Exchanging repertoires, and best ways to create repertoires for over the board games

[Jonathan003]

Has someone ideas how I can split a big repertoire file to small parts, (maybe max 40 games for one part). And give logical names?

I do it this way. I open the big repertoire file with Hiarcs Chess Explorer, I go to players. Then I can see what openings are played. I export the games and create new databases with the opening names. The problem is that for exemple with the repertoire database from the white games of Anand. There are still more than a 1000 of games for the opening: Sicilian Defence, Scheveningen (B80-B85), only. I changed the results to drawn games and created bin books. So I want to further divide the games. 
What I do is open the pgn database Sicilian Defence, Scheveningen (B80-B85).pgn in chessbase 15, I convert to cbh, and create a search booster.
Than I go to players, and choose to prepare against black for the opponent of Anand, (I created fictive names to make this work).
That way I can further divide the more than 1000 games in a logical way.
I call it Part__001_, Par__002__, ect...
Finally if the pgn files are still to big I use gameSplit included in 40H-PGN-20181110 utilities.
I divide the pgn files further so the have a max size of 15 kB, so I have small pgn files I can easy study with chesstempo opening trainer beta.
Dos anyone have any ideas how I can achief this in a more convenient way?

If you want to know how I create repertoires send me a private message.

[Jonathan003]

I'm looking for a way to to determine which grangmaster games fit's most my style to base my repertoire on. I have made a repertoire based on the games of Magnus Carlsen. It is tempting because he is the world champion. But maybe there are other top GM's hose playing style fit's more my own playing style.

Here is what I have in mind. I would create repertoires from the top 10 GM's. Than I use the player dossier function in chessbase 15 to see the most played positions for white and for black for the GM.
Then I would like to create some system to know what positions I like the most, to know with what GM I come out.
I'm not a programmer so I don't exactly know how to do it. But you got my Idea I hope.

I already dit the quiz here: http://www.chesspersonality.com/

I came out with Magnus Carlsen.
It is a nice test but it has flaws. So I look for beter ways to do it. I want to make repertoires for other chess players to in the future.

I would like to hear some ideas how to do it please.

[Ovyron]

I think Chess Openings Wizard would work for this. The procedure would be as follows:

-Use some tool to identify the ECO codes of the games played in the PGN, and tag them correctly.
-Separate the games into white (the player is white) and black (the player is black) so their repertoires aren't confused by opponent's repertoire.
-Split the games in relevant ECO chunks (how to do this optimally is hard, but you can get creative on this part. Usually the Najdorf takes several of these chunks by itself.) How many chunks do you use depends on how "fine-grained" you want the repertoire analysis to be.
-Create a book in COW for each of those chunks.
-Backsolve each book so it shows the number of positions in the book (you can write that in the comment of the opening position for the book.)

And that's it.

Once this is done, all you need to do is comparing what is your biggest ECO chunk (representing your repertoire) with the chunks of the other players (say, B97 is the file with most positions, you sort the players by how large is their B97 chunk.) After you have found this, you can repeat the process splitting these relevant ECO position into smaller chunks.

As a bonus, you can import games truncated up to some move, say 37, then have an engine analyze the leaf nodes of all the games in the relevant chunk, and after backsolving you will be able to tell with what moves that player fared better, not just what moves they played the most. So that, if your pet line was the pet line of some master, but they didn't fare well with it, you can find a previous move that deviates from that, that worked better for them (even if it wasn't played as much), and try it out, that being a way to improve your repertoire.

[Jonathan003]

Thanks for the tip. I have Chess Openings Wizard so I will try to follow the steps.

Once this is done, all you need to do is comparing what is your biggest ECO chunk (representing your repertoire) with the chunks of the other players (say, B97 is the file with most positions, you sort the players by how large is their B97 chunk.) After you have found this, you can repeat the process splitting these relevant ECO position into smaller chunks.

That's in case if I already have a repertoire. I don't really have a repertoire, I have studied very little opening's theory.
I would like to categorize a player into 'type'
based on their opening repertoire.
To see if this could be helping me choose a player to use as
'model' reference when choosing which opening(s) to play.

Something like this quiz here: http://www.chesspersonality.com/

I think it is tricky to do a good job of it though.

[Ovyron]

That's in case if I already have a repertoire. I don't really have a repertoire, I have studied very little opening's theory.

That's actually very good, opening theory can be actually hurtful and irrelevant at some levels, that's why I don't teach any of it to my students, letting them discover what are the positions that are best to them and that make them feel comfortable by themselves, pointing out general opening principles and moves to avoid (b4/g4/b5/g5 seem to be favorite moves from people, for some reason?) Studying opening theory should be done much later, and that's when you already have enough played game that a repertoire exists already (where a "repertoire" is just what you like playing against what your opponents play.)

