Author: Robert Hyatt
Date: 10:03:42 07/20/00
Go up one level in this thread
On July 20, 2000 at 10:58:34, Chris Carson wrote: >On July 20, 2000 at 10:33:34, Robert Hyatt wrote: > >>On July 20, 2000 at 09:12:04, Chris Carson wrote: >> >>>On July 20, 2000 at 08:25:13, Robert Hyatt wrote: >>> >>>[snip] >>> >>>>I am not even sure about the 25%. If DB has a 90% chance of beating every >>>>program there, and I think that is reasonable, then in a 5 round event, the >>>>probability of winning every round is .9 ^ 5 which is 60% for 5 rounds. At >>>>a WMCCC-type event, we go for maybe 10 rounds... that comes out to a 34% chance >>>>of winning all games. >>>> >>>>However, as I have mentioned before, I saw DT forfeit in round one at the last >>>>ACM event, and _still_ it won the tournament, regardless of the hardware others >>>>were using. >>> >>>Show me the data that shows DB can beat today's programs on >>>today's fastest hardware 90% of the time at 40/2. I do not want >>>opinion here, you say 90% of the time, that is a bold statement >>>and I want to see hard data support. >>> >>>DJ 6 on 8x-700 => TPR=2702 >>>97 DB => TPR=2862 >>> >>>A difference of 160 points in favor of 97 DB. That does not imply >>>that 97 DB can beat DJ6 90% of the time. That gives a 72% probability, >>>keeping in mind that 50% means equality, 72-50=22% or about what >>>Ed said! A TPR of +366 points is required for a 90% expectation. >>>You must filp a very weighted coin. :) >>> >>>Best Regards, >>>Chris Carson >> >> >>That is an extrapolation based on simple fact. For 10 years, deep thought >>accomplished that at ACM and WCCC events. DB is about 100 times faster than >>Deep Thought. Since deep thought played its last game about 5 years ago, I >>don't think you will find that machines have gotten 100 times faster over the >>past five years. > >You say 10 years and then 5 years both in the same paragraph. Which >is it? I said (a) deep thought dominated computer chess for over 10 years. From 1988 through 1995. (b) deep thought last played in 1995 and at that time it was blowing all micros away, having (I believe) a total of two losses during that 10+ year period. So at it's last public match, at the 1995 ACM event, it was far better than anything else around. DB is over 100 times faster than DT was. So DB (if it played today) would be at the same level above today's programs as deep thought was in 1995. Or even further ahead since we know DB picked up well over 100X the speed of DT in 1995, and computers have not increased anywhere near 100x since 1995. HW today is more than 100 times faster than 10 years ago. >Easily. The 8xP3-700 is more than 50 times as fast as the P90 >of 5 years ago. You are comparing NPS of DB >and making an incorrect extrapolation. No external validity. > Nope. Read again. I compared deep thought to programs available in 1995. There was no comparison then. DB has gained more speed since 1995 than the micros have gained over that same period. The p90 wasn't a 5 year ago machine. I was using a P5/133's and P6/200s in early 1996. >I do not expect you to ever admit a mistake or that someelse >might have a valid point. I have never seen that. I do see that >96 DB is a dead issue and you try to prove 97 DB 90% supriority >based on DT vs 386 and 486 machines. This is just plain not valid. I admit mistakes _when_ I make them. In this case I did not. If you read my analysis carefully, you can see why I said what I did. I _know_ how DT did against the micros from 1988 through 1995. I know how much better DB is than DT, and I am only counting raw speed here, not the qualitative differences that were obviously made in the evaluation. I'd say that a program that lost exactly two games to micros over a 17 year period (1988 through 1995) was pretty dominant, wouldn't you agree? And since DB _gained_ ground speed-wise over the remaining 5 years, to bring us to today, I doubt that gap was magically closed. All based on past history, which is a pretty good indicator, IMHO. > >> >>The math is pretty simple. >> >>Your ratings are meaningless. You will find that the rating between two >>computer programs is completely unrelated to the ratings two programs produce >>when playing humans... You only have to look at some of the SSDF numbers and >>compare them to human tournament results to see that... > >Show me the data that supports DB to be 366 points better than DJ6 >on 8x700 hardware. You have none, you just make this up and >quote 10 years ago as valid. Today is 2000, not 1990. > 2 years ago a single DB chip played several matches with top commercial programs. This DB chip was running at 1/10th of its normal speed, and yet it won 36 out of 40 games. This has been reported several times here on CCC, by several that have heard Hsu and Campbell give talks about the DB hardware. If it could win 90% of the games running at 1/10th the normal speed for one chip, what does 480 chips at full speed get (hint: 4,800 times faster). Would you think it might have a pretty easy time with today's programs? quads or 8-way boxes as you want? >Do pigs really fly in your neighborhood? :) No, but they must fly in yours... In my neighborhood, DB still flies... :) > >Best Regards, >Chris Carson
This page took 0.03 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.