Author: Chessfun
Date: 08:18:49 07/24/01
Go up one level in this thread
On July 24, 2001 at 10:43:59, Harald Faber wrote: >On July 24, 2001 at 09:51:08, Chessfun wrote: > >>On July 24, 2001 at 01:14:39, Harald Faber wrote: >> >>>On July 23, 2001 at 17:25:55, G. R. Morton wrote: >>> >>>>I seem to have mixed impressions from some of the Info I've read. Opinions? >>>>Thanks in advance. >>> >>>SSDF *pretends* that GambitTiger 2 is stronger, >> >>They don't "pretend" their results and others suggest it is so. > > >This is twisting words, you know what I meant. I had no intention of twisting anything. I simply took what you had written at face value. >> but see that they test >>>ChessTiger 14 in the Chessbase-version >> >>As I recall from beta testing this was Christophe's idea. > > >I know. >>which has NO ADJUSTED opening book in >>>opposite to the Chesspartner-version. >> >>While it may not have an adjusted book it has a better book learner. > > >But it takes some games so that learning works. In the June SSDF list Tiger 14 CB has played 308 games. So now they continue using it, isn't this enough games for that learning to work? >> And the difference between them in SSDF is >>>marginal, >> >>That is true, as it also likely is, if you compute the ratings from your games. > > >Sure, never meant anythng else. > > >> so, in spite of less games I think that my tourney says more the >>>"truth" >> >>huh, did I miss something. You play a few games and that says more of >>the truth. This dispite the fact that you previously posted that if you >>played more games you would likely get totally different results. > > >Please read carefully. I *probably* get different results if I'd replay the >games. BUT it also might be possible that I get the same percentage result after >500 games as I did after 10 games. This is indeed a stretch. >I might get a 6-4 after 10 games. 10 more >games give a 4-6, but after 500 games it steps back to 300-200. >And BTW, if you'd take a look at the games, you'd see what is/was going on on >the board. Again this is still stretching the imagination to assume that your 10 games would translate into 500. Also I don't look at games, I "throw bones"...;-) >>than SSDF because in my comparison both Tigers play the same opponents >>>(did they in SSDF? I don't think so) >> >>Again as you have yourself said, the limited number of games your playing >>leads even you to believe that with more games you would have totally different >>results. So the fact your playing the same opponents is meaningless. > > >For you, not for me. It is one important point for real comparison. The only >condition for a 100% comparison I couldn't realize is that I am not able to >replay the same openings with the autoplayers. The ONLY way to *compare* two >programs is to play the same opponents with the same openings. Part two is what >is not fulfilled in my tourney. But I don't mind, I see enough. And all what is >written in the 5 lines above is NOT the issue of SSDF. This issue has been debated endlessly. First with Nunn1 then Nunn2. However even those openings are likely more suited to specific programs than others. In your case playing 10 games would give 5 opening selections, clearly this isn't enough. Sarah.
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.