Author: Christophe Theron
Date: 08:57:56 02/01/00
Go up one level in this thread
On February 01, 2000 at 11:06:30, Dave Gomboc wrote: >On February 01, 2000 at 00:43:26, Christophe Theron wrote: > >>On January 31, 2000 at 02:24:16, Dave Gomboc wrote: >> >>>On January 30, 2000 at 21:39:16, Christophe Theron wrote: >>> >>>>So all that is needed is that the SSDF guys drop Chess Tiger DOS and start >>>>testing the Rebel-Tiger instead. >>>> >>>>I cannot be sure the Rebel-Tiger would be on top again, but I am sure it plays >>>>significantly better. So... >>> >>>This is exactly what I thought would happen, even though you previously said >>>that SSDF could start testing with Chess Tiger DOS because things wouldn't >>>change for Rebel-Tiger's release. You really should take credit when it is your >>>due, Christophe! ;) >>> >>>Dave >> >> >>Come on, didn't you follow what happened? > >Yes, I did. > >>Rebel-Tiger, commercial version, is by default EXACTLY the same engine than the >>DOS engine that has been tested by the SSDF, with the only exception, that has >>been checked by the SSDF and mentionned in their comments, that it computes >>somewhat faster (about 8%). The difference only comes from the compiler used to >>build the commercial engine, it is not a matter of optimized code. >> >>But while the engine was tested by the SSDF, I was still working to improve it >>(I had less than one month to do that). As I wanted to keep my promise that the >>commercial version would play exactly like the version tested by the SSDF, the >>improvements have been included in the commercial version but they are not >>activated by default. I wasn't even sure when the master CD went to duplication >>that these new settings were better or not! > >I think it is good that you kept your promise. > >>The engine has been "frozen" by the end of November, so I had enough time since >>that to check if the new settings are really better or not. >> >>They DO improve the program's strength. So now there are 4 good reasons for the >>SSDF to test the commercial product: >> >>* it is somewhat faster >>* it can use bigger hash tables >>* it has improved algorithms >>* there will be soon an opening book patch >> >>I don't feel like I have betrayed anybody. I have kept my promise to the SSDF, >>and my customers do get the latest and best engine I have produced. >> >>What do you say about Fritz6 followed several weeks after by Fritz6a, and >>Junior6 followed by Junior6a? > >I think they shouldn't be tested separately either, and my understanding is that >they won't be. Doesn't your original statement imply that there should be a >separate entry for Rebel-Tiger than for Chess-Tiger on the SSDF list? That is >certainly how it reads to me. Yes, that's what I mean. Rebel-Tiger with the new improved setting is, I estimate, 20-25 elo points stronger than Tiger 12.0 DOS. If the SSDF melts the results of Tiger 12.0 DOS and RT+new_settings, we are not going to see much change in strength until they play more than 400 games with it. But it happens that these 20-25 points is the difference that has been measured between Tiger 12.0 DOS and our new #1, Junior 6. So this strength difference is of significant importance. I don't see why asking to test them separately is so wild. Look at the SSDF list, you'll see that several programs of almost equal version have been tested as separate entities: Hiarcs 7 and Hiarcs 7.32, Nimzo 99 and Nimzo 7.32 for example. Christophe
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.