My new book is out: Noomen.ctg

[Frank Quisinsky]

Hi Ferdinand,

I checked that too last week ... (dublicate games, most important if you need many games).

After my short test I have 28% double openings with the book by Jeroen (after your stats 30%).

This is a lot ... and I have with around 3000 games with the book by Jeroen 280 ECO codes only.

With my opening book after around 3.000 games ... 140 double openings = 4.5% with around 430 ECO codes (many ECO codes are very rarely, not often played in my book but all can be play). Sure that rarely lines will be played rarely with the book by Jeroen too.

4.5% ... now 24 months hard work in optimation the priorities between the lines! I think very good but not good enough ... 3% is possible ... think so!

For a ChessBase Book a normely stat the new Noomen book have. Sure, that Jeroen added good lines but again ... not enough if you need many games. To many double lines.

One of my stats (and this is the problem with my book):
Between 50 games matches = 25x white and 25x black I have after a test-run around 10 double opening lines or 30 double lines undo end of move 6. That is in each book really a problem. I have 43.000 playable lines in my book ... 99% are equal.

After 225.000 FCP games 10.5% of the 45.000 lines of good know GM and correspondence games by strongest players are deactivated (bad lines, games not started with equal positions). Also this one are normely stats.

It's a lot of work to select out such lines and looking for ... all 500 ECO codes should be play. Means not easy to find to rarely lines good playable equal examples.

Best
Frank

Only one ChessBase Book (I collect all) have good stats ... the Hiarcs books have here the best statistics and I found not many not playable lines ... last ChessBase Version = 17% double lines after around 3.000 games.

[Ferdy]

I checked that too last week ... (dublicate games, most important if you need many games).

I expect high duplicate numbers because both engines use the same book - noomen.ctg. Also my book variety of play setting used in sf 7 was not actually full.

[Frank Quisinsky]

Hi Ferdy,

yes ...

Example:
Load the Komodo 10 test-run games in your selection "download games by Players". Now give all 57 players the name Human.

Hint:
Search bad lines in the 2.800 games (you will not find it).

As next ...
truncate 20

Now search the doubles lines undo move 10 (20 half moves).
Result: 136 double lines after 2.800 games.

I do that after each test-run and optimate my book with the games I have.



Best
Frank

More and more bad lines found (during a test-run ... directly replayed and deactivated in my book) ...
But the secret is with such optimation (set priorities) that with lesser lines (after I filter out bad lines) more lines can be play (set better priorities).

For around 3 months ... around 150 double lines after a test-run ... at the moment around 130-140! ... a very hard work!

And ...
All the tools you created helps me a lot.
I think my book is over optimated ... but each day I am thinking ... can be better and better and I am working on it around 1 hour daily since years for get such stats.

[Jeroen]

Hi Peter,

Thanks a lot for your kind comments.

Regarding the "influence = 0" setting: throughout the book I have used weights to alter priorities of moves. In testing I have noticed that a high "influence" setting will lead to less variety in chosen variations, even if the slider for variety is maximised. This is an unwanted phenomenon, that will not take place with influence of learn value = 0.

Of course feel free to experiment with the learn values. For me it was necessary to get the desired variety and to stop the probability that some lines will be played very seldomly.

Kind regards, Jeroen

[flok]

I could take a look at that code? Please ask if it is allowed to be made in an open source bsd or whatever license.

[PK]

@Frank, I made a Polyglot book from Your pgn. You set all games as draws, so all the percentages depend on move frequencies - and some of them are quite strange, like over 70% for 1.d4.

[Jamal Bubker]

Thank you very much Jeroen for releasing and providing us this top high quality chess openings book