Author: Kurt Utzinger
Date: 04:12:29 11/08/05
Go up one level in this thread
On November 08, 2005 at 05:42:08, Uri Blass wrote: >On November 08, 2005 at 05:24:09, Kurt Utzinger wrote: > >>On November 08, 2005 at 03:12:20, Uri Blass wrote: >> >>>On November 08, 2005 at 02:01:56, Sandro Necchi wrote: >>> >>>>On November 08, 2005 at 01:39:25, Kurt Utzinger wrote: >>>> >>>>>On November 07, 2005 at 13:32:29, Sandro Necchi wrote: >>>>> >>>>>>Hi, >>>>>> >>>>>>I think I have improved Columbus' egg 9g with the latest 9L (Lyon) version. >>>>>> >>>>>>I am testing it against "Fritz friend" (Fritz 9) at 90 minutes game on 2 equal >>>>>>AMD XP 3000+, so similar to SSDF on A1200 >>>>>> >>>>>>Columbus' egg 9L + Shrddr10.bkt prc (preliminary c version) - Fritz 9 = 7 - 4 >>>>>>(+5, =4, -2) 63.6% TP = +97 Elo >>>>>> >>>>>>P.N. >>>>>> >>>>>>1. Shredder10 prc is in std. mode, so not the strongest mode! >>>>>> >>>>>>2. The improvements on Columbus' egg 9L are mainly strategical in the first >>>>>>part/middle game as the new settings are more refined. >>>>>> >>>>>>If this version will score very well after several games I will post the >>>>>>settings. >>>>>> >>>>>>Ciao >>>>>>Sandro >>>>> >>>>> Hi Sandro >>>>> You are very courageous to conclude something >>>>> out of 11 games only. In our Fruit 2.2.1 test >>>>> Fritz 9 lost in the first 10 games with 2,0-8,0 >>>>> but was almost equal after 50 games. In my >>>>> Nunn2 match, Fritz 9 won the first serie vs >>>>> Fruit 2.2.1 with 8,0-2,0 only to loose the >>>>> second serie with 3,0-7,0 and from such experience >>>>> I can't take matches under 50 games seriously. A >>>>> further good examply of how things can go/vary >>>>> to be seen at my comments re our RK 2004 tournament >>>>> http://www.utzingerk.com/rk_2004_english.htm >>>>> Regards >>>>> Kurt >>>> >>>>Hi Kurt, >>>> >>>>sorry but I said "I think", not I am sure. So you cannot say that I concluded >>>>something, however I made several tests before starting the match looking this >>>>setting on several positions, so I was not relying on this score, but on a >>>>overall judgement. >>> >>>Did you look at games from CEGT? >>>I guess that the games can give you productive knowledge about the mistakes of >>>colombus egg 9g so the question is if you found that the new setting can avoid >>>significant part of the mistakes of 9g >>> >>>Uri >> >> >> Hi Uri >> You are surely aware that testing is a very >> hard job. If a new setting can avoid the >> mistakes of 9g this does not necessarily >> mean that the new setting will not produce >> other - so far not known - mistakes -:) >> Regards >> Kurt > >I know it but I think that good testing should start from mistakes of previous >personality because if it even does not avoid significant part of the mistakes >of previous personality the chances that it is going to be better is very small. > >My opinion is that playing games should be only last part and not first part >when you think that you may have an improvement. > >Uri Hi Uri As far as I can see we have different opinions here. From my own experience with many CM-settings, I would tend to say that the "correctness" of a new setting is more reliable if tested with many games than by (only) finding better/stronger test moves. Regards Kurt
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.