Author: Roy Eassa
Date: 10:28:06 01/18/02
Go up one level in this thread
Albert, I understand just what you are saying. However, my understanding was that the problem in Chess Tiger was essentially a bug in that program such that it will not use many of the other, actually necessary (non-trivial) 5-man TBs correctly if these "obvious-win" ones are not also present. Once it starts looking into TBs in a given variation, it is not able to leave TBs and use normal calculations any more in that variation. Thus, when there's a variation in which it CAN achieve an endgame where it's ahead by vast amounts of material (say by promotions), it won't even play the entire variation, which may evolve from a closely contested position, due to this bug. Thus it may fail to win certain winnable endgames. This in fact occurred twice, as I recall, in its recent match against Gandalf. On January 17, 2002 at 19:50:03, Albert Silver wrote: >On January 17, 2002 at 17:23:11, Roy Eassa wrote: > >>On January 17, 2002 at 17:18:24, Albert Silver wrote: >> >>>On January 17, 2002 at 14:21:36, Roy Eassa wrote: >>> >>>>But I have a question: if the complete 3-4-5 set requires 11 CD-ROMs of 650 MB >>>>apiece, how can it fit on only 9 CDs of 700 MB apiece? >>>> >>>>Even if the last of the 11 CDs is only 1% full (i.e., 10.01 CDs-worth of 650 MB >>>>apiece), that would still be more than 9 CDs of 700 MB. >>>> >>>>Is this really the full set? >>> >>>A fair question. No, there are a few that were deliberately removed in order to >>>save space as they had absolutely no practical value. Truly. For example: >>> >>>King + 3 queens vs. King >>>King + 3 rooks vs. King >>>King + 2 queens + 1 rook vs. King >>> >>>This reduced the number of files from a maximum of 290 files to 240 files. >> >> >>I was under the impression that Chess Tiger was weaker using a subset (even a >>very common-sense subset) of the 5-man TBs than not using any 5-man TBs at all. >>I guess one could download the missing TBs if one knew exactly which ones they >>were, just to be on the safe (and complete) side. > >I understand your concern, and remember perfectly what you are referring to, >however I'll add the following points: > >- The TBs that might have caused problems were from a 3 CD set and truly >incomplete. It was missing many piece combinations however unlikely. This is a 9 >CD set where the only missing combinations are ones in which Tiger (or any >engine) would find a mate in a fraction of a second. > >- If Tiger were to reach one of the postions mentioned at any point of its >calculations, it shouldn't play any weaker. After all, it would be effectively >up some 20-30 pawns in its eval. > >- Tiger has undergone significant testing on the machines at Convekta in order >to test the sets as well as some other functions used in Chess Assistant and no >problems of any kind have shown up in this time. The engine-related functions I >speak of are use of the TBs by Tiger, and use of the TBs by the GUI to >complement an engine that doesn't support them internally. > >Personally, I wouldn't fill up more than a Gb of space on my HD to put in KQQQ >vs. K type combinations. > > Albert > >> >>Thanks, >> >> -Roy.
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.