rosgnl12
User
Junior Member
| Posts: 40 |   |
Karma: 0
|
re:Wie funktioniert die Computer-Eröffnungsdatenbank (allgemei - 2006/09/19 05:13
The moves of the games you import into your openings book are added to the openingsbook. Fritz also looks at the results of the games, and the possible Elo of players. With that it adds those statistics to those imported opening moves. As far as those opening moves already are present in your openings book, the statistics are updated.
If there are 100,000 games with e4, and 45,000 ended in a win for white,
20,000 in a draw and 35,000 in a win for black, you'll see those percentages reflected in the statistics. In the small statistics windows below you can see the actual numbers 45% / 20% / 35%; in the move list you'll see a result of 55% for e4 (wins + half the draws). This figure is always from the viewpoint of the side to play.
Using a high quality database with recent games gives you a view of current opening knowledge, and e.g. new refutations. Adding them to the opening tree adds the stats of those games and of their opponents and updates the value of the results of those individual moves.
Fritz doesn't analyse the game, it only looks at the result and the Elo
0f the opponents. So it's yes and no. If you see a bad scoring percentage for white with e4 and you see average Elo of those white players is under 2000, you bet you just added a lot of games by weak white players.
In reality this doesn't happen. When adding games to the database you only use either carefully selected games for a specific opening or a high quality database with games without those kind of blunders. If bad gamer are imported in your openings book, you're the only one to blame for it.
The Fritz help files have quite some information about this theme.
Using those stats can be a tricky business. It depends on a lot more than just statistics. Not every game is imported to the same extend as the other.
All games might be imported to move 20. You browse through the tree and arrive at move 20. You see the statistics giving you a 66% chance with
20.b6 from four games. So you (or Fritz) play the move. Suddenly you see the only response for black: 20...a5 and notice a 100% _win_ chance for black with that move with only _one_ game. Reason: in three out of those four games the game was ended at move 20.b6. One win for white because of time forfeiting by black, two draws because the players were happy with their tourney result and decided to draw, and one game which gave a loss for white after 20...a5 21. b7? and Qxe2#
That's why after a certain point Fritz offers you the possibility to
"grey those moves out", meaning that there's no statistical relevance to the moves anymore.
_Annotating_ the moves in the tree (tournament move yes or no, good or bad move etc) is done manually by the book builder or comes from a learning file by the chess engine, marking moves it often loses with ad bad or not preferrable..
---------
Before God we are equally wise and equally foolish.
Popular posts by rosgnl12
Don't delete multiple icons in v...
Chess engines & programs
Chess Tiger 15.0
|
post new topic