Author: Chris Carson
Date: 15:19:53 05/13/01
Go up one level in this thread
On May 13, 2001 at 17:18:13, Chris Carson wrote: >On May 13, 2001 at 17:08:55, Joshua Lee wrote: > >>500 games is not an arbitrary number, this is the number of Games at a minimum >>on the SSDF list where the ratings were steady +- 30 instead of +- 50 or 60 so >>there goes one of your assumptions the other I know you are not a part of SSDF, >>That's number two and as far as your example goes those GM's have most likely >>played more than 500 Games so 1 tournament proves nothing ask around i am sure >>the people in here will agree that you can't tell much by just 1 tournament. >>But you can believe what you want. Like i said Junior is too High there is no >>way for example that it can be so close in rating to 9x faster hardware if you >>believe that i would worry think about it it is like saying junior on a 286 is >>100pts below junior on 450Mhz????????????????????? Hello. Like i said before it >>is only 1 tournament but still you can tell the ssdf results are wrong just by >>puting this to mind , junior has played withing the ssdf rating pool which >>includes some games against humans still with a decent margin of error and has >>recieved a rating of 2523 which is almost 100pts too high still on 450 which is >>9 times slower than an 8 way 700Mhz machine which is trippling the speed twice >>or 90 x 2 180 like before how much memory did deep junior use this does make a >>difference. I believe SSDF is wrong and i am sure most people can do the >>research and find out the same if they don't believe me. FYI the examples from >>tony's page were not used in the ssdf's rating pool so only 183 games are being >>used but you would be right about one thing if they used those games they would >>have the so called arbitrary number (500 games)i mentioned but ofcourse you >>didn't say that. > >I stand by my previous posts. :) > >Best Regards, >Chris Carson Joshua, The calibration games on Tony's page are the 337 (plenty) Human vs programs the SSDF used. I think they also used the additional 183 (plenty) to re-calibrate, so a total of 520 (plenty of games). Also, the following are the ratings: 1. DJ6 8x700 H vs C 2702 2. DJ6 450MHZ SSDF 2595 3. DJ6 450MHZ H vs C 2519 The bottom two SSDF and actual DJ6 vs human games are pretty close and expected (similiar to how humans perform at different tournaments as I have shown in previous posts). Only 76 points difference not unresonable a difference, humans have a larger difference than this on a regular basis, just look at any ot the TWIC issues or the next tournament result. The 8x700 is not 9 times faster than the 450 and there is almost 200 poinst difference (100 from the SSDF). This looks very reasonable to me. So again, I stand by my previous posts, but believe what you want. :) Best Regards, Chris Carson
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.