty did what I said, _not_ what I _meant_. :) I finally decided to two-prong this issue. I catch blocked pawns, but I also count 'useful levers' and get real concerned when the number gets too low... a measure of 'blockedness'... >>I also do not think that it played clearly open positions against Bareev or >>Huebner. > >Huebner tried to block the position but he failed. And at the first opportunity >he made a big tactical mistake, and committed another one by resigning too >early. Bareev sacrificed a pawn early on, so it had nothing to do with the type >of the position. The only question in that game was: Can Bareev prove he has >enough compensation or not. The ending was quite tricky, also having nothing to >do with blocked positions. > >Besides, you (and all the others!) do not come up with arguments to prove that I >am wrong on my thesis that computers obtain GM results, but that their knowlegde >is far behind. Does nobody want to disagree with this!? Nobody wants to question >my remark that calculation power doesn't solve the problems computer programs >are still facing!? Come on guys! I happen to agree. But then I have watched _far_ more GM games than most have, on ICC, and chess.net. I see what they do to programs when they learn how to do it. I have had many pleasurable battles of wits with a couple of GMs that decide "I am going to draw this thing every game if I can...". Their success ratio is very low for doing this now. But it is difficult to do it well and not wreck something else when doing so... > >I started my love for computer chess in 1980. I played the challenger 7, 10, >Voice and won easily. They were weak in all aspects of the game. Then came >Turbostar, Super Conny and still these were weak in closed positions and king's >attacks. Then I bought my first 'real' computer, the Mephisto MM 4. I liked that >one very much, because it played positional chess. Ed did a great job there. >With fast hardware he would have beaten Richard Lang easily with the MM 4 >program. Still, the same problem appeared: Although the MM 4 knew more about >blocked structures, it could be beaten by King's attacks or long term plans. > >After MM 4 I got interested in PC programs. Gideon, The ChessMachine, Genius >were the first one. Still the same problem: No good plans in blocked positions, >no understanding for King's attacks. Now we are 10 years further in time. PC's >are 100 times faster (I am not sure about this figure, I estimate it) and DJ >plays on a super PC, equalling 4,5 GHz. And what do I see: NO understanding of >blocked positions, no understanding of how to prevent or defend against a king's >attack. > >Conclusion: I am in the computerworld for 20 years. I saw them starting with >less than 1000 Elo. Now they have above 2600. But two problems remain, they were >not solved in 1980, not in 1990 and not even in 2000: Blocked positions and >king's attacks. > >Jeroen > >Jeroen I wouldn't say totally unsolved. (a) progress has been made; (b) deep blue seemed to have no problems with this issue, based on a bunch of games it played while winning the fredkin prize on thru Kasparov. Of course, they could do whatever their GM coaches wanted with no regard to computational cost, since there was none...
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.