At some levels chess improvement is about creativity and learning tactical pattens as much as possible, even, setting up traps (the kind of traps that will be abandoned when you're stronger and your opponents don't fall into them, when you have nothing to do in a position setting up a high level trap will be the way to win games. Traps aren't hoping that your opponent doesn't see something, but creating a scenario on the board that tests your opponent's knowledge.) Opening theory is nothing but memorization, so it should be left for later because you have to understand the moves you are making in the opening, and once you do that you'll play the best opening moves by yourself without having seen a Master play them, or reading them from a book, and you'll play the best opening moves even when there's no opening theory about them, because the opponent played a move you've never seen before, and you are in novel territory.

Otherwise, you do have a repertoire, gather all the PGNs you have played, import them into a book, that's your repertoire.

----------------

Tacking this from the other side: there's no way to extract playing style from openings played, this has been a problem plaguing computer chess for years on end, that we still don't know the style chess engines play at, that all the perceived style of them has been subjective, with no objective way to decern them.

The difference between engine games and human games is sample size, with engine we have hundreds of thousands of games, yet no way to take a look at them and see what style the engine is using. With human games it must be much harder, because if this was easy, a single game would be enough to know a player's style.

Creating player "types" requires knowing the style they use (say, a type would play in some style in closed positions, and in another in open positions, etc.) so if you managed to find a way to separate players into types by looking at their games, you'd be the first and I think your findings would be very worthwhie.

If you did this, the process could be reverse-engineered, and someone could program a chess engine with the same type and style of some human player, or anybody, for that matter, I guess that's something I always wished to see (imagine a program that can play like Paul Morphy, and we'd get to see how he'd handled chess opening theory today.)

[EroSennin]

The more extreme a style is, the easier it's to find the pattern. Some people sac something almost in every game. Some try to avoid exchanging queens and some play endgames in all games.

[Ferdy]

Has someone ideas how I can split a big repertoire file to small parts, (maybe max 40 games for one part). And give logical names?

I do it this way. I open the big repertoire file with Hiarcs Chess Explorer, I go to players. Then I can see what openings are played. I export the games and create new databases with the opening names. The problem is that for exemple with the repertoire database from the white games of Anand. There are still more than a 1000 of games for the opening: Sicilian Defence, Scheveningen (B80-B85), only. I changed the results to drawn games and created bin books. So I want to further divide the games. 
What I do is open the pgn database Sicilian Defence, Scheveningen (B80-B85).pgn in chessbase 15, I convert to cbh, and create a search booster.
Than I go to players, and choose to prepare against black for the opponent of Anand, (I created fictive names to make this work).
That way I can further divide the more than 1000 games in a logical way.
I call it Part__001_, Par__002__, ect...
Finally if the pgn files are still to big I use gameSplit included in 40H-PGN-20181110 utilities.
I divide the pgn files further so the have a max size of 15 kB, so I have small pgn files I can easy study with chesstempo opening trainer beta.
Dos anyone have any ideas how I can achief this in a more convenient way?

If you want to know how I create repertoires send me a private message.

I am developing a chess gui that would allow a user to save a game in white or black repertoire. These repertoire pgn files can then be used in CPT or COW.

You can load 2 polyglot books as user references plus an Adviser engine for other move recommendation/ideas. You can add your ideas in the comment, and it will be saved in the game.

Although I played as white, if I find that blacks play is good and like it, I can actually save this to black_repertoire.pgn.

[Jonathan003]

The more extreme a style is, the easier it's to find the pattern. Some people sac something almost in every game. Some try to avoid exchanging queens and some play endgames in all games.

I changed my mind.
I thing to define a player as for example a positional player like Karpov, or an aggressive player like Bobby Fisher is a thing of the past.
Top players al study opening theory and they will play wat the position demands.

There is stil some difference in playing style but not like it was in the old days.

[Jonathan003]

There is an option in Chessbase to display unplayed transpositions in a CTG book, but when you play engine against engine matches it seems like Fritz ignores these moves.
I want Fritz to play these unplayed transpositions to, if they are good moves.
If I make a tree from the same games in Aquarium, and I export
the tree to EPD format, and then I import this EPD file back to the same tree, and choose the option 'generate moves'. This tree also contains the moves that are displayed as unplayed transpositions in the CTG book.

I have made a video capture of it so you can see what I mean.



I would like to be able to include these unplayed transpositions automatically in 'opening lines' in Lucas Chess. Or some other way to convert to pgn with these unplayed transpositions included.

Maybe if there would be a way to include these unplayed transpositions in a polyglot book somehow. And then import all the best moves from the bin book to the opening lines.
Because I think it is useful information.
If I make a repertoire based on the games of Magnus Carlsen for exemple.
I can always delete variations I don't want, but it is more work to find good variations to add.

I hope someone can help me with this.