Author: Robert Hyatt
Date: 15:40:40 12/05/97
Go up one level in this thread
On December 05, 1997 at 16:00:21, Howard Exner wrote: >On December 05, 1997 at 08:52:51, Robert Hyatt wrote: > >>>Place Name Points Rounds Buchholz >>> 1 REBEL 9.0 10.0 11 82.0 >>> 2 Hiarcs 6.0 8.5 11 83.0 >>> 3 Nelson Pinal (IM) 8.5 11 80.0 >>> 4 Manuel Dominguez (FM) 8.5 11 79.5 >>> 5 Gustavo Hermandez (IM) 8.0 11 78.0 >>> 6 Fernando Cabrera 8.0 11 74.5 >>> 7 ChessMaster 5000 8.0 11 73.0 >>> 8 Luis Arzeno 8.0 11 71.0 >>> 9 Salvador Potentini 8.0 11 66.5 >>>10 Ramon Mateo (IM) 7.5 11 81.5 >>> >>>Bob, do you now finally believe the current micro's play "at least" >>>on IM level on tournament time control :)) >>> >>>- Ed Schroder - >> >>Unless I am terribly forgetful, I believe I have said this all along. I >>don't believe they play at the GM level yet. They are tactically strong >>enough that anyone can get into trouble until humans learn how to play >>against computers. But the GM players know "oh so much more" than all >>of >>the micros added together that the micros are simply not there yet... >> >>I don't know any of the IM players above. I do know several on the >>chess >>servers and every one of them are stronger against computers than the >>above >>group. > >How do you know they are stronger at 40/2 in a real tournament? > First, I know their FIDE ratings. Second, I have watched a couple of these players completely shred commercial micros in blitz. Third, I've played chess myself for 40 years now, and am decent enough that I can see the differences between the best programs and the best humans. Fourth, I chat with lots of these IM and GM players regularly, as do others on the servers, and they can tell you exactly what the program they just trashed didn't understand, be it a bad bishop, a good knight, an exchange sacrifice, a mobile majority, a weak pawn, weak squares or whatever. Things that the programs understand after a fashion, but only *barely*. So there's plenty to convince me of the difference between the strong IM players (not to mention the GM's) and the computers. As I said, we are mostly surviving on tactics at present, with occasional nice games where positional knowledge fits real well and helps the program play like a GM at times. But only "at times". And when a good GM (Roman is a killer for example) learns when these "times" are, he will avoid them every time. >>The problem is that humans adapt. > >Yes they do. > >> >>Again, rig up an automatic interface and play on any server where there >>are IM and GM players. > >Didn't the gm's get their butts kicked last year in the game in 30 >internet tournament( I believe Crafty was part of this butt >kicking contingent)? Of course like you I believe that the >computers are not GM strength. Where do you place them? >Strong Expert? Weak IM? I peg them somewhere between weak IM and "middle" IM class. But they are erratic as the devil. IE even a weak IM has more chess knowledge than the best of the programs, but is probably weak in tactics in some positions. But even weak IM's are to be avoided in endings for example, because often the winning plan is far too deep to search, and knowledge has enough holes in it that the exceptions come more often than not... Yes 4 or 5 computers rolled up an equal number of GM players in a game/30 match a while back (Crafty and Ferret were two of the programs). But beating them in those games is simply fortunate. You can just "see" what the GM's can do... > >>Then come back and tell me how far behind the >>good >>humans you really think you are, after watching. > >Why bother when a real tournament result has been presented here. >Ditto for all the excellent Junior results in real tournaments. I'm not saying "toss 'em out". I'm saying that most of the GM's that play are *not* computer-aware *yet*. the ones on ICC *are*. There is a world of difference between a strong player, and a strong player that has computer experience. We used to call this "computer-shock". It still exists. > > >IE from what I have >>seen, >>both Rebel and Genius will have severe problems with the "attackers" >>because >>of the way you do king safety. In the few games I've seen Rebel 8/9 >>play >>against these players, it has a hard time deciding what is going on, and >>often attacks when it thinks it is appropriate, but it seems to not >>under- >>stand diagonals and attacks from pieces across the board. > >Could you post these games? I'm interested in playing over >these games(include time controls) but they seem harder to find. >Recall how swiftly Bronstein disposed of computers in past Aegon >events using bizarre anti-computer strategies but how in the >most recent event this didn't work. > I'll see what I can dig up. Many of the GM's only play very fast time controls so the games won't end up in ICC's databases, as that seems to be a "GM nightmare"... > >>Against >>*most* >>players, this isn't a problem. Against "computer-aware" opponents, it >>is >>a *serious* flaw. you just haven't encountered those "computer-aware" >>opponents in any numbers yet. > >Are you saying the Aegon human players are not "computer-aware"? I'm saying that they are not "anti-computer" yet. IE I play dozens of games against GM's every day. Almost every game is what I would call "legitimate chess" with no anti-computer stuff, because the GM's are using this as an opportunity to brush up on something, not learn how to bust computers since they don't see 'em in the tournaments where they try to make money. The IM's are more likely to play anti-computer for some reason, and they often give programs more problems than the GM's do for this reason... > >>If you visit ICC, I'll be happy to >>introduce >>you to a few. :) I have met them myself. They are a tough lot. But >>you >>have to do this with an automatic interface so they can play enough >>games >>to become familiar with what you can/can't handle... one or two isn't >>enough. I've entered tournaments similar to the one above and finished >>in first place as well. But in looking at the games, I can clearly see >>who should have won. IE Crafty won last year's Pan Am tournament here >>in >>the states, with some really strong competition in that event. But it >>isn't >>a GM. It isn't even a "near-the-top IM". I don't think any are. >>*yet*. >> >>We are getting by on a little knowledge and a lot of search. There are >>players that can search as well, and have a *lot* more knowledge. They >>are >>a big problem... > >This a becoming a common response whenever someone posts good results >from a 40/2 tournament. Bob, do you believe that ICC results are an >equal >indicator of computer strength as are real tournament condition results? difficult to say, because we don't have *many* non-ICC results to look at. IE AEGON once a year, Crafty in the Pan Am once a year, etc. The body of games on ICC is 1000X larger. There are dozens of genius clones, CM5000 clones, Fritz Clones, Hiarcs Clones, Rebel Clones, and hundreds of Crafty and GNU clones. And don't forget, look *carefully* at the 40/2 tournament. How did *all* the programs do? IE don't just look at the one that finished near the top, because with 50-60 different ones (and multiple copies) playing, a couple are bound to do well. But what about the "average"? In the 80's a program getting a good result was earth-shaking, because I played in USCF events where I was the only one, or perhaps had one or two others playing with me out of a field of a couple of hundred.
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.