PDA

View Full Version : Fritz openings book: design flaw?



Miguel
25-06-2007, 08:29 PM
I think the way that transposition possibilities are managed in the Fritz openings book is perhaps flawed. Specifically, the importance of moves that transpose is overemphasised, since in any given position move probability is based on the frequency of the resulting position rather than the actual move frequency. While this may be okay for engine-engine games, for human training and practice it can be bad: it can lead to "over-training" in positions that result from transposition and "under-training" in other important lines. (Note that this problem is not specific to Fritz10.ctg, but also happens with user-created openings books.)

An example which emerged in my practice came after the moves 1. e4 c6 2. Nc3 d5. This Caro-Kann position occurs in 2,567 games from the Fritz 10 database. Of these games White continued with the Two Knights Variation in 1,751 games (68.2%) by playing 3. Nf3, and only in 369 games (14.4%) did White transpose into the Open Variation with 3. d4. So it's clear that 3. Nf3 is the regular move, yet because the Open Variation (through the normal move order: 1. e4 c6 2. d4 d5 3. Nc3) is so popular the openings book skews the probabilities such that the Two Knights Variation is rarely played: 3. d4 - 76.7% and 3. Nf3 - 23.3% (with "normal" book options).

Other examples:
1. d4 e6 - The main moves should be 2. c4 and 2. Nf3, but instead the main move given by the openings book is 2. e4;
1. Nf3 Nf6 - The main moves should be 2. c4 and 2. g3, but instead the main move given is 2. d4;
1. c4 Nf6 - The main move should be 2. Nc3, but instead the main moves given are 2. d4 and 2. Nf3.

These can be confirmed with the Shredder Chess Opening Database [1].

I have mentioned this problem to ChessBase support (two weeks ago) and to ChessBase Workshop columnist Steve Lopez (mid-2006), but so far I have received no replies.

Is my complaint justified, or am I making a mountain out of a molehill?

[1] http://www.shredderchess.com/online-chess/online-databases/opening-database.html

- Miguel