Author: Ed Schröder
Date: 16:49:36 02/10/01
Go up one level in this thread
On February 10, 2001 at 15:46:30, James T. Walker wrote: >On February 10, 2001 at 13:36:06, Howard Exner wrote: > >>Rebel's play in these two games is agressive. >>In Fritz's win I'm curious if Fritz shared Rebel's eval just prior to this >>point, where Rebel played Qb6. Rebel was quite happy with its position prior to >>Qb6. What did Fritz expect here, Qe5 maybe? >> >>[D]4r3/1p2r1kp/pq2PppN/2RRnP2/1P6/7P/P2Q2P1/6K1 w - - >> >>Rebel's win again was lively, playing Rxe5 to cap off a nice build up. >>Other moves seem good also. >> >>[D]5k2/3R4/p3P2p/q5b1/1pKN1pp1/1P6/2R5/8 b - - >> >>Do you feel the autoplay hurts Rebel significantly or is there too little data >>out there to conclude much? >> >>Either way it appears once the Windows version is available that the dos >>autoplayer will be defunct. Ed, what GUI is in the works for Century 3.X ? >>Something new, original Rebel, Lex's Chess Partner, or some hybrid? > > >Hello Howard, >At move 46 (second diagram) Fritz had a score of -2.50 meaning it was losing >badly. The score slowly dropped with each move until 50. Nd4 (+0.88). >I honestly like the way Rebel plays most of the time. It seems like >occasionally it is completely blind to what is happening. Examples are the mate >in 6 posted here a few days ago. Rebel could not find it after 18 plies of >search. I have seen Rebel lose games which should have been easily drawn. At >the same time it sometimes beats Fritz and Junior easily. It's too early to >tell for sure if the auto232 is hurting Rebel because I only have 44 games. But >my feeling is that it is hurting but I can't tell to what extent yet. I can >only say that the manual results are already taking a lead vs the auto232 >results so is it a coincidence or is it already showing up? By the way, thanks >to Mogens, I found another game which I played with Fritz not using the >"Optimized" book settings which Rebel drew as white. I replayed the game and >Rebel won so the score against Fritz in manual games ended at 11-9 favor Fritz >6. The auto232 games are not affected by my errors since the optimum book >setting is set automatically. It was 14-6 favor Fritz. Also I changed to G/30 >minutes to go faster and in the first 3 games vs Junior (manual) the score is >2.5-0.5 favor Rebel. So it's easy to understand why I feel there is a definite >cause-effect working with auto232. I have to sit and watch all the games played >manually and can't help but see how Rebel plays. >Jim >Jim I am glad somebody is doing such a massive check. In the months before Paderborn 1999 I have done the same, about 30 manual 40/40 games and the results were about 15% higher than under the autoplayer (measured over 400 games). I realize that 30 games is not much but I had no time to play more manual games. From Rebel 10 till Century 3 I remember the bad auto232 result in almost every autoplayer tournament this in sharp contradiction with its manual performance which were good hardly without exception. In the meantime my suspect to the DOS autoplayer software goes a bit further. It would not be unlikely at all that ALL auto232 DOS drivers had the (same?!) bug. I still have an example of a Mchess-Rebel game where Mchess8 in a good position gave away a full rook which (of course) in no way is reproducible. Also when Christophe was autoplaying Tiger with the DOS driver it frequently happened the autoplayer was hanging in the permanent brain. Those who have tested the DOS edition of Tiger (see the SSDF list) can confirm this typical behavior only happening during the autoplayer. As Christophe explained to me any incoming move is checked with the contents of the hash table. Of course the move must be present and if not something is very wrong with the contents in the hash table which indeed would force the program to hang. It only happened during the autoplayer and not in manual games, the "hanging" of Tiger simply disappeared moving to the Windows autoplayer code programmed by Lex Loep. I asked Christophe specifically if he made some changes, he said no and that the "hanging" problem in autoplayer games simply never appeared again. I know this is not sufficient evidence to put the whole auto232 DOS autoplayer into the trash including all the results from 1994 when it was introduced but the list of complaints is quite incriminating even more if you add-up a similar complaint from Cstal-1 (DOS) hurting the playing strength. Also Bob (Crafty) frequently has complained about the DOS autoplayer driver although he (to my best knowledge) never said the autoplayer hurted the strength of Crafty. Ed
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.