Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Results of John Merlino's Chessmaster survey?

Author: John Merlino

Date: 08:46:27 11/30/01

Go up one level in this thread


On November 30, 2001 at 05:16:33, Marc van Hal wrote:

>On November 29, 2001 at 19:30:01, John Merlino wrote:
>
>>On November 29, 2001 at 17:01:07, Mike S. wrote:
>>
>>>On November 29, 2001 at 16:38:41, John Merlino wrote:
>>>
>>>>On November 29, 2001 at 13:55:40, Roy Eassa wrote:
>>>>>Sorry if I missed them, John, but do you have any results or statistics on the
>>>>>results of the Chessmaster survey you posted awhile back?
>>>>
>>>>I didn't get very much useful data because only 24 people responded, and most of
>>>>the answers were either almost unanimous (and unsurprising) or completely
>>>>random. (...)
>>>>
>>>>That's really about it. Not much of use, except as, perhaps, an amusement.
>>>
>>>What about the "tasc database as an opening book" issue, in other words, to have
>>>Chessmaster use the "database findings" like an opening book?
>>>
>>>I've seen then and when CM being out of it's book and at the same time the
>>>database findings still provided useful move statistics from many master games.
>>>CM could use these like an opening book (maybe as an additional option, or
>>>adjustable). IIRC, that was how the tasc base program used to do it btw when
>>>playing a game.
>>>
>>>Now, these infos are displayed for the user, but CM itself doesn't utilize it
>>>for it's own play.
>>>
>>>I think this could really be an improvement, and shouldn't be too difficult to
>>>implement (as you have the database findings feature already). Furthermore, the
>>>benefit could even be bigger when the game collection is enlarged and updated
>>>either by you, or even by the users themselves. Or they could use special
>>>databases for that purpose, i.e. computer chess games when The King plays
>>>against another engine.
>>>
>>>Regards,
>>>M.Scheidl
>>
>>This has been done, in a roundabout way.... The new opening book (thanks to a
>>much larger format) has been created using CM8000's database (specifically, all
>>games where both players were at least 2400). All moves up to 40 plies from
>>these games were stored, and then any move that was made only once was removed.
>>The resulting book has over 280,000 moves (more than seven times the size of the
>>previous book).
>>
>>So far, testing has shown that this book is no worse than the CM8000 book. ;-)
>>
>>jm
>
>Games of 2400 versus 2400 can be prety bad though
>Then again if you know grap 1 game from the SSDF list you already can win from
>Chessmaster when it plays 1.d4
>A game against Tiger
>
>Regards Marc van Hal

That's why all moves that were played only once were removed. It (for the most
part) eliminates blunders.

jm



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.