Author: Howard Exner
Date: 18:25:53 04/26/98
Go up one level in this thread
On April 24, 1998 at 11:52:29, Don Dailey wrote: >On April 24, 1998 at 03:47:44, Howard Exner wrote: > >>On April 23, 1998 at 23:36:30, Robert Hyatt wrote: >> >>>On April 23, 1998 at 21:13:32, Howard Exner wrote: >>> >> >>>>>IMHO, you should change 6 months to a year to 6-10 years. The Micros >>>>>are *not* a factor of two away from playing evenly with GM players yet. >>>>>Not even close... >>>>> >>>>>We already have a PII/400 at the office. Intel says 450 late this >>>>>year, and (possibly) Merced next year. We already have machines as >>>>>fast as merced... they are spelled "alpha"... no one's *close* to a >>>>>GM yet... >>>> >>>>IMHO you should change "no one's *close* to GM yet" to no one's >>>>close to a super GM yet. I believe that the top micro's on this fast >>>>hardware >>>>play at a weak GM level (2450 - 2550). >>> >>>at faster time controls, yes. at 40/2, no. not yet... >> >>Are too :) I was hoping Bob would counter here with "Are not!". I had the witty, "Are too - infinity! nyah nyah" ready for him. Boy he's lucky to have dodged that one. >>> >>>an occasional win will happen. but when you see a "top micro" get >>>squashed by a USCF master, you *know* it is not a GM. >> >>Since we are talking elo and 40/2 post such an example of this >>"squashing". >>There isn't any on record that I've seen. >> >>>GM's don't lose >>>to masters very often... yet on ICC these programs lose to them on a >>>regular basis... except that the "operators" won't play such a player >>>more than once... >> >>We've been here before on this topic as it seems to emerge from time to >>time. >>I know of your arguments that depict weaknesses in computer play and >>have >>no argument with any of them. I am largely basing my belief on years of >>playing over thousands of games. While I am no more than an Expert level >>player >>I do believe that I can preview a game and recognize good vs bad play. >>Not >>unlike recognizing high levels of play in Sports while not being able to >>duplicate what I am viewing. (as others do also) >> >>I believe I'm right in claiming weak GM levels of play for the micros. >>I could change that belief but have read or seen nothing that can show >>me that "micros are not close to GM level. Remember that there are very >>many GM's rated 2450 -2500. I'm interpreting not close to GM as to mean >>roughly 2350 or so. >> >>One of your major arguments is still flawed. The one that goes like >>"I watched an IM beat Crafty on the servers last night". Those are >>definately >>not 40/2 encounters. I thought the highest ratings on the servers belong >>to computers. Is that so? >> >>Anyway it's fun to rehash this "how strong are the micros". We will no >>doubt >>have the same discussion in a few months from now. :) > > >It's probably appropriate to revisit this question every few months >since >the answer changes every few months (How strong are the micros). > >By the way, I want to comment on the time control thing. We are >picking >a typical world championship time control which I think is appropriate. >HOWEVER, the assumption seems to be anything much faster is hugely >distorted >in favor of the computers. I completely disagree. In time controls >like >the Aegon, in my humble opinion, there is a little, but not a whole lot >of >difference in the relative strengths of humans vs computers. When you >get down to speed chess (game in 5 seconds) there is even more >difference >but even then I'll bet it's not as much as most of us imagine. Don't >forget that computers play TREMENDOUSLY weaker at 5 seconds than they do >at 3 minutes. > >I think this is another example of the human's inability to make fine >grained judgements. We see most things as all or nothing (especially >if it supports our point of view.) Most of my computer assumptions on this topic (ie: how time effects chess playing strength of computers) comes from my reading the ICD computer chess reports (the old printed ones). There always seemed to be some article on programs playing so extremely strong in blitz tournaments with strong GM opposition. Then came the articles on rapid events (Genius beating Gary in London) showing that the were making inroads there as well.) >There is only a gradual weakening >of the humans with respect to computers which at speed chess finally >makes a fairly significant difference. Does anyone have any hard >data on how much the average difference is between game in 5 vs 40/2? I have something posted in relation to this on Bob's recent "Crafty on the Net" thread. I do recall some of Larry Kaufman's articles on this. Did he not write that there was an appreciable gap between blitz and 40/2? > >I'll bet the curve is very sharp too. I would guess that humans do >remarkably better at game in 10 than they do at game in 5 and that >at game in 1 hour there is very little difference. My guess is that the increase of the level of play as humans receive more time is linear (that based on no data, just a personal opinion). >If you factor >out time pressure issue by using Aegon like time controls (fischer >clock etc.) I'll bet this difference is even less. >- Don
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